public inbox for [email protected]help / color / mirror / Atom feed
[PATCH 5/5] Remove the GUC stats_temp_directory 110+ messages / 5 participants [nested] [flat]
* [PATCH 5/5] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 ++++----- src/backend/replication/basebackup.c | 13 ++------- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- src/test/perl/PostgresNode.pm | 4 --- 10 files changed, 14 insertions(+), 94 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index 9d4c000df0..bbae70221e 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1145,8 +1145,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 84341a30e5..6ad619de47 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6827,25 +6827,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index 990995c17b..2a2adaa0f7 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -195,12 +195,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index 1047c77a63..b47f8d084e 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index 514ea78a68..64e1374288 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -108,15 +108,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; #define StatsLock (&StatsShmem->StatsMainLock) diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index c2978a949a..e784088747 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -230,11 +230,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -265,13 +262,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 92c4fee8f8..7feb44c39d 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -194,7 +194,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -4085,17 +4084,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11368,35 +11356,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 5ee5e09ddf..0ba984b074 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -562,7 +562,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index c0bbf8a7d5..2c7c799ca9 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -31,7 +31,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ diff --git a/src/test/perl/PostgresNode.pm b/src/test/perl/PostgresNode.pm index 6019f37f91..a107683c0d 100644 --- a/src/test/perl/PostgresNode.pm +++ b/src/test/perl/PostgresNode.pm @@ -451,10 +451,6 @@ sub init print $conf TestLib::slurp_file($ENV{TEMP_CONFIG}) if defined $ENV{TEMP_CONFIG}; - # XXX Neutralize any stats_temp_directory in TEMP_CONFIG. Nodes running - # concurrently must not share a stats_temp_directory. - print $conf "stats_temp_directory = 'pg_stat_tmp'\n"; - if ($params{allows_streaming}) { if ($params{allows_streaming} eq "logical") -- 2.16.3 ----Next_Part(Thu_Jul_04_19_27_54_2019_888)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH 5/5] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 ++++----- src/backend/replication/basebackup.c | 13 ++------- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- 9 files changed, 14 insertions(+), 90 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index a73fd4d044..95285809c2 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1119,8 +1119,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index d383de2512..546a9c6296 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6734,25 +6734,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index 3092c7b5a3..809459e919 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -195,12 +195,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index cbdad0c3fb..133eb3ff19 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index aba43ce1dd..681b331a76 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -109,15 +109,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; #define StatsLock (&StatsShmem->StatsMainLock) diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 537f09e342..722f171f56 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -230,11 +230,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -265,13 +262,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index aa564d153a..e8771b8024 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -192,7 +192,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -3996,17 +3995,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11141,35 +11129,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index cccb5f145a..9983d6668c 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -554,7 +554,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index 5d15026db3..f4e3c9c9dc 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -31,7 +31,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ -- 2.16.3 ----Next_Part(Wed_Mar_27_16_36_25_2019_836)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH 5/5] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/replication/basebackup.c | 13 ++------- src/backend/statmon/pgstat.c | 13 ++++----- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- 9 files changed, 14 insertions(+), 90 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index a73fd4d044..95285809c2 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1119,8 +1119,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 8bd57f376b..79f704cc99 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6709,25 +6709,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index 5c1408bdf5..b538799ff6 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -197,12 +197,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index cbdad0c3fb..133eb3ff19 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index e30b2dbcf0..a567aacf73 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -231,11 +231,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -266,13 +263,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/statmon/pgstat.c b/src/backend/statmon/pgstat.c index 3849c6ec05..3108cf9c5e 100644 --- a/src/backend/statmon/pgstat.c +++ b/src/backend/statmon/pgstat.c @@ -89,15 +89,12 @@ typedef enum PgStat_TableLookupState bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; /* Shared stats bootstrap information */ typedef struct StatsShmemStruct { diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 62a07727d0..49123204c5 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -192,7 +192,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -3989,17 +3988,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11003,35 +10991,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 194f312096..fdb088dbfd 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -554,7 +554,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index 90464769db..f0804013db 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -29,7 +29,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ -- 2.16.3 ----Next_Part(Tue_Feb_19_21_40_07_2019_247)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v22 5/5] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 ++++----- src/backend/replication/basebackup.c | 13 ++------- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- src/test/perl/PostgresNode.pm | 4 --- 10 files changed, 14 insertions(+), 94 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index bdc9026c62..2885540362 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1146,8 +1146,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 7f9ce8fcba..a8bed31232 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6818,25 +6818,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index ea6aad4d1e..33ad2b8be8 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -195,12 +195,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index 1c19e863d2..2f04bb68bb 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index c0b20763b0..6b8025ad13 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -108,15 +108,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; #define StatsLock (&StatsShmem->StatsMainLock) diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 6aab8d7b5f..2eb49924b9 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -242,11 +242,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -277,13 +274,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 90ffd89339..753e30ebb7 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -194,7 +194,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -4072,17 +4071,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11352,35 +11340,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 0fc23e3a61..66f539c4bb 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -558,7 +558,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index c9fbcead3f..e9e18ed27a 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -30,7 +30,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ diff --git a/src/test/perl/PostgresNode.pm b/src/test/perl/PostgresNode.pm index 270bd6c856..c604c5e90b 100644 --- a/src/test/perl/PostgresNode.pm +++ b/src/test/perl/PostgresNode.pm @@ -455,10 +455,6 @@ sub init print $conf TestLib::slurp_file($ENV{TEMP_CONFIG}) if defined $ENV{TEMP_CONFIG}; - # XXX Neutralize any stats_temp_directory in TEMP_CONFIG. Nodes running - # concurrently must not share a stats_temp_directory. - print $conf "stats_temp_directory = 'pg_stat_tmp'\n"; - if ($params{allows_streaming}) { if ($params{allows_streaming} eq "logical") -- 2.16.3 ----Next_Part(Tue_Sep_10_17_58_58_2019_765)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH 6/6] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 ++++----- src/backend/replication/basebackup.c | 13 ++------- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- 9 files changed, 14 insertions(+), 90 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index a73fd4d044..95285809c2 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1119,8 +1119,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 8bd57f376b..79f704cc99 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6709,25 +6709,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index 339afb2de9..eb6e0eecdd 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -195,12 +195,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index cbdad0c3fb..133eb3ff19 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index 61b0bd161d..6f65d707ec 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -102,15 +102,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; LWLock StatsMainLock; #define StatsLock (&StatsMainLock) diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index def6c03dd0..58ba33e822 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -230,11 +230,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -265,13 +262,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 156d147c85..058d97075f 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -191,7 +191,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -3988,17 +3987,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11002,35 +10990,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 194f312096..fdb088dbfd 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -554,7 +554,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index cb70d00b8f..265ed4d118 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -31,7 +31,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ -- 2.16.3 ----Next_Part(Mon_Feb_25_13_52_14_2019_191)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH 5/5] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 ++++----- src/backend/replication/basebackup.c | 13 ++------- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- src/test/perl/PostgresNode.pm | 4 --- 10 files changed, 14 insertions(+), 94 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index 9d4c000df0..bbae70221e 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1145,8 +1145,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index c91e3e1550..b85f6f421a 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6827,25 +6827,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index 990995c17b..2a2adaa0f7 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -195,12 +195,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index 1047c77a63..b47f8d084e 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index 514ea78a68..64e1374288 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -108,15 +108,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; #define StatsLock (&StatsShmem->StatsMainLock) diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 57f17e1418..c854d7e193 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -230,11 +230,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -265,13 +262,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index fc463601ff..05af81ac47 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -194,7 +194,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -4085,17 +4084,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11368,35 +11356,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index cfad86c02a..29b7ebca46 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -562,7 +562,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index c0bbf8a7d5..2c7c799ca9 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -31,7 +31,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ diff --git a/src/test/perl/PostgresNode.pm b/src/test/perl/PostgresNode.pm index 6019f37f91..a107683c0d 100644 --- a/src/test/perl/PostgresNode.pm +++ b/src/test/perl/PostgresNode.pm @@ -451,10 +451,6 @@ sub init print $conf TestLib::slurp_file($ENV{TEMP_CONFIG}) if defined $ENV{TEMP_CONFIG}; - # XXX Neutralize any stats_temp_directory in TEMP_CONFIG. Nodes running - # concurrently must not share a stats_temp_directory. - print $conf "stats_temp_directory = 'pg_stat_tmp'\n"; - if ($params{allows_streaming}) { if ($params{allows_streaming} eq "logical") -- 2.16.3 ----Next_Part(Thu_Jul_11_16_40_59_2019_397)-- Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="statbench.pl" #! /usr/bin/perl use strict; use IPC::Open2; use Time::HiRes qw( usleep ualarm gettimeofday tv_interval ); use Errno qw/ECHILD/; my $totaldbs = 100; my $totaltbls = 100; my $ndbs = 50; my $ntbls = 100; my $loops = 20; my $chiter = 20000; my $chtrlen = 1000; my $nprocs = 90; my $trg_file = '/tmp/dbrun_pl.trg'; my $refresult_file = '/tmp/refresult.txt'; # title loops,ndbs,clnts,ntbls,childiter , xactlen,refprocs # testrun("A1", $loops, 1, 1, 1, $chiter*10, $chtrlen, 0); # testrun("A2", $loops, 1, 1, 1, $chiter*10, $chtrlen, 1); # testrun("B1", $loops, 1, 90, 1, $chiter , $chtrlen, 0); # testrun("B2", $loops, 1, 90, 1, $chiter , $chtrlen, 1); # testrun("C1", $loops, 1, 90, 50, $chiter , $chtrlen, 0); # testrun("C2", $loops, 1, 90, 50, $chiter , $chtrlen, 10); # testrun("D1", $loops, 50, 90, 1, $chiter , $chtrlen, 0); # testrun("D2", $loops, 50, 90, 1, $chiter , $chtrlen, 1); # testrun("E1", $loops, 50, 90, 1, $chiter , $chtrlen, 10); # testrun("F1", $loops , 1, 10, 1, $chiter , $chtrlen, 90); testrun("G1", 10, 1, 400, 1000, 20000, 1000, 0); testrun("G2", 10, 1, 400, 1000, 20000, 1000, 1); exit; sub testrun { my ($test_name, $loops, $ndbs, $nprocs, $ntbls, $childiter, $childtrlen, $nreferers) = @_; my @results = (); my @refresults = (); # run each iteration for (my $l = 0 ; $l < $loops ; $l++) { my %starttime = (); my %endtime = (); # This file is used for stopping free-running processes. open(OUT, '>', $trg_file) || die "failed to open file:$!\n"; print OUT "$$\n"; close(OUT); # It's very mysterious that one dummy subprocess makes things # stable. Perl runs slowly if there's only one psql process # without this... { my $pid = fork; if ($pid == 0) { while (-f $trg_file) { usleep (500000); } exit; } } # start referer processes, collecting how many times the query ran. pipe(my $refresrd, my $refreswr); for(my $i = 0 ; $i < $nreferers ; $i++) { my $pid = fork; if ($pid < 0) { die "fork referer process failed : $!\n"; } if ($pid == 0) { my $pid = open2(my $psqlrd, my $psqlwr, "psql postgres"); my $count = 0; close($refresrd); if ($pid < 0) { die "fork psql failed : $!\n"; } while (-f $trg_file) { print $psqlwr "select * from pg_stat_user_tables;\n"; while (<$psqlrd>) { last if ($_ =~ /rows\)/); } $count++; } print $refreswr "$count\n"; exit; } } close($refreswr); # launch updator processes for (my $i = 0 ; $i < $nprocs ; $i++) { #my $dbn = rand($ndbs); # equal dist. my $dbn = $i % $totaldbs; # round robbin my $dbname = sprintf("db%03d", $dbn); my $pid = fork; if ($pid < 0) { die "fork failed: $!\n"; } elsif ($pid == 0) { my $pid = open2(my $rd, my $wr, "psql $dbname > /dev/null"); if ($pid < 0) { die "sub fork failed: $!\n"; } my $ncmd = $childtrlen; #print $wr "set log_min_duration_statement to 0;\n"; print $wr "begin;\n"; for (my $i = 0 ; $i < $childiter ; $i++) { #my $tbn = rand($ntbls); # equal dist. my $tbn = $i % $totaltbls; # round robbin printf $wr "select /* $dbname\[$i\]*/ count(*) from t%03d;\n", $tbn; if (--$ncmd == 0) { print $wr "commit;begin;\n"; $ncmd = $childtrlen; } } print $wr "commit;\n"; print $wr "\\q\n"; my $res = <$rd>; exit; } (my $sec, my $usec) = &gettimeofday(); $starttime{$pid} = $sec * 1000 + $usec / 1000; # print "start\[$pid\] = $starttime{$pid}\n"; } # wait for updateors to finish for (my $i = 0 ; $i < $nprocs ; $i++) { my $pid = wait(); if ($pid < 0) { if ($! != ECHILD) {die "???: $!\n"; }} redo if (!defined $starttime{$pid}); (my $sec, my $usec) = &gettimeofday(); $endtime{$pid} = $sec * 1000 + $usec / 1000; # printf "%d[%d]: %d - %d = %d ms\n", $i, $pid, $endtime{$pid}, $starttime{$pid}, $endtime{$pid} - $starttime{$pid}; } my $sum = 0; foreach my $pid (keys %starttime) { $sum += $endtime{$pid} - $starttime{$pid}; # printf "[%d]: %d ms (%d, %d)\n", $pid, $endtime{$pid} - $starttime{$pid}, $starttime{$pid}, $endtime{$pid}; } push(@results, $sum / $nprocs); # kill referers if any (and dummy process) unlink($trg_file); while (wait() == 0) {} my $nrefs = 0; my $refcount = 0; while(<$refresrd>) { chomp; $refcount += $_; $nrefs++; } close($refresrd); push (@refresults, $refcount / $nrefs) if ($nrefs > 0); } # calculate stdev (my $updmean, my $updstdev) = stdev(@results); (my $refmean, my $refstdev) = stdev(@refresults); printf "$test_name (l:%d, d:%d, t:%d, i:%d, tr:%d): %.2f ms (stdev %.2f) / %d updprocs, %.2f refs (stdev %.2f) / %d refprocs\n", $loops, $ndbs, $ntbls, $childiter, $childtrlen, $updmean, $updstdev, $nprocs, $refmean, $refstdev, $nreferers; } sub stdev { my $sum = 0; my $sqsum = 0; my $count = $#_ + 1; return (0, 0) if $count == 0; foreach my $el (@_) { $sum += 1.0 * $el; $sqsum += 1.0 * $el * $el; } my $mean = $sum / $count; my $stdev = sqrt($sqsum / $count - $mean * $mean); return ($mean, $stdev); } ----Next_Part(Thu_Jul_11_16_40_59_2019_397)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH 5/5] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/replication/basebackup.c | 13 ++------- src/backend/statmon/pgstat.c | 13 ++++----- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- 9 files changed, 14 insertions(+), 90 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index a73fd4d044..95285809c2 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1119,8 +1119,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 8bd57f376b..79f704cc99 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6709,25 +6709,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index 5c1408bdf5..b538799ff6 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -197,12 +197,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index cbdad0c3fb..133eb3ff19 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index e30b2dbcf0..a567aacf73 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -231,11 +231,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -266,13 +263,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/statmon/pgstat.c b/src/backend/statmon/pgstat.c index 9dcd7ab540..805b64e758 100644 --- a/src/backend/statmon/pgstat.c +++ b/src/backend/statmon/pgstat.c @@ -89,15 +89,12 @@ typedef enum PgStat_TableLookupState bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; /* Shared stats bootstrap information */ typedef struct StatsShmemStruct { diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 62a07727d0..49123204c5 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -192,7 +192,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -3989,17 +3988,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11003,35 +10991,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 194f312096..fdb088dbfd 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -554,7 +554,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index eaa9d11c2a..28e039f4bc 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -29,7 +29,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ -- 2.16.3 ----Next_Part(Mon_Feb_18_21_35_31_2019_949)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH 5/5] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/replication/basebackup.c | 13 ++------- src/backend/statmon/pgstat.c | 13 ++++----- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- 9 files changed, 14 insertions(+), 90 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index a73fd4d044..95285809c2 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1119,8 +1119,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 07b847a8e9..ae226b9e3d 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6709,25 +6709,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index 7a84f51340..b37a2fd165 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -197,12 +197,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index cbdad0c3fb..133eb3ff19 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index e30b2dbcf0..a567aacf73 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -231,11 +231,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -266,13 +263,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/statmon/pgstat.c b/src/backend/statmon/pgstat.c index 66fb1e9341..1bc2557388 100644 --- a/src/backend/statmon/pgstat.c +++ b/src/backend/statmon/pgstat.c @@ -85,15 +85,12 @@ typedef enum PgStat_TableLookupState; bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; /* Shared stats bootstrap infomation */ typedef struct StatsShmemStruct { diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index fb7856517e..ed4a14587c 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -192,7 +192,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -3988,17 +3987,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11002,35 +10990,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index ad6c436f93..c7648dcb47 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -554,7 +554,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index 92c9adf48e..9ab6df5fae 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -28,7 +28,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ -- 2.16.3 ----Next_Part(Fri_Feb_15_17_29_00_2019_199)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH 5/5] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 ++++----- src/backend/replication/basebackup.c | 13 ++------- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- 9 files changed, 14 insertions(+), 90 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index b67da8916a..f0daffbc93 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1145,8 +1145,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 84341a30e5..6ad619de47 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6827,25 +6827,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index 0fe888e4db..31fc3604ea 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -195,12 +195,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index c4bac87e80..c264f839e7 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index 072f3ec62e..cec1ece065 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -109,15 +109,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; #define StatsLock (&StatsShmem->StatsMainLock) diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 36dcb28754..5b6de31ff6 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -230,11 +230,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -265,13 +262,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index ed51da4234..5c7e83ffdb 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -194,7 +194,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -4085,17 +4084,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11363,35 +11351,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 5ee5e09ddf..0ba984b074 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -562,7 +562,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index 2196b0bd38..4e0a05b2ce 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -31,7 +31,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ -- 2.16.3 ----Next_Part(Fri_May_17_14_27_22_2019_078)-- Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="run.sh" #! /usr/bin/bash DURATION=300 DURATION2=1500 ROOT=/home/horiguti BIN1PATH=$ROOT/bin/pgsql_master_o2/bin BIN2PATH=$ROOT/bin/pgsql_shmemstat/bin DATAPATH=$ROOT/data/data_shmemstat run() { local BINARY=$1 echo "## $BINARY-A1" pgbench -j 1 -c 1 -T $DURATION -b select-only postgres echo "## $BINARY-B1" pgbench -j 100 -c 100 -T $DURATION -b select-only postgres echo "## $BINARY-C1" pgbench -j 1 -c 1 -T $DURATION -f tr.sql postgres echo "## $BINARY-D1" pgbench -j 100 -c 100 -T $DURATION -f tr.sql postgres echo "## $BINARY-A2" ((pgbench -j 1 -c 1 -T $DURATION -f ref.sql postgres | sed -e 's/^/1:/') & (pgbench -j 1 -c 1 -T $DURATION -b select-only postgres | sed -e 's/^/2:/')); sleep 1 echo "## $BINARY-B2" ((pgbench -j 1 -c 1 -T $DURATION -f ref.sql postgres | sed -e 's/^/1:/') & (pgbench -j 97 -c 97 -T $DURATION -b select-only postgres | sed -e 's/^/2:/')); sleep 1 echo "## $BINARY-C2" ((pgbench -j 1 -c 1 -T $DURATION -f ref.sql postgres | sed -e 's/^/1:/') & (pgbench -j 1 -c 1 -T $DURATION -f tr.sql postgres | sed -e 's/^/2:/')); sleep 1 echo "## $BINARY-D2" ((pgbench -j 1 -c 1 -T $DURATION -f ref.sql postgres | sed -e 's/^/1:/') & (pgbench -j 97 -c 97 -T $DURATION -f tr.sql postgres | sed -e 's/^/2:/')); sleep 1 } $BIN1PATH/pg_ctl -D $DATAPATH start run "o2" $BIN1PATH/pg_ctl -D $DATAPATH stop -m s $BIN2PATH/pg_ctl -D $DATAPATH start run "shmem" $BIN2PATH/pg_ctl -D $DATAPATH stop -m s ----Next_Part(Fri_May_17_14_27_22_2019_078)-- Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="gencr.pl" #! /usr/bin/perl print "drop table if exists p;\n"; print "create table p (a int) partition by list (a);\n"; for ($i = 0 ; $i < 1000 ; $i++) { printf("create table c%03d partition of p for values in (%d);\n", $i, $i); printf("insert into p values(%d);\n", $i); } ----Next_Part(Fri_May_17_14_27_22_2019_078)-- Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="tr.sql" select sum(a) from p; ----Next_Part(Fri_May_17_14_27_22_2019_078)-- Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="ref.sql" select sum(seq_scan) from pg_stat_user_tables; ----Next_Part(Fri_May_17_14_27_22_2019_078)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH 5/5] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 ++++----- src/backend/replication/basebackup.c | 13 ++------- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- 9 files changed, 14 insertions(+), 90 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index b67da8916a..f0daffbc93 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1145,8 +1145,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 84341a30e5..6ad619de47 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6827,25 +6827,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index 0fe888e4db..31fc3604ea 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -195,12 +195,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index c4bac87e80..c264f839e7 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index 24dab5aa54..94df252597 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -109,15 +109,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; #define StatsLock (&StatsShmem->StatsMainLock) diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 36dcb28754..5b6de31ff6 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -230,11 +230,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -265,13 +262,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index ed51da4234..5c7e83ffdb 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -194,7 +194,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -4085,17 +4084,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11363,35 +11351,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 5ee5e09ddf..0ba984b074 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -562,7 +562,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index 2196b0bd38..4e0a05b2ce 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -31,7 +31,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ -- 2.16.3 ----Next_Part(Fri_May_17_15_47_20_2019_554)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v23 5/5] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 ++++----- src/backend/replication/basebackup.c | 13 ++------- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- src/test/perl/PostgresNode.pm | 4 --- 10 files changed, 14 insertions(+), 94 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index bdc9026c62..2885540362 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1146,8 +1146,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 6612f95f9f..b346809c11 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6818,25 +6818,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index ea6aad4d1e..33ad2b8be8 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -195,12 +195,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index 1c19e863d2..2f04bb68bb 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index c0b20763b0..6b8025ad13 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -108,15 +108,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; #define StatsLock (&StatsShmem->StatsMainLock) diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index d0f210de8c..39fcf29ff2 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -242,11 +242,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -277,13 +274,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 2178e1cf5e..50625421ab 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -194,7 +194,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -4072,17 +4071,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11352,35 +11340,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 0fc23e3a61..66f539c4bb 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -558,7 +558,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index c9fbcead3f..e9e18ed27a 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -30,7 +30,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ diff --git a/src/test/perl/PostgresNode.pm b/src/test/perl/PostgresNode.pm index 270bd6c856..c604c5e90b 100644 --- a/src/test/perl/PostgresNode.pm +++ b/src/test/perl/PostgresNode.pm @@ -455,10 +455,6 @@ sub init print $conf TestLib::slurp_file($ENV{TEMP_CONFIG}) if defined $ENV{TEMP_CONFIG}; - # XXX Neutralize any stats_temp_directory in TEMP_CONFIG. Nodes running - # concurrently must not share a stats_temp_directory. - print $conf "stats_temp_directory = 'pg_stat_tmp'\n"; - if ($params{allows_streaming}) { if ($params{allows_streaming} eq "logical") -- 2.16.3 ----Next_Part(Fri_Sep_27_09_46_47_2019_292)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH 6/6] Remove the GUC stats_temp_directory @ 2018-11-27 05:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2018-11-27 05:42 UTC (permalink / raw) The guc used to specifie the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contirb, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward comptibility. --- doc/src/sgml/backup.sgml | 2 -- doc/src/sgml/config.sgml | 19 ------------- doc/src/sgml/monitoring.sgml | 7 +---- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 ++++----- src/backend/replication/basebackup.c | 13 ++------- src/backend/utils/misc/guc.c | 41 --------------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 +++- 9 files changed, 14 insertions(+), 90 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index a73fd4d044..95285809c2 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1119,8 +1119,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 8bd57f376b..79f704cc99 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -6709,25 +6709,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml index 83a0be1965..bf94bd87d2 100644 --- a/doc/src/sgml/monitoring.sgml +++ b/doc/src/sgml/monitoring.sgml @@ -195,12 +195,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser <para> The statistics collector transmits the collected information to other - <productname>PostgreSQL</productname> processes through temporary files. - These files are stored in the directory named by the - <xref linkend="guc-stats-temp-directory"/> parameter, - <filename>pg_stat_tmp</filename> by default. - For better performance, <varname>stats_temp_directory</varname> can be - pointed at a RAM-based file system, decreasing physical I/O requirements. + <productname>PostgreSQL</productname> processes through shared memory. When the server shuts down cleanly, a permanent copy of the statistics data is stored in the <filename>pg_stat</filename> subdirectory, so that statistics can be retained across server restarts. When recovery is diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index cbdad0c3fb..133eb3ff19 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index ba4c274f9a..8f8ac0b356 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -102,15 +102,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward comptibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; LWLock StatsMainLock; #define StatsLock (&StatsMainLock) diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index def6c03dd0..58ba33e822 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -230,11 +230,8 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -265,13 +262,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 156d147c85..058d97075f 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -191,7 +191,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -3988,17 +3987,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11002,35 +10990,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index 194f312096..fdb088dbfd 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -554,7 +554,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index cb70d00b8f..265ed4d118 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -31,7 +31,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ -- 2.16.3 ----Next_Part(Thu_Feb_21_16_05_55_2019_560)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v27 7/7] Remove the GUC stats_temp_directory @ 2020-03-13 08:00 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-03-13 08:00 UTC (permalink / raw) The GUC used to specify the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contrib, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward compatibility. --- doc/src/sgml/backup.sgml | 2 - doc/src/sgml/config.sgml | 19 --------- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 +++--- src/backend/replication/basebackup.c | 13 ++---- src/backend/utils/misc/guc.c | 41 ------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 ++- src/test/perl/PostgresNode.pm | 4 -- 9 files changed, 13 insertions(+), 88 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index bdc9026c62..2885540362 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1146,8 +1146,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 680e1c3564..43d0d303ad 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -7111,25 +7111,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index 1c19e863d2..2f04bb68bb 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index c0760854f4..053cd467fd 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -95,15 +95,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward compatibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; /* * Shared stats bootstrap information, protected by StatsLock. diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index a2e28b064c..7b7d87b938 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -254,7 +254,6 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; backup_total = 0; @@ -273,8 +272,6 @@ perform_base_backup(basebackup_options *opt) backup_total); } - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -306,13 +303,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index af876d1f01..cabeb806c5 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -197,7 +197,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static bool check_maintenance_io_concurrency(int *newval, void **extra, GucSource source); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -4231,17 +4230,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11518,35 +11506,6 @@ check_maintenance_io_concurrency(int *newval, void **extra, GucSource source) return true; } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index aa44f0c9bf..207e042e99 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -573,7 +573,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index 6fad13c4be..4971a88c70 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -32,7 +32,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ diff --git a/src/test/perl/PostgresNode.pm b/src/test/perl/PostgresNode.pm index 9575268bd7..f3340f726c 100644 --- a/src/test/perl/PostgresNode.pm +++ b/src/test/perl/PostgresNode.pm @@ -455,10 +455,6 @@ sub init print $conf TestLib::slurp_file($ENV{TEMP_CONFIG}) if defined $ENV{TEMP_CONFIG}; - # XXX Neutralize any stats_temp_directory in TEMP_CONFIG. Nodes running - # concurrently must not share a stats_temp_directory. - print $conf "stats_temp_directory = 'pg_stat_tmp'\n"; - if ($params{allows_streaming}) { if ($params{allows_streaming} eq "logical") -- 2.18.2 ----Next_Part(Fri_Mar_27_16_31_15_2020_716)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v28 7/7] Remove the GUC stats_temp_directory @ 2020-03-13 08:00 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-03-13 08:00 UTC (permalink / raw) The GUC used to specify the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contrib, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward compatibility. --- doc/src/sgml/backup.sgml | 2 - doc/src/sgml/config.sgml | 19 --------- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 +++--- src/backend/replication/basebackup.c | 13 ++---- src/backend/utils/misc/guc.c | 41 ------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 ++- src/test/perl/PostgresNode.pm | 4 -- 9 files changed, 13 insertions(+), 88 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index bdc9026c62..2885540362 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1146,8 +1146,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 7ed2b3884c..0c251c8ac6 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -7120,25 +7120,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index 1c19e863d2..2f04bb68bb 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index 8c65df9bfd..fbb6287fa5 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -95,15 +95,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward compatibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; /* * Shared stats bootstrap information, protected by StatsLock. diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index a2e28b064c..7b7d87b938 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -254,7 +254,6 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; backup_total = 0; @@ -273,8 +272,6 @@ perform_base_backup(basebackup_options *opt) backup_total); } - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -306,13 +303,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 79bc7ac8ca..f94cec4677 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -197,7 +197,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static bool check_maintenance_io_concurrency(int *newval, void **extra, GucSource source); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -4251,17 +4250,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11538,35 +11526,6 @@ check_maintenance_io_concurrency(int *newval, void **extra, GucSource source) return true; } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index e9f8ca775d..4fd040b9c7 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -572,7 +572,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index 69bd794806..263f9ace1f 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -32,7 +32,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ diff --git a/src/test/perl/PostgresNode.pm b/src/test/perl/PostgresNode.pm index 9575268bd7..f3340f726c 100644 --- a/src/test/perl/PostgresNode.pm +++ b/src/test/perl/PostgresNode.pm @@ -455,10 +455,6 @@ sub init print $conf TestLib::slurp_file($ENV{TEMP_CONFIG}) if defined $ENV{TEMP_CONFIG}; - # XXX Neutralize any stats_temp_directory in TEMP_CONFIG. Nodes running - # concurrently must not share a stats_temp_directory. - print $conf "stats_temp_directory = 'pg_stat_tmp'\n"; - if ($params{allows_streaming}) { if ($params{allows_streaming} eq "logical") -- 2.18.2 ----Next_Part(Mon_Mar_30_09_29_45_2020_582)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v25 8/8] Remove the GUC stats_temp_directory @ 2020-03-13 08:00 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-03-13 08:00 UTC (permalink / raw) The GUC used to specify the directory to store temporary statistics files. It is no longer needed by the stats collector but still used by the programs in bin and contrib, and maybe other extensions. Thus this patch removes the GUC but some backing variables and macro definitions are left alone for backward compatibility. --- doc/src/sgml/backup.sgml | 2 - doc/src/sgml/config.sgml | 19 --------- doc/src/sgml/storage.sgml | 3 +- src/backend/postmaster/pgstat.c | 13 +++--- src/backend/replication/basebackup.c | 13 ++---- src/backend/utils/misc/guc.c | 41 ------------------- src/backend/utils/misc/postgresql.conf.sample | 1 - src/include/pgstat.h | 5 ++- src/test/perl/PostgresNode.pm | 4 -- 9 files changed, 13 insertions(+), 88 deletions(-) diff --git a/doc/src/sgml/backup.sgml b/doc/src/sgml/backup.sgml index bdc9026c62..2885540362 100644 --- a/doc/src/sgml/backup.sgml +++ b/doc/src/sgml/backup.sgml @@ -1146,8 +1146,6 @@ SELECT pg_stop_backup(); <filename>pg_snapshots/</filename>, <filename>pg_stat_tmp/</filename>, and <filename>pg_subtrans/</filename> (but not the directories themselves) can be omitted from the backup as they will be initialized on postmaster startup. - If <xref linkend="guc-stats-temp-directory"/> is set and is under the data - directory then the contents of that directory can also be omitted. </para> <para> diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 8cd86beb9d..7f6056b9e9 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -7056,25 +7056,6 @@ COPY postgres_log FROM '/full/path/to/logfile.csv' WITH csv; </listitem> </varlistentry> - <varlistentry id="guc-stats-temp-directory" xreflabel="stats_temp_directory"> - <term><varname>stats_temp_directory</varname> (<type>string</type>) - <indexterm> - <primary><varname>stats_temp_directory</varname> configuration parameter</primary> - </indexterm> - </term> - <listitem> - <para> - Sets the directory to store temporary statistics data in. This can be - a path relative to the data directory or an absolute path. The default - is <filename>pg_stat_tmp</filename>. Pointing this at a RAM-based - file system will decrease physical I/O requirements and can lead to - improved performance. - This parameter can only be set in the <filename>postgresql.conf</filename> - file or on the server command line. - </para> - </listitem> - </varlistentry> - </variablelist> </sect2> diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml index 1c19e863d2..2f04bb68bb 100644 --- a/doc/src/sgml/storage.sgml +++ b/doc/src/sgml/storage.sgml @@ -122,8 +122,7 @@ Item <row> <entry><filename>pg_stat_tmp</filename></entry> - <entry>Subdirectory containing temporary files for the statistics - subsystem</entry> + <entry>Subdirectory containing ephemeral files for extensions</entry> </row> <row> diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c index 34a4005791..4cd8530e91 100644 --- a/src/backend/postmaster/pgstat.c +++ b/src/backend/postmaster/pgstat.c @@ -96,15 +96,12 @@ bool pgstat_track_counts = false; int pgstat_track_functions = TRACK_FUNC_OFF; int pgstat_track_activity_query_size = 1024; -/* ---------- - * Built from GUC parameter - * ---------- +/* + * This used to be a GUC variable and is no longer used in this file, but left + * alone just for backward compatibility for extensions, having the default + * value. */ -char *pgstat_stat_directory = NULL; - -/* No longer used, but will be removed with GUC */ -char *pgstat_stat_filename = NULL; -char *pgstat_stat_tmpname = NULL; +char *pgstat_stat_directory = PG_STAT_TMP_DIR; /* Shared stats bootstrap information, protected by StatsLock */ typedef struct StatsShmemStruct diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 806d013108..c086ab781b 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -251,15 +251,12 @@ perform_base_backup(basebackup_options *opt) TimeLineID endtli; StringInfo labelfile; StringInfo tblspc_map_file = NULL; - int datadirpathlen; List *tablespaces = NIL; backup_total = 0; backup_streamed = 0; pgstat_progress_start_command(PROGRESS_COMMAND_BASEBACKUP, InvalidOid); - datadirpathlen = strlen(DataDir); - backup_started_in_recovery = RecoveryInProgress(); labelfile = makeStringInfo(); @@ -291,13 +288,9 @@ perform_base_backup(basebackup_options *opt) * Calculate the relative path of temporary statistics directory in * order to skip the files which are located in that directory later. */ - if (is_absolute_path(pgstat_stat_directory) && - strncmp(pgstat_stat_directory, DataDir, datadirpathlen) == 0) - statrelpath = psprintf("./%s", pgstat_stat_directory + datadirpathlen + 1); - else if (strncmp(pgstat_stat_directory, "./", 2) != 0) - statrelpath = psprintf("./%s", pgstat_stat_directory); - else - statrelpath = pgstat_stat_directory; + + Assert(strchr(PG_STAT_TMP_DIR, '/') == NULL); + statrelpath = psprintf("./%s", PG_STAT_TMP_DIR); /* Add a node for the base directory at the end */ ti = palloc0(sizeof(tablespaceinfo)); diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 4c6d648662..417fbbdc5d 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -197,7 +197,6 @@ static bool check_max_wal_senders(int *newval, void **extra, GucSource source); static bool check_autovacuum_work_mem(int *newval, void **extra, GucSource source); static bool check_effective_io_concurrency(int *newval, void **extra, GucSource source); static void assign_effective_io_concurrency(int newval, void *extra); -static void assign_pgstat_temp_directory(const char *newval, void *extra); static bool check_application_name(char **newval, void **extra, GucSource source); static void assign_application_name(const char *newval, void *extra); static bool check_cluster_name(char **newval, void **extra, GucSource source); @@ -4193,17 +4192,6 @@ static struct config_string ConfigureNamesString[] = NULL, NULL, NULL }, - { - {"stats_temp_directory", PGC_SIGHUP, STATS_COLLECTOR, - gettext_noop("Writes temporary statistics files to the specified directory."), - NULL, - GUC_SUPERUSER_ONLY - }, - &pgstat_temp_directory, - PG_STAT_TMP_DIR, - check_canonical_path, assign_pgstat_temp_directory, NULL - }, - { {"synchronous_standby_names", PGC_SIGHUP, REPLICATION_MASTER, gettext_noop("Number of synchronous standbys and list of names of potential synchronous ones."), @@ -11489,35 +11477,6 @@ assign_effective_io_concurrency(int newval, void *extra) #endif /* USE_PREFETCH */ } -static void -assign_pgstat_temp_directory(const char *newval, void *extra) -{ - /* check_canonical_path already canonicalized newval for us */ - char *dname; - char *tname; - char *fname; - - /* directory */ - dname = guc_malloc(ERROR, strlen(newval) + 1); /* runtime dir */ - sprintf(dname, "%s", newval); - - /* global stats */ - tname = guc_malloc(ERROR, strlen(newval) + 12); /* /global.tmp */ - sprintf(tname, "%s/global.tmp", newval); - fname = guc_malloc(ERROR, strlen(newval) + 13); /* /global.stat */ - sprintf(fname, "%s/global.stat", newval); - - if (pgstat_stat_directory) - free(pgstat_stat_directory); - pgstat_stat_directory = dname; - if (pgstat_stat_tmpname) - free(pgstat_stat_tmpname); - pgstat_stat_tmpname = tname; - if (pgstat_stat_filename) - free(pgstat_stat_filename); - pgstat_stat_filename = fname; -} - static bool check_application_name(char **newval, void **extra, GucSource source) { diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index aa44f0c9bf..207e042e99 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -573,7 +573,6 @@ #track_io_timing = off #track_functions = none # none, pl, all #track_activity_query_size = 1024 # (change requires restart) -#stats_temp_directory = 'pg_stat_tmp' # - Monitoring - diff --git a/src/include/pgstat.h b/src/include/pgstat.h index 4e137140bd..062f393941 100644 --- a/src/include/pgstat.h +++ b/src/include/pgstat.h @@ -32,7 +32,10 @@ #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" -/* Default directory to store temporary statistics data in */ +/* + * This used to be the directory to store temporary statistics data in but is + * no longer used. Defined here for backward compatibility. + */ #define PG_STAT_TMP_DIR "pg_stat_tmp" /* Values for track_functions GUC variable --- order is significant! */ diff --git a/src/test/perl/PostgresNode.pm b/src/test/perl/PostgresNode.pm index 9575268bd7..f3340f726c 100644 --- a/src/test/perl/PostgresNode.pm +++ b/src/test/perl/PostgresNode.pm @@ -455,10 +455,6 @@ sub init print $conf TestLib::slurp_file($ENV{TEMP_CONFIG}) if defined $ENV{TEMP_CONFIG}; - # XXX Neutralize any stats_temp_directory in TEMP_CONFIG. Nodes running - # concurrently must not share a stats_temp_directory. - print $conf "stats_temp_directory = 'pg_stat_tmp'\n"; - if ($params{allows_streaming}) { if ($params{allows_streaming} eq "logical") -- 2.18.2 ----Next_Part(Thu_Mar_19_20_30_04_2020_284)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v9 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 +++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 593 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 256 +++++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 ++++++++++---- src/backend/storage/smgr/md.c | 93 +++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 + src/common/relpath.c | 47 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + 22 files changed, 1465 insertions(+), 207 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..d251f22207 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..b344bbe511 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 1e1fbe957f..59f4c2eacf 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4564,6 +4565,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7824,6 +7833,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index c5ad28d71f..03fccc3c3b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,18 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (1 << 0) +#define PDOP_UNLINK_FORK (1 << 1) +#define PDOP_UNLINK_MARK (1 << 2) +#define PDOP_SET_PERSISTENCE (1 << 3) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +86,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,22 +172,47 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file works as the signal of orphan files. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); - /* Add the relation to the list of stuff to delete at abort */ + /* + * Add the relation to the list of stuff to delete at abort. We don't + * remove the mark file at commit. It needs to persists until the main fork + * file is actually deleted. See SyncPostCheckpoint. + */ pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_MARK; + pending->unlink_forknum = MAIN_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +222,226 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have entries for init-fork operations on this relation, that means + * that we have already registered pending delete entries to drop an + * init-fork preexisting since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + + /* + * We don't touch unrelated entries. Although init-fork related entries + * are not useful if the relation is created or dropped in this + * transaction, we don't bother to avoid registering entries for such + * relations here. + */ + if (!RelFileNodeEquals(rnode, pending->relnode) || + pending->unlink_forknum != INIT_FORKNUM || + (pending->op & PDOP_DELETE) != 0) + { + prev = pending; + continue; + } + + /* make sure the entry is what we're expecting here */ + Assert(((pending->op & (PDOP_UNLINK_FORK|PDOP_UNLINK_MARK)) != 0 && + pending->unlink_forknum == INIT_FORKNUM) || + (pending->op & PDOP_SET_PERSISTENCE) != 0); + + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + + if (!create) + return; + + /* + * We are going to create an init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The mark file works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_UNLINK_MARK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop mark file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just register pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + + /* + * We don't touch unrelated entries. Although init-fork related entries + * are not useful if the relation is created or dropped in this + * transaction, we don't bother to avoid registering entries for such + * relations here. + */ + if (!RelFileNodeEquals(rnode, pending->relnode) || + pending->unlink_forknum != INIT_FORKNUM || + (pending->op & PDOP_DELETE) != 0) + { + prev = pending; + continue; + } + + /* make sure the entry is what we're expecting here */ + Assert(((pending->op & (PDOP_UNLINK_FORK|PDOP_UNLINK_MARK)) != 0 && + pending->unlink_forknum == INIT_FORKNUM) || + (pending->op & PDOP_SET_PERSISTENCE) != 0); + + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +461,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +556,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -618,59 +975,104 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op & PDOP_DELETE) + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; + } + + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PDOP_UNLINK_MARK) + { + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -840,7 +1242,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -853,7 +1256,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -933,6 +1337,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1434,120 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingRelDelete *prev = NULL; + + for (pending = pendingDeletes; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PDOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingDeletes = pending->next; + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + PendingRelDelete *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingDeletes; pending != NULL; pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PDOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingDeletes = pending->next; + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index bf42587e38..0d9c801535 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -52,6 +52,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5329,6 +5330,177 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother to allow in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = rel->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5459,47 +5631,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index ec0485705d..45e1a5d817 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1070,6 +1070,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1120,7 +1121,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index b4532948d3..dab74bf99a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3154,6 +3155,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 263057841d..8487ae1f02 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 0ae3fb6902..0137902bb2 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index b4bca7eed6..580b74839f 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,81 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1025,6 +1101,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1378,12 +1463,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0fcef4994b..110e64b0b2 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index d4083e8a56..9563940d45 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -89,7 +89,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -222,7 +223,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -236,6 +238,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 436df54120..dbc0da5da5 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..67f24890d6 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[10]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, 10, ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..12346ed7f6 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..106a5cf508 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index cfce23ecbc..f5a7df87a4 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 34602ae006..2dc0357ad5 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..99620816b5 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..e1f97e9b89 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..201ecace8a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Mon_Dec_20_16_53_20_2021_853)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v9-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v19 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 49 ++ src/backend/access/transam/README | 9 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlogrecovery.c | 18 + src/backend/catalog/storage.c | 548 +++++++++++++++++++++- src/backend/commands/tablecmds.c | 266 +++++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 86 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 ++++++++++---- src/backend/storage/smgr/md.c | 94 +++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 22 + src/bin/pg_rewind/pg_rewind.c | 1 - src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + 24 files changed, 1459 insertions(+), 183 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7547813254..225ffbafef 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,46 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action = "<none>"; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +95,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..2ecd8c8c7c 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -725,6 +725,15 @@ then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index adf763a8ea..559666b802 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2198,6 +2198,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2448,6 +2451,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2773,6 +2779,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index f9f212680b..2923b8ef8c 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -40,6 +40,7 @@ #include "access/xlogrecovery.h" #include "access/xlogutils.h" #include "catalog/pg_control.h" +#include "catalog/storage.h" #include "commands/tablespace.h" #include "miscadmin.h" #include "pgstat.h" @@ -53,6 +54,7 @@ #include "storage/pmsignal.h" #include "storage/proc.h" #include "storage/procarray.h" +#include "storage/reinit.h" #include "storage/spin.h" #include "utils/builtins.h" #include "utils/guc.h" @@ -1746,6 +1748,14 @@ PerformWalRecovery(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { @@ -3022,6 +3032,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 9b8075536a..cd1445713a 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup *pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -117,7 +136,8 @@ AddPendingSync(const RelFileNode *rnode) SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; SMgrRelation srel; BackendId backend; bool needs_wal; @@ -143,21 +163,41 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); /* Add the relation to the list of stuff to delete at abort */ - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { @@ -168,6 +208,200 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init-fork exists since before the current transaction + * started. This function reverts that change just by removing the entry. + * See RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the commit-sentinel file */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * init fork for indexes needs further initialization. ambuildempty should + * do WAL-log and file sync by itself but otherwise we do that by + * ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion is canceled. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init fork is created in the current transaction. We remove + * both the init fork and mark file immediately in that case. Otherwise + * just register a pending-unlink for the existing init fork. See + * RelationCreateInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks are never loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +421,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -673,6 +989,95 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * part of the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert ((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + + if (pending->op & PCOP_UNLINK_FORK) + { + /* + * Unlink the fork file. Currently we use this only for + * init forks and we're sure that the init fork is not + * loaded on shared buffers. For RelationDropInitFork + * case, the function dropped that buffers. For + * RelationCreateInitFork case, PCOP_SET_PERSISTENCE(true) + * is set and the buffers have been dropped just before. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -933,6 +1338,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1435,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 3e83f375b5..9e5b77e94a 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -53,6 +53,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5347,6 +5348,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5477,47 +5659,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 0bf28b55d7..17185f4e55 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1209,6 +1209,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1259,7 +1260,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index f5459c68f8..6cd010429a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -38,6 +38,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3155,6 +3156,91 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(bufHdr->tag.forkNum != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 14b77f2861..2fc9f17c28 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index f053fe0495..f28f55baa6 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlogrecovery.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 879f647dbc..4d44bdd78b 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,82 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + close(fd); + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1031,6 +1108,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1384,12 +1470,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index d71a557a35..0710e8b145 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -63,6 +63,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -84,6 +88,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -337,6 +343,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -664,6 +690,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index e161d57761..f5ded7cb34 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -90,7 +90,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -223,7 +224,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -237,6 +239,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 56df08c64f..f1382d4c4f 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,28 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore there. We'll see that the file don't exist in + * the target data dir, and copy them in from the source system. No + * need to do anything special here. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these, The file will be removed from the + * target, if it doesn't exist in the source system. The files are + * empty so we don't need to bother the content. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/bin/pg_rewind/pg_rewind.c b/src/bin/pg_rewind/pg_rewind.c index efb82a4034..b289df4060 100644 --- a/src/bin/pg_rewind/pg_rewind.c +++ b/src/bin/pg_rewind/pg_rewind.c @@ -412,7 +412,6 @@ main(int argc, char **argv) if (showprogress) pg_log_info("reading source file list"); source->traverse_files(source, &process_source_file); - if (showprogress) pg_log_info("reading target file list"); traverse_datadir(datadir_target, &process_target_file); diff --git a/src/common/relpath.c b/src/common/relpath.c index 636c96efd3..1c19e16fea 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 9ffc741913..d362d62ed2 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -41,6 +43,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 622de22b03..8139308634 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a4b5dc853b..a864c91614 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index dd01841c30..739b386216 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 29209e2724..8bf746bf45 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 6e46d8d96a..ef5fdaf4f8 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -24,6 +24,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); extern void mdrelease(void); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -42,12 +46,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index bf2c10d443..e399aec0c7 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index 8e3ef92cda..022654b7b2 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Mar__1_17_50_30_2022_341)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v19-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v14 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 545 +++++++++++++++++- src/backend/commands/tablecmds.c | 266 +++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 +++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 +++++++---- src/backend/storage/smgr/md.c | 93 ++- src/backend/storage/smgr/smgr.c | 32 + src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 + src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + src/test/recovery/t/027_persistence_change.pl | 247 ++++++++ 24 files changed, 1707 insertions(+), 182 deletions(-) create mode 100644 src/test/recovery/t/027_persistence_change.pl diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..d251f22207 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..b344bbe511 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index e7b0bc804d..b41186d6d8 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2197,6 +2197,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2447,6 +2450,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2772,6 +2778,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 87cd05c945..243860fcb1 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4564,6 +4565,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7824,6 +7833,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index c5ad28d71f..d6b30387e9 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup *pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -117,7 +136,8 @@ AddPendingSync(const RelFileNode *rnode) SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; SMgrRelation srel; BackendId backend; bool needs_wal; @@ -143,21 +163,41 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); /* Add the relation to the list of stuff to delete at abort */ - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { @@ -168,6 +208,203 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have entries for init-fork operations on this relation, that means + * that we have already registered pending delete entries to drop an + * init-fork preexisting since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* + * We are going to create an init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The mark file works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just register pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +424,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -255,6 +574,7 @@ RelationPreserveStorage(RelFileNode rnode, bool atCommit) prev->next = next; else pendingDeletes = next; + pfree(pending); /* prev does not change */ } @@ -673,6 +993,88 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * emitted before the commit record for the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert ((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -933,6 +1335,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1432,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 3631b8a929..848fda40ca 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -52,6 +52,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5329,6 +5330,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5459,47 +5641,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index ec0485705d..45e1a5d817 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1070,6 +1070,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1120,7 +1121,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index b4532948d3..dab74bf99a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3154,6 +3155,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 263057841d..8487ae1f02 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 0ae3fb6902..0137902bb2 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index b4bca7eed6..580b74839f 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,81 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1025,6 +1101,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1378,12 +1463,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0fcef4994b..110e64b0b2 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index d4083e8a56..9563940d45 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -89,7 +89,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -222,7 +223,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -236,6 +238,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 436df54120..dbc0da5da5 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..4945b111cc 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..584ebac391 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -41,6 +43,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..12346ed7f6 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..106a5cf508 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index cfce23ecbc..f5a7df87a4 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 34602ae006..2dc0357ad5 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..99620816b5 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..e1f97e9b89 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..201ecace8a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/test/recovery/t/027_persistence_change.pl b/src/test/recovery/t/027_persistence_change.pl new file mode 100644 index 0000000000..c2f7076ea9 --- /dev/null +++ b/src/test/recovery/t/027_persistence_change.pl @@ -0,0 +1,247 @@ + +# Copyright (c) 2021, PostgreSQL Global Development Group + +# Test relation persistence change +use strict; +use warnings; +use PostgreSQL::Test::Cluster; +use PostgreSQL::Test::Utils; +use Test::More; +use Test::More tests => 30; +use IPC::Run qw(pump finish timer); +use Config; + +my $data_unit = 2000; + +# Initialize primary node. +my $node = PostgreSQL::Test::Cluster->new('node'); +$node->init; +# we don't want checkpointing +$node->append_conf('postgresql.conf', qq( +checkpoint_timeout = '24h' +)); +$node->start; +create($node); + +my $relfilenodes1 = relfilenodes(); + +# correctly recover empty tables +$node->stop('immediate'); +$node->start; +insert($node, 0, $data_unit); + +# data persists after a crash +$node->stop('immediate'); +$node->start; +checkdataloss($data_unit, 'crash logged 1'); + +set_unlogged($node); +# SET UNLOGGED didn't change relfilenode +my $relfilenodes2 = relfilenodes(); +checkrelfilenodes($relfilenodes1, $relfilenodes2, 'logged->unlogged'); + +# data cleanly vanishes after a crash +$node->stop('immediate'); +$node->start; +checkdataloss(0, 'crash unlogged'); + +insert($node, 0, $data_unit); +set_logged($node); + +$node->stop('immediate'); +$node->start; +# SET LOGGED didn't change relfilenode and data survive a crash +my $relfilenodes3 = relfilenodes(); +checkrelfilenodes($relfilenodes2, $relfilenodes3, 'unlogged->logged'); +checkdataloss($data_unit, 'crash logged 2'); + +# unlogged insert -> graceful stop +set_unlogged($node); +insert($node, $data_unit, $data_unit, 0); +$node->stop; +$node->start; +checkdataloss($data_unit * 2, 'unlogged graceful restart'); + +# crash during transaction +set_logged($node); +$node->stop('immediate'); +$node->start; +insert($node, $data_unit * 2, $data_unit); + +my $h = insert($node, $data_unit * 3, $data_unit, 1); ## this is aborted +$node->stop('immediate'); + +# finishing $h stalls this case, just tear it off. +$h = undef; + +# check if indexes are working +$node->start; +# drop first half of data to reduce run time +$node->safe_psql('postgres', 'DELETE FROM t WHERE bt < ' . $data_unit * 2); +check($node, $data_unit * 2, $data_unit * 3 - 1, 'final check'); + +sub create +{ + my ($node) = @_; + + $node->psql('postgres', qq( + CREATE TABLE t (bt int, gin int[], gist point, hash int, + brin int, spgist point); + CREATE INDEX i_bt ON t USING btree (bt); + CREATE INDEX i_gin ON t USING gin (gin); + CREATE INDEX i_gist ON t USING gist (gist); + CREATE INDEX i_hash ON t USING hash (hash); + CREATE INDEX i_brin ON t USING brin (brin); + CREATE INDEX i_spgist ON t USING spgist (spgist);)); +} + + +sub insert +{ + my ($node, $st, $num, $interactive) = @_; + my $ed = $st + $num - 1; + my $query = qq(BEGIN; +INSERT INTO t + (SELECT i, ARRAY[i, i * 2], point(i, i * 2), i, i, point(i, i) + FROM generate_series($st, $ed) i); +); + + if ($interactive) + { + my $in = ''; + my $out = ''; + my $timer = timer(10); + + my $h = $node->interactive_psql('postgres', \$in, \$out, $timer); + like($out, qr/psql/, "print startup banner"); + + $timer->start(10); + $in .= $query . "SELECT 'END';\n"; + pump $h until ($out =~ /\nEND/ || $timer->is_expired); + ok(($out =~ /\nEND/ && !$timer->is_expired), "inserted-$st-$num"); + return $h + # the trasaction is not terminated + } + else + { + $node->psql('postgres', $query . "COMMIT;"); + return undef; + } +} + +sub check +{ + my ($node, $st, $ed, $head) = @_; + my $num_data = $ed - $st + 1; + + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO true; + SET enable_indexscan TO false; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE bt = i)), + $num_data, "$head: heap is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE bt = i)), + $num_data, "$head: btree is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE gin = ARRAY[i, i * 2];)), + $num_data, "$head: gin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE gist <@ box(point(i-0.5, i*2-0.5),point(i+0.5, i*2+0.5));)), + $num_data, "$head: gist is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE hash = i;)), + $num_data, "$head: hash is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE brin = i;)), + $num_data, "$head: brin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE spgist <@ box(point(i-0.5,i-0.5),point(i+0.5,i+0.5));)), + $num_data, "$head: spgist is not broken"); +} + +sub set_unlogged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t SET UNLOGGED; +)); +} + +sub set_logged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t SET LOGGED; +)); +} + +sub relfilenodes +{ + my $result = $node->safe_psql('postgres', qq{ + SELECT relname, relfilenode FROM pg_class + WHERE relname + IN ('t', 'i_bt','i_gin','i_gist','i_hash','i_brin','i_spgist');}); + + my %relfilenodes; + + foreach my $l (split(/\n/, $result)) + { + die "unexpected format: $l" if ($l !~ /^([^|]+)\|([0-9]+)$/); + $relfilenodes{$1} = $2; + } + + # the number must correspond to the in list above + is (scalar %relfilenodes, 7, "number of relations is correct"); + + return \%relfilenodes; +} + +sub checkrelfilenodes +{ + my ($rnodes1, $rnodes2, $s) = @_; + + foreach my $n (keys %{$rnodes1}) + { + if ($n eq 'i_gist') + { + # persistence of GiST index is not changed in-place + isnt($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is changed: $n"); + } + else + { + # otherwise all relations are processed in-place + is($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is not changed: $n"); + } + } +} + +sub checkdataloss +{ + my ($expected, $s) = @_; + + is($node->safe_psql('postgres', "SELECT count(*) FROM t;"), $expected, + "$s: data in table t is in the expected state"); +} -- 2.27.0 ----Next_Part(Thu_Jan__6_13_30_17_2022_597)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v14-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v22 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 49 ++ src/backend/access/transam/README | 10 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlogrecovery.c | 18 + src/backend/catalog/storage.c | 561 +++++++++++++++++++++- src/backend/commands/tablecmds.c | 267 ++++++++-- src/backend/replication/basebackup.c | 9 +- src/backend/storage/buffer/bufmgr.c | 85 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 340 +++++++++---- src/backend/storage/smgr/md.c | 95 +++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 21 +- src/bin/pg_rewind/parsexlog.c | 22 + src/bin/pg_rewind/pg_rewind.c | 1 - src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 43 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + src/tools/pgindent/typedefs.list | 7 + 25 files changed, 1485 insertions(+), 183 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index e0ee8a078a..2f92c06f70 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,46 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rlocator, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rlocator.dbOid, + xlrec->rlocator.spcOid, + xlrec->rlocator.relNumber, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action = "<none>"; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +95,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 734c39a4d0..f08bd7f42d 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +================================ +The Smgr MARK files +-------------------------------- + +An smgr mark file is an empty file that is created when a new relation +storage file is created to signal that the storage file needs to be +cleaned up at recovery time. In contrast to the four actions above, +failure to remove smgr mark files will lead to data loss, in which +case the server will shut down. + Skipping WAL for New RelFileLocator -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index 116de1175b..a1d97150dd 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2213,6 +2213,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2463,6 +2466,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2788,6 +2794,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index 5d6f1b5e46..42b4f6b5c8 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -41,6 +41,7 @@ #include "access/xlogrecovery.h" #include "access/xlogutils.h" #include "catalog/pg_control.h" +#include "catalog/storage.h" #include "commands/tablespace.h" #include "miscadmin.h" #include "pgstat.h" @@ -54,6 +55,7 @@ #include "storage/pmsignal.h" #include "storage/proc.h" #include "storage/procarray.h" +#include "storage/reinit.h" #include "storage/spin.h" #include "utils/builtins.h" #include "utils/guc.h" @@ -1773,6 +1775,14 @@ PerformWalRecovery(void) RmgrCleanup(); + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + ereport(LOG, (errmsg("redo done at %X/%X system usage: %s", LSN_FORMAT_ARGS(xlogreader->ReadRecPtr), @@ -3052,6 +3062,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index d708af19ed..bf21b35ba5 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileLocator rlocator; /* relation that need a cleanup */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + ForkNumber unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileLocator rlocator; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup * pendingCleanups = NULL; /* head of linked list */ static HTAB *pendingSyncHash = NULL; @@ -123,6 +142,7 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, SMgrRelation srel; BackendId backend; bool needs_wal; + PendingCleanup *pendingclean; Assert(!IsInParallelMode()); /* couldn't update pendingSyncHash */ @@ -145,9 +165,23 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + * We don't need this during WAL-loggged CREATE DATABASE. See + * CreateAndCopyRelationData for detail. + */ srel = smgropen(rlocator, backend); + + if (register_delete) + { + log_smgrcreatemark(&rlocator, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + } + smgrcreate(srel, MAIN_FORKNUM, false); - + if (needs_wal) log_smgrcreate(&srel->smgr_rlocator.locator, MAIN_FORKNUM); @@ -157,16 +191,29 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, */ if (register_delete) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->rlocator = rlocator; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->rlocator = rlocator; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->rlocator = rlocator; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; } if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) @@ -178,6 +225,204 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init-fork exists since before the current transaction + * started. This function reverts that change just by removing the entry. + * See RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the commit-sentinel file */ + srel = smgropen(rlocator, InvalidBackendId); + log_smgrcreatemark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * init fork for indexes needs further initialization. ambuildempty should + * do WAL-log and file sync by itself but otherwise we do that by + * ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion is canceled. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init fork is created in the current transaction. We remove + * both the init fork and mark file immediately in that case. Otherwise + * register an at-commit pending-unlink for the existing init fork. See + * RelationCreateInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + + /* + * Some AMs initializes INIT fork via buffer manager. Drop all buffers + * for the INIT fork then unlink the INIT fork along with the mark + * file. + */ + DropRelFileLocatorBuffers(srel, &forknum, 1, &firstblock); + log_smgrunlinkmark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -197,6 +442,88 @@ log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileLocator *rlocator, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator *rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = *rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -711,6 +1038,95 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * part of the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->rlocator, pending->backend); + + Assert((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + + if (pending->op & PCOP_UNLINK_FORK) + { + /* + * Unlink the fork file. Currently we use this only for + * init forks and we're sure that the init fork is not + * loaded on shared buffers. For RelationDropInitFork + * case, the function dropped that buffers. For + * RelationCreateInitFork case, PCOP_SET_PERSISTENCE(true) + * is set and the buffers have been dropped just before. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->rlocator, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->rlocator, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->rlocator, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -971,6 +1387,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rlocator, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1059,6 +1484,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index ef5b34a312..ab8ec38929 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -54,6 +54,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5372,6 +5373,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take the hard way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. We don't + * emit this if wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5502,48 +5684,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 95440013c0..0ef4b2bf01 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1167,6 +1167,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1217,7 +1218,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) @@ -1442,6 +1443,7 @@ is_checksummed_file(const char *fullpath, const char *filename) strncmp(fullpath, "/", 1) == 0) { int excludeIdx; + char *p; /* Compare file against noChecksumFiles skip list */ for (excludeIdx = 0; noChecksumFiles[excludeIdx].name != NULL; excludeIdx++) @@ -1455,6 +1457,11 @@ is_checksummed_file(const char *fullpath, const char *filename) return false; } + /* exclude mark files */ + p = strchr(filename, '.'); + if (p && isalpha(p[1]) && p[2] == 0) + return false; + return true; } else diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index e4de4b306c..e8b8b33780 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3161,6 +3161,91 @@ DropRelFileLocatorBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileLocatorBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(bufHdr->tag.forkNum != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileLocatorsAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index f904f60c08..bcdbbad0f1 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3696,7 +3694,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index f053fe0495..4c2b19ada4 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlogrecovery.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relation files cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relation file information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the storage file is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then + * the smgr object before initializing the unlogged relation. This is + * safe as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while ((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileLocatorBackend rel; + + /* + * The relation is persistent and stays persistent. Don't drop the + * buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.locator.spcOid = tspid; + rel.locator.dbOid = dbid; + rel.locator.relNumber = ent->reloid; + + srels[nrels++] = smgropen(rel.locator, InvalidBackendId); + } + + DropRelFileLocatorsAllBuffers(srels, nrels); + + for (i = 0; i < nrels; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; - unlogged_relation_entry ent; + Oid key; + relfile_entry *ent; + RelFileLocatorBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed transaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relation files. + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path)); + + rel.backend = InvalidBackendId; + rel.locator.spcOid = tspid; + rel.locator.dbOid = dbid; + rel.locator.relNumber = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 3998296a62..4249df657a 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -141,7 +141,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -173,6 +174,82 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rlocator.locator.spcOid, + reln->smgr_rlocator.locator.dbOid, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not create mark file \"%s\": %m", path)); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path)); + pfree(path); + + if (fd < 0) + return false; + + close(fd); + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1032,6 +1109,16 @@ register_forget_request(RelFileLocatorBackend rlocator, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileLocatorBackend rlocator, + ForkNumber forknum) +{ + register_forget_request(rlocator, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1390,12 +1477,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rlocator, MAIN_FORKNUM); + p = GetRelationPath(ftag->rlocator.dbOid, ftag->rlocator.spcOid, + ftag->rlocator.relNumber,InvalidBackendId, + MAIN_FORKNUM, mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index b21d8c3822..b461c0d583 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -371,6 +377,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -693,6 +719,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rlocator, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index e1fb631003..1a72dd12bb 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -91,7 +91,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -235,7 +236,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -244,6 +246,21 @@ SyncPostCheckpoint(void) * here. rmtree() also has to ignore ENOENT errors, to deal with * the possibility that we delete the file first. */ + if (errno != ENOENT) + ereport(WARNING, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path)); + } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, + path, + SMGR_MARK_UNCOMMITTED) + < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork file has been successfully removed. It's ok if the file + * does not exist. + */ if (errno != ENOENT) ereport(WARNING, (errcode_for_file_access(), diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 53f011a2fe..4da62f71ce 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -418,6 +418,28 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore there. We'll see that the file don't exist in + * the target data dir, and copy them in from the source system. No + * need to do anything special here. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these, The file will be removed from the + * target, if it doesn't exist in the source system. The files are + * empty so we don't need to bother the content. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/bin/pg_rewind/pg_rewind.c b/src/bin/pg_rewind/pg_rewind.c index 1ff8da1676..c48a60cfe1 100644 --- a/src/bin/pg_rewind/pg_rewind.c +++ b/src/bin/pg_rewind/pg_rewind.c @@ -419,7 +419,6 @@ main(int argc, char **argv) if (showprogress) pg_log_info("reading source file list"); source->traverse_files(source, &process_source_file); - if (showprogress) pg_log_info("reading target file list"); traverse_datadir(datadir_target, &process_target_file); diff --git a/src/common/relpath.c b/src/common/relpath.c index 1b6b620ce8..46cfe38fd5 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbOid, Oid spcOid) */ char * GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcOid == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, Assert(dbOid == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNumber, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNumber, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNumber); + path = psprintf("global/%u%s", relNumber, markstr); } else if (spcOid == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbOid, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbOid, relNumber); + path = psprintf("base/%u/%u%s", + dbOid, relNumber, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbOid, backendId, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbOid, backendId, relNumber); + path = psprintf("base/%u/t%d_%u%s", + dbOid, backendId, relNumber, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcOid, TABLESPACE_VERSION_DIRECTORY, dbOid, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcOid, TABLESPACE_VERSION_DIRECTORY, - dbOid, relNumber); + dbOid, relNumber, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcOid, TABLESPACE_VERSION_DIRECTORY, dbOid, backendId, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcOid, TABLESPACE_VERSION_DIRECTORY, - dbOid, backendId, relNumber); + dbOid, backendId, relNumber, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 9964c312aa..ee4179699a 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -25,6 +25,8 @@ extern PGDLLIMPORT int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileLocator rlocator, char relpersistence, bool register_delete); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileLocator rlocator, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -43,6 +45,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileLocator **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 44a5e2043b..9f48fb5e6f 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilelocator.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileLocator rlocator; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileLocator rlocator; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,13 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileLocator *rlocator, + ForkNumber forkNum, StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileLocator *rlocator, + ForkNumber forkNum, StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileLocator *rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index 3ab713247f..eba7a05f4e 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbOid, Oid spcOid); extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, /* First argument is a RelFileLocator */ #define relpathbackend(rlocator, backend, forknum) \ GetRelationPath((rlocator).dbOid, (rlocator).spcOid, (rlocator).relNumber, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileLocator */ #define relpathperm(rlocator, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, #define relpath(rlocator, forknum) \ relpathbackend((rlocator).locator, (rlocator).backend, forknum) +/* First argument is a RelFileLocatorBackend */ +#define markpath(rlocator, forknum, mark) \ + GetRelationPath((rlocator).locator.dbOid, (rlocator).locator.spcOid, \ + (rlocator).locator.relNumber, \ + (rlocator).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index 7bcfaac272..a4271ebd69 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -213,6 +213,8 @@ extern void DropRelFileLocatorBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelFileLocatorsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 2b4a8e0ffe..eb33a9ba4c 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern void SyncDataDirectory(void); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 10aa1b0109..294a09444c 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileLocatorBackend rlocator, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index bf2c10d443..e399aec0c7 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a07715356b..75b9d41e4b 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilelocator.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -87,7 +99,12 @@ extern void smgrcloseall(void); extern void smgrcloserellocator(RelFileLocatorBackend rlocator); extern void smgrrelease(SMgrRelation reln); extern void smgrreleaseall(void); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 34a76ceb60..5445826998 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -1981,6 +1981,7 @@ PatternInfoArray Pattern_Prefix_Status Pattern_Type PendingFsyncEntry +PendingMarkCleanup PendingRelDelete PendingRelSync PendingUnlinkEntry @@ -2611,6 +2612,7 @@ StdRdOptIndexCleanup StdRdOptions Step StopList +StorageMarks StrategyNumber StreamCtl String @@ -3612,6 +3614,7 @@ registered_buffer regmatch_t regoff_t regproc +relfile_entry relopt_bool relopt_enum relopt_enum_elt_def @@ -3665,6 +3668,7 @@ slist_iter slist_mutable_iter slist_node slock_t +smgr_mark_action socket_set socklen_t spgBulkDeleteState @@ -3863,8 +3867,11 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create +xl_smgr_mark xl_smgr_truncate +xl_smgr_unlink xl_standby_lock xl_standby_locks xl_tblspc_create_rec -- 2.31.1 ----Next_Part(Thu_Jul__7_17_24_59_2022_451)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="v22-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v10 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 +++ src/backend/access/transam/README | 8 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 545 ++++++++++++++++++++++++- src/backend/commands/tablecmds.c | 256 ++++++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 +++++++++++----- src/backend/storage/smgr/md.c | 93 ++++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 ++ src/common/relpath.c | 47 ++- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + 23 files changed, 1450 insertions(+), 182 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..d251f22207 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..b344bbe511 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index e7b0bc804d..b41186d6d8 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2197,6 +2197,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2447,6 +2450,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2772,6 +2778,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 1e1fbe957f..59f4c2eacf 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4564,6 +4565,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7824,6 +7833,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index c5ad28d71f..d6b30387e9 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup *pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -117,7 +136,8 @@ AddPendingSync(const RelFileNode *rnode) SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; SMgrRelation srel; BackendId backend; bool needs_wal; @@ -143,21 +163,41 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); /* Add the relation to the list of stuff to delete at abort */ - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { @@ -168,6 +208,203 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have entries for init-fork operations on this relation, that means + * that we have already registered pending delete entries to drop an + * init-fork preexisting since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* + * We are going to create an init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The mark file works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just register pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +424,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -255,6 +574,7 @@ RelationPreserveStorage(RelFileNode rnode, bool atCommit) prev->next = next; else pendingDeletes = next; + pfree(pending); /* prev does not change */ } @@ -673,6 +993,88 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * emitted before the commit record for the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert ((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -933,6 +1335,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1432,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index bf42587e38..0d9c801535 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -52,6 +52,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5329,6 +5330,177 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother to allow in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = rel->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5459,47 +5631,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index ec0485705d..45e1a5d817 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1070,6 +1070,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1120,7 +1121,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index b4532948d3..dab74bf99a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3154,6 +3155,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 263057841d..8487ae1f02 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 0ae3fb6902..0137902bb2 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index b4bca7eed6..580b74839f 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,81 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1025,6 +1101,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1378,12 +1463,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0fcef4994b..110e64b0b2 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index d4083e8a56..9563940d45 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -89,7 +89,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -222,7 +223,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -236,6 +238,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 436df54120..dbc0da5da5 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..4945b111cc 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..584ebac391 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -41,6 +43,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..12346ed7f6 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..106a5cf508 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index cfce23ecbc..f5a7df87a4 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 34602ae006..2dc0357ad5 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..99620816b5 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..e1f97e9b89 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..201ecace8a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Dec_21_20_04_55_2021_483)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v10-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v12 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 545 +++++++++++++++++- src/backend/commands/tablecmds.c | 266 +++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 +++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 +++++++---- src/backend/storage/smgr/md.c | 93 ++- src/backend/storage/smgr/smgr.c | 32 + src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 + src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + src/test/recovery/t/027_persistence_change.pl | 268 +++++++++ 24 files changed, 1728 insertions(+), 182 deletions(-) create mode 100644 src/test/recovery/t/027_persistence_change.pl diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..d251f22207 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..b344bbe511 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index e7b0bc804d..b41186d6d8 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2197,6 +2197,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2447,6 +2450,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2772,6 +2778,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 1e1fbe957f..59f4c2eacf 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4564,6 +4565,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7824,6 +7833,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index c5ad28d71f..d6b30387e9 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup *pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -117,7 +136,8 @@ AddPendingSync(const RelFileNode *rnode) SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; SMgrRelation srel; BackendId backend; bool needs_wal; @@ -143,21 +163,41 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); /* Add the relation to the list of stuff to delete at abort */ - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { @@ -168,6 +208,203 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have entries for init-fork operations on this relation, that means + * that we have already registered pending delete entries to drop an + * init-fork preexisting since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* + * We are going to create an init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The mark file works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just register pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +424,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -255,6 +574,7 @@ RelationPreserveStorage(RelFileNode rnode, bool atCommit) prev->next = next; else pendingDeletes = next; + pfree(pending); /* prev does not change */ } @@ -673,6 +993,88 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * emitted before the commit record for the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert ((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -933,6 +1335,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1432,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 45e59e3d5c..41e77e1072 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -52,6 +52,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5329,6 +5330,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5459,47 +5641,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index ec0485705d..45e1a5d817 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1070,6 +1070,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1120,7 +1121,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index b4532948d3..dab74bf99a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3154,6 +3155,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 263057841d..8487ae1f02 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 0ae3fb6902..0137902bb2 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index b4bca7eed6..580b74839f 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,81 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1025,6 +1101,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1378,12 +1463,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0fcef4994b..110e64b0b2 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index d4083e8a56..9563940d45 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -89,7 +89,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -222,7 +223,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -236,6 +238,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 436df54120..dbc0da5da5 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..4945b111cc 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..584ebac391 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -41,6 +43,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..12346ed7f6 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..106a5cf508 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index cfce23ecbc..f5a7df87a4 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 34602ae006..2dc0357ad5 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..99620816b5 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..e1f97e9b89 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..201ecace8a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/test/recovery/t/027_persistence_change.pl b/src/test/recovery/t/027_persistence_change.pl new file mode 100644 index 0000000000..38c5388093 --- /dev/null +++ b/src/test/recovery/t/027_persistence_change.pl @@ -0,0 +1,268 @@ + +# Copyright (c) 2021, PostgreSQL Global Development Group + +# Test relation persistence change +use strict; +use warnings; +use PostgreSQL::Test::Cluster; +use PostgreSQL::Test::Utils; +use Test::More; +use Test::More tests => 57; +use IPC::Run qw(pump finish timer); +use Config; + +my $data_unit = 2000; + +# Initialize primary node. +my $node = PostgreSQL::Test::Cluster->new('node'); +$node->init; +# we don't want checkpointing +$node->append_conf('postgresql.conf', qq( +checkpoint_timeout = '24h' +)); +$node->start; +create($node); + +my $relfilenodes1 = relfilenodes(); + +# correctly recover empty tables +$node->stop('immediate'); +$node->start; +insert($node, 0, $data_unit); + +# data persists after a crash +$node->stop('immediate'); +$node->start; +checkdataloss($data_unit, 'crash logged 1'); + +set_unlogged($node); +# SET UNLOGGED didn't change relfilenode +my $relfilenodes2 = relfilenodes(); +checkrelfilenodes($relfilenodes1, $relfilenodes2, 'logged->unlogged'); + +# data cleanly vanishes after a crash +$node->stop('immediate'); +$node->start; +checkdataloss(0, 'crash unlogged'); + +insert($node, 0, $data_unit); +set_logged($node); + +$node->stop('immediate'); +$node->start; +# SET LOGGED didn't change relfilenode and data survive a crash +my $relfilenodes3 = relfilenodes(); +checkrelfilenodes($relfilenodes2, $relfilenodes3, 'unlogged->logged'); +checkdataloss($data_unit, 'crash logged 2'); + +# unlogged insert -> graceful stop +set_unlogged($node); +insert($node, $data_unit, $data_unit, 0); +$node->stop; +$node->start; +checkdataloss($data_unit * 2, 'unlogged graceful restart'); + +# crash during transaction +set_logged($node); +$node->stop('immediate'); +$node->start; +insert($node, $data_unit * 2, $data_unit); + +my $h = insert($node, $data_unit * 3, $data_unit, 1); ## this is aborted +$node->stop('immediate'); + +# finishing $h stalls this case, just tear it off. +$h = undef; + +# check if indexes are working +$node->start; +check($node, $data_unit * 3, 'final'); + +sub create +{ + my ($node) = @_; + + $node->psql('postgres', qq( + CREATE TABLE t_bt (a int); + CREATE INDEX t_bt_i ON t_bt USING btree (a); + CREATE TABLE t_gin (a int[]); + CREATE INDEX t_gin_i ON t_gin USING gin (a); + CREATE TABLE t_gist (a point); + CREATE INDEX t_gist_i ON t_gist USING gist (a); + CREATE TABLE t_hash (a int); + CREATE INDEX t_hash_i ON t_hash USING hash (a); + CREATE TABLE t_brin (a int); + CREATE INDEX t_brin_i ON t_brin USING brin (a); + CREATE TABLE t_spgist (a point); + CREATE INDEX t_spgist_i ON t_spgist USING spgist (a);)); +} + + +sub insert +{ + my ($node, $st, $num, $interactive) = @_; + my $ed = $st + $num - 1; + my $query = qq(BEGIN; +INSERT INTO t_bt (SELECT i FROM generate_series($st, $ed) i); +INSERT INTO t_gin + (SELECT ARRAY[i, i * 2] FROM generate_series($st, $ed) i); +INSERT INTO t_gist + (SELECT point(i, i * 2) FROM generate_series($st, $ed) i); +INSERT INTO t_hash (SELECT i FROM generate_series($st, $ed) i); +INSERT INTO t_brin (SELECT i FROM generate_series($st, $ed) i); +INSERT INTO t_spgist + (SELECT point(i,i) FROM generate_series($st, $ed) i); +); + + if ($interactive) + { + my $in = ''; + my $out = ''; + my $timer = timer(10); + + my $h = $node->interactive_psql('postgres', \$in, \$out, $timer); + like($out, qr/psql/, "print startup banner"); + + $timer->start(10); + $in .= $query . "SELECT 'END';\n"; + pump $h until ($out =~ /\nEND/ || $timer->is_expired); + ok(($out =~ /\nEND/ && !$timer->is_expired), "inserted"); + return $h + # the trasaction is not terminated + } + else + { + $node->psql('postgres', $query . "COMMIT;"); + return undef; + } +} + +sub check +{ + my ($node, $num_data, $head) = @_; + my $st = 0; + my $ed = $num_data - 1; + + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO true; + SET enable_indexscan TO false; + SELECT COUNT(*) FROM t_bt, generate_series($st, $ed) i + WHERE a = i)), + $num_data, "$head: heap is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t_bt, generate_series($st, $ed) i + WHERE a = i)), + $num_data, "$head: btree is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t_gin, generate_series($st, $ed) i + WHERE a = ARRAY[i, i * 2];)), + $num_data, "$head: gin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t_gist, generate_series($st, $ed) i + WHERE a <@ box(point(i-0.5, i*2-0.5),point(i+0.5, i*2+0.5));)), + $num_data, "$head: gist is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t_hash, generate_series($st, $ed) i + WHERE a = i;)), + $num_data, "$head: hash is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t_brin, generate_series($st, $ed) i + WHERE a = i;)), + $num_data, "$head: brin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t_spgist, generate_series($st, $ed) i + WHERE a <@ box(point(i-0.5,i-0.5),point(i+0.5,i+0.5));)), + $num_data, "$head: spgist is not broken"); +} + +sub set_unlogged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t_bt SET UNLOGGED; + ALTER TABLE t_gin SET UNLOGGED; + ALTER TABLE t_gist SET UNLOGGED; + ALTER TABLE t_hash SET UNLOGGED; + ALTER TABLE t_brin SET UNLOGGED; + ALTER TABLE t_spgist SET UNLOGGED;)); +} + +sub set_logged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t_bt SET LOGGED; + ALTER TABLE t_gin SET LOGGED; + ALTER TABLE t_gist SET LOGGED; + ALTER TABLE t_hash SET LOGGED; + ALTER TABLE t_brin SET LOGGED; + ALTER TABLE t_spgist SET LOGGED;)); +} + +sub relfilenodes +{ + my $result = $node->safe_psql('postgres', qq( + SELECT relname, relfilenode FROM pg_class + WHERE relname + IN + (SELECT unnest(ARRAY[n, n||'_i']) + FROM unnest(ARRAY['t_bt','t_gin','t_gist','t_hash','t_brin','t_spgist']) + as n(n)); +)); + + my %relfilenodes; + + foreach my $l (split(/\n/, $result)) + { + die "unexpected format: $l" if ($l !~ /^([^|]+)\|([0-9]+)$/); + $relfilenodes{$1} = $2; + } + + return \%relfilenodes; +} + +sub checkrelfilenodes +{ + my ($rnodes1, $rnodes2, $s) = @_; + + foreach my $n (keys %{$rnodes1}) + { + if ($n eq 't_gist_i') + { + # persistence of GiST index is not changed in-place + isnt($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is changed: $n"); + } + else + { + # otherwise all relations are processed in-place + is($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is not changed: $n"); + } + } +} + +sub checkdataloss +{ + my ($expected, $s) = @_; + + foreach my $n ('t_bt','t_gin','t_gist','t_hash','t_brin','t_spgist') + { + is($node->safe_psql('postgres', "SELECT count(*) FROM $n;"), $expected, + "$s: data in table $n is in the expected state"); + } +} -- 2.27.0 ----Next_Part(Thu_Dec_23_15_01_41_2021_829)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v12-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v15 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 545 +++++++++++++++++- src/backend/commands/tablecmds.c | 266 +++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 +++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 +++++++---- src/backend/storage/smgr/md.c | 93 ++- src/backend/storage/smgr/smgr.c | 32 + src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 + src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + src/test/recovery/t/027_persistence_change.pl | 247 ++++++++ 24 files changed, 1707 insertions(+), 182 deletions(-) create mode 100644 src/test/recovery/t/027_persistence_change.pl diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..d251f22207 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..b344bbe511 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index e7b0bc804d..b41186d6d8 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2197,6 +2197,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2447,6 +2450,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2772,6 +2778,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 87cd05c945..243860fcb1 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4564,6 +4565,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7824,6 +7833,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index c5ad28d71f..d6b30387e9 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup *pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -117,7 +136,8 @@ AddPendingSync(const RelFileNode *rnode) SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; SMgrRelation srel; BackendId backend; bool needs_wal; @@ -143,21 +163,41 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); /* Add the relation to the list of stuff to delete at abort */ - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { @@ -168,6 +208,203 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have entries for init-fork operations on this relation, that means + * that we have already registered pending delete entries to drop an + * init-fork preexisting since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* + * We are going to create an init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The mark file works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just register pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +424,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -255,6 +574,7 @@ RelationPreserveStorage(RelFileNode rnode, bool atCommit) prev->next = next; else pendingDeletes = next; + pfree(pending); /* prev does not change */ } @@ -673,6 +993,88 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * emitted before the commit record for the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert ((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -933,6 +1335,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1432,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 89bc865e28..51fcf9ca5f 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -52,6 +52,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5346,6 +5347,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5476,47 +5658,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index ec0485705d..45e1a5d817 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1070,6 +1070,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1120,7 +1121,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index b4532948d3..dab74bf99a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3154,6 +3155,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 263057841d..8487ae1f02 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 0ae3fb6902..0137902bb2 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index b4bca7eed6..580b74839f 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,81 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1025,6 +1101,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1378,12 +1463,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0fcef4994b..110e64b0b2 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index d4083e8a56..9563940d45 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -89,7 +89,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -222,7 +223,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -236,6 +238,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 436df54120..dbc0da5da5 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..4945b111cc 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..584ebac391 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -41,6 +43,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..12346ed7f6 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..106a5cf508 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index cfce23ecbc..f5a7df87a4 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 34602ae006..2dc0357ad5 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..99620816b5 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..e1f97e9b89 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..201ecace8a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/test/recovery/t/027_persistence_change.pl b/src/test/recovery/t/027_persistence_change.pl new file mode 100644 index 0000000000..526b19cbda --- /dev/null +++ b/src/test/recovery/t/027_persistence_change.pl @@ -0,0 +1,247 @@ + +# Copyright (c) 2021, PostgreSQL Global Development Group + +# Test relation persistence change +use strict; +use warnings; +use PostgreSQL::Test::Cluster; +use PostgreSQL::Test::Utils; +use Test::More; +use Test::More tests => 30; +use IPC::Run qw(pump finish timer); +use Config; + +my $data_unit = 2000; + +# Initialize primary node. +my $node = PostgreSQL::Test::Cluster->new('node'); +$node->init; +# we don't want checkpointing +$node->append_conf('postgresql.conf', qq( +checkpoint_timeout = '24h' +)); +$node->start; +create($node); + +my $relfilenodes1 = relfilenodes(); + +# correctly recover empty tables +$node->stop('immediate'); +$node->start; +insert($node, 0, $data_unit, 0); + +# data persists after a crash +$node->stop('immediate'); +$node->start; +checkdataloss($data_unit, 'crash logged 1'); + +set_unlogged($node); +# SET UNLOGGED didn't change relfilenode +my $relfilenodes2 = relfilenodes(); +checkrelfilenodes($relfilenodes1, $relfilenodes2, 'logged->unlogged'); + +# data cleanly vanishes after a crash +$node->stop('immediate'); +$node->start; +checkdataloss(0, 'crash unlogged'); + +insert($node, 0, $data_unit, 0); +set_logged($node); + +$node->stop('immediate'); +$node->start; +# SET LOGGED didn't change relfilenode and data survive a crash +my $relfilenodes3 = relfilenodes(); +checkrelfilenodes($relfilenodes2, $relfilenodes3, 'unlogged->logged'); +checkdataloss($data_unit, 'crash logged 2'); + +# unlogged insert -> graceful stop +set_unlogged($node); +insert($node, $data_unit, $data_unit, 0); +$node->stop; +$node->start; +checkdataloss($data_unit * 2, 'unlogged graceful restart'); + +# crash during transaction +set_logged($node); +$node->stop('immediate'); +$node->start; +insert($node, $data_unit * 2, $data_unit, 0); + +my $h = insert($node, $data_unit * 3, $data_unit, 1); ## this is aborted +$node->stop('immediate'); + +# finishing $h stalls this case, just tear it off. +$h = undef; + +# check if indexes are working +$node->start; +# drop first half of data to reduce run time +$node->safe_psql('postgres', 'DELETE FROM t WHERE bt < ' . $data_unit * 2); +check($node, $data_unit * 2, $data_unit * 3 - 1, 'final check'); + +sub create +{ + my ($node) = @_; + + $node->psql('postgres', qq( + CREATE TABLE t (bt int, gin int[], gist point, hash int, + brin int, spgist point); + CREATE INDEX i_bt ON t USING btree (bt); + CREATE INDEX i_gin ON t USING gin (gin); + CREATE INDEX i_gist ON t USING gist (gist); + CREATE INDEX i_hash ON t USING hash (hash); + CREATE INDEX i_brin ON t USING brin (brin); + CREATE INDEX i_spgist ON t USING spgist (spgist);)); +} + + +sub insert +{ + my ($node, $st, $num, $interactive) = @_; + my $ed = $st + $num - 1; + my $query = qq(BEGIN; +INSERT INTO t + (SELECT i, ARRAY[i, i * 2], point(i, i * 2), i, i, point(i, i) + FROM generate_series($st, $ed) i); +); + + if ($interactive) + { + my $in = ''; + my $out = ''; + my $timer = timer(10); + + my $h = $node->interactive_psql('postgres', \$in, \$out, $timer); + like($out, qr/psql/, "print startup banner"); + + $in .= "$query\n"; + pump $h until ($out =~ /[\n\r]+INSERT 0 $num[\n\r]+/ || + $timer->is_expired); + ok(($out =~ /[\n\r]+INSERT 0 $num[\n\r]+/), "inserted-$st-$num"); + return $h + # the trasaction is not terminated + } + else + { + $node->psql('postgres', $query . "COMMIT;"); + return undef; + } +} + +sub check +{ + my ($node, $st, $ed, $head) = @_; + my $num_data = $ed - $st + 1; + + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO true; + SET enable_indexscan TO false; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE bt = i)), + $num_data, "$head: heap is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE bt = i)), + $num_data, "$head: btree is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE gin = ARRAY[i, i * 2];)), + $num_data, "$head: gin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE gist <@ box(point(i-0.5, i*2-0.5),point(i+0.5, i*2+0.5));)), + $num_data, "$head: gist is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE hash = i;)), + $num_data, "$head: hash is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE brin = i;)), + $num_data, "$head: brin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE spgist <@ box(point(i-0.5,i-0.5),point(i+0.5,i+0.5));)), + $num_data, "$head: spgist is not broken"); +} + +sub set_unlogged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t SET UNLOGGED; +)); +} + +sub set_logged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t SET LOGGED; +)); +} + +sub relfilenodes +{ + my $result = $node->safe_psql('postgres', qq{ + SELECT relname, relfilenode FROM pg_class + WHERE relname + IN ('t', 'i_bt','i_gin','i_gist','i_hash','i_brin','i_spgist');}); + + my %relfilenodes; + + foreach my $l (split(/\n/, $result)) + { + die "unexpected format: $l" if ($l !~ /^([^|]+)\|([0-9]+)$/); + $relfilenodes{$1} = $2; + } + + # the number must correspond to the in list above + is (scalar %relfilenodes, 7, "number of relations is correct"); + + return \%relfilenodes; +} + +sub checkrelfilenodes +{ + my ($rnodes1, $rnodes2, $s) = @_; + + foreach my $n (keys %{$rnodes1}) + { + if ($n eq 'i_gist') + { + # persistence of GiST index is not changed in-place + isnt($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is changed: $n"); + } + else + { + # otherwise all relations are processed in-place + is($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is not changed: $n"); + } + } +} + +sub checkdataloss +{ + my ($expected, $s) = @_; + + is($node->safe_psql('postgres', "SELECT count(*) FROM t;"), $expected, + "$s: data in table t is in the expected state"); +} -- 2.27.0 ----Next_Part(Fri_Jan__7_17_29_55_2022_965)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v15-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v16 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 545 +++++++++++++++++- src/backend/commands/tablecmds.c | 266 +++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 +++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 +++++++---- src/backend/storage/smgr/md.c | 94 ++- src/backend/storage/smgr/smgr.c | 32 + src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 + src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + src/test/recovery/t/027_persistence_change.pl | 263 +++++++++ 24 files changed, 1724 insertions(+), 182 deletions(-) create mode 100644 src/test/recovery/t/027_persistence_change.pl diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..d251f22207 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..b344bbe511 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index e7b0bc804d..b41186d6d8 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2197,6 +2197,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2447,6 +2450,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2772,6 +2778,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 87cd05c945..243860fcb1 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4564,6 +4565,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7824,6 +7833,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index c5ad28d71f..d6b30387e9 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup *pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -117,7 +136,8 @@ AddPendingSync(const RelFileNode *rnode) SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; SMgrRelation srel; BackendId backend; bool needs_wal; @@ -143,21 +163,41 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); /* Add the relation to the list of stuff to delete at abort */ - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { @@ -168,6 +208,203 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have entries for init-fork operations on this relation, that means + * that we have already registered pending delete entries to drop an + * init-fork preexisting since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* + * We are going to create an init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The mark file works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just register pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +424,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -255,6 +574,7 @@ RelationPreserveStorage(RelFileNode rnode, bool atCommit) prev->next = next; else pendingDeletes = next; + pfree(pending); /* prev does not change */ } @@ -673,6 +993,88 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * emitted before the commit record for the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert ((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -933,6 +1335,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1432,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 89bc865e28..51fcf9ca5f 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -52,6 +52,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5346,6 +5347,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5476,47 +5658,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index ec0485705d..45e1a5d817 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1070,6 +1070,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1120,7 +1121,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index b4532948d3..dab74bf99a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3154,6 +3155,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 263057841d..8487ae1f02 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 0ae3fb6902..0137902bb2 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index b4bca7eed6..1f3aac5bcc 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,82 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + close(fd); + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1025,6 +1102,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1378,12 +1464,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0fcef4994b..110e64b0b2 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index d4083e8a56..9563940d45 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -89,7 +89,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -222,7 +223,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -236,6 +238,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 436df54120..dbc0da5da5 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..4945b111cc 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..584ebac391 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -41,6 +43,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..12346ed7f6 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..106a5cf508 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index cfce23ecbc..f5a7df87a4 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 34602ae006..2dc0357ad5 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..99620816b5 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..e1f97e9b89 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..201ecace8a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/test/recovery/t/027_persistence_change.pl b/src/test/recovery/t/027_persistence_change.pl new file mode 100644 index 0000000000..261c4cf943 --- /dev/null +++ b/src/test/recovery/t/027_persistence_change.pl @@ -0,0 +1,263 @@ + +# Copyright (c) 2021, PostgreSQL Global Development Group + +# Test relation persistence change +use strict; +use warnings; +use PostgreSQL::Test::Cluster; +use PostgreSQL::Test::Utils; +use Test::More; +use Test::More tests => 30; +use IPC::Run qw(pump finish timer); +use Config; + +my $data_unit = 2000; + +# Initialize primary node. +my $node = PostgreSQL::Test::Cluster->new('node'); +$node->init; +# we don't want checkpointing +$node->append_conf('postgresql.conf', qq( +checkpoint_timeout = '24h' +)); +$node->start; +create($node); + +my $relfilenodes1 = relfilenodes(); + +# correctly recover empty tables +$node->stop('immediate'); +$node->start; +insert($node, 0, $data_unit, 0); + +# data persists after a crash +$node->stop('immediate'); +$node->start; +checkdataloss($data_unit, 'crash logged 1'); + +set_unlogged($node); +# SET UNLOGGED shouldn't change relfilenode +my $relfilenodes2 = relfilenodes(); +checkrelfilenodes($relfilenodes1, $relfilenodes2, 'logged->unlogged'); + +# data cleanly vanishes after a crash +$node->stop('immediate'); +$node->start; +checkdataloss(0, 'crash unlogged'); + +insert($node, 0, $data_unit, 0); +set_logged($node); + +$node->stop('immediate'); +$node->start; +# SET LOGGED shouldn't change relfilenode and data should survive the crash +my $relfilenodes3 = relfilenodes(); +checkrelfilenodes($relfilenodes2, $relfilenodes3, 'unlogged->logged'); +checkdataloss($data_unit, 'crash logged 2'); + +# unlogged insert -> graceful stop +set_unlogged($node); +insert($node, $data_unit, $data_unit, 0); +$node->stop; +$node->start; +checkdataloss($data_unit * 2, 'unlogged graceful restart'); + +# crash during transaction +set_logged($node); +$node->stop('immediate'); +$node->start; +insert($node, $data_unit * 2, $data_unit, 0); + +my $h; + +# insert(,,,1) requires IO::Pty. Skip the test if the module is not +# available, but do the insert to make the expected situation for the +# later tests. +eval { require IO::Pty; }; +if ($@) +{ + insert($node, $data_unit * 3, $data_unit, 0); + ok (1, 'SKIPPED: IO::Pty is needed'); + ok (1, 'SKIPPED: IO::Pty is needed'); +} +else +{ + $h = insert($node, $data_unit * 3, $data_unit, 1); ## this is aborted +} + +$node->stop('immediate'); + +# finishing $h stalls this case, just tear it off. +$h = undef; + +# check if indexes are working +$node->start; +# drop first half of data to reduce run time +$node->safe_psql('postgres', 'DELETE FROM t WHERE bt < ' . $data_unit * 2); +check($node, $data_unit * 2, $data_unit * 3 - 1, 'final check'); + +sub create +{ + my ($node) = @_; + + $node->psql('postgres', qq( + CREATE TABLE t (bt int, gin int[], gist point, hash int, + brin int, spgist point); + CREATE INDEX i_bt ON t USING btree (bt); + CREATE INDEX i_gin ON t USING gin (gin); + CREATE INDEX i_gist ON t USING gist (gist); + CREATE INDEX i_hash ON t USING hash (hash); + CREATE INDEX i_brin ON t USING brin (brin); + CREATE INDEX i_spgist ON t USING spgist (spgist);)); +} + + +sub insert +{ + my ($node, $st, $num, $interactive) = @_; + my $ed = $st + $num - 1; + my $query = qq(BEGIN; +INSERT INTO t + (SELECT i, ARRAY[i, i * 2], point(i, i * 2), i, i, point(i, i) + FROM generate_series($st, $ed) i); +); + + if ($interactive) + { + my $in = ''; + my $out = ''; + my $timer = timer(10); + + my $h = $node->interactive_psql('postgres', \$in, \$out, $timer); + like($out, qr/psql/, "print startup banner"); + + $in .= "$query\n"; + pump $h until ($out =~ /[\n\r]+INSERT 0 $num[\n\r]+/ || + $timer->is_expired); + ok(($out =~ /[\n\r]+INSERT 0 $num[\n\r]+/), "inserted-$st-$num"); + return $h + # the trasaction is not terminated + } + else + { + $node->psql('postgres', $query . "COMMIT;"); + return undef; + } +} + +sub check +{ + my ($node, $st, $ed, $head) = @_; + my $num_data = $ed - $st + 1; + + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO true; + SET enable_indexscan TO false; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE bt = i)), + $num_data, "$head: heap is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE bt = i)), + $num_data, "$head: btree is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE gin = ARRAY[i, i * 2];)), + $num_data, "$head: gin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE gist <@ box(point(i-0.5, i*2-0.5),point(i+0.5, i*2+0.5));)), + $num_data, "$head: gist is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE hash = i;)), + $num_data, "$head: hash is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE brin = i;)), + $num_data, "$head: brin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE spgist <@ box(point(i-0.5,i-0.5),point(i+0.5,i+0.5));)), + $num_data, "$head: spgist is not broken"); +} + +sub set_unlogged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t SET UNLOGGED; +)); +} + +sub set_logged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t SET LOGGED; +)); +} + +sub relfilenodes +{ + my $result = $node->safe_psql('postgres', qq{ + SELECT relname, relfilenode FROM pg_class + WHERE relname + IN ('t', 'i_bt','i_gin','i_gist','i_hash','i_brin','i_spgist');}); + + my %relfilenodes; + + foreach my $l (split(/\n/, $result)) + { + die "unexpected format: $l" if ($l !~ /^([^|]+)\|([0-9]+)$/); + $relfilenodes{$1} = $2; + } + + # the number must correspond to the in list above + is (scalar %relfilenodes, 7, "number of relations is correct"); + + return \%relfilenodes; +} + +sub checkrelfilenodes +{ + my ($rnodes1, $rnodes2, $s) = @_; + + foreach my $n (keys %{$rnodes1}) + { + if ($n eq 'i_gist') + { + # persistence of GiST index is not changed in-place + isnt($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is changed: $n"); + } + else + { + # otherwise all relations are processed in-place + is($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is not changed: $n"); + } + } +} + +sub checkdataloss +{ + my ($expected, $s) = @_; + + is($node->safe_psql('postgres', "SELECT count(*) FROM t;"), $expected, + "$s: data in table t is in the expected state"); +} -- 2.27.0 ----Next_Part(Fri_Jan_14_11_43_10_2022_351)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v16-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v2 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 394 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 213 ++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 164 +++++----- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 800 insertions(+), 137 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index a7c0cb1bc3..097dacfee6 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index d538f25726..f4dddbad55 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -57,9 +58,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -143,7 +151,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +171,37 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +211,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +439,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +490,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -606,43 +897,70 @@ smgrDoPendingDeletes(bool isCommit) prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + srel = smgropen(pending->relnode, pending->backend); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + BlockNumber block = 0; + RelFileNodeBackend rbnode; + + rbnode.node = pending->relnode; + rbnode.backend = InvalidBackendId; + + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + smgrclose(srel); + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } @@ -824,7 +1142,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1156,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1237,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1334,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 1fa9f19f08..45be633d9f 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -4917,6 +4917,138 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5169,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index c5e8707151..6ff46fb86d 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 8700f7f19a..80a1e61408 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -250,13 +251,12 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) errmsg("could not remove file \"%s\": %m", rm_path))); else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + elog(LOG, "unlinked file \"%s\"", rm_path); } } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(LOG, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 9889ad6ad8..32dad72ed3 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 072bdd118f..2a1d87dc33 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index ad733d1363..2a5e5fa990 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 30c38e0ca6..c2259cd7e3 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 7b21cab2e0..dcf1e605c0 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index 869cabcc0d..f6e1a74a38 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ee91b8fa26..9697449938 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index f28a842401..5d74631006 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Dec_25_09_12_52_2020_873)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v2-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v6 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 +++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 520 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 346 +++++++++++----- src/backend/storage/smgr/md.c | 92 ++++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 ++ src/common/relpath.c | 47 ++- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + 22 files changed, 1384 insertions(+), 206 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..d251f22207 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..7cf77e4a02 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +A smgr mark files is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to 4 above, failure to remove smgr mark files will lead to +data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 6f8810e149..27bbe17395 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4458,6 +4459,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7577,6 +7586,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..7302a3fad4 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,18 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (1 << 0) +#define PDOP_UNLINK_FORK (1 << 1) +#define PDOP_UNLINK_MARK (1 << 2) +#define PDOP_SET_PERSISTENCE (1 << 3) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +86,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,22 +172,48 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The SMGR_MARK_UNCOMMITED mark file works as + * the signal of that situation. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); - /* Add the relation to the list of stuff to delete at abort */ + /* + * Add the relation to the list of stuff to delete at abort. We don't + * remove the mark file at commit. It needs to persiste until the main fork + * file is actually deleted. See SyncPostCheckpoint. + */ pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_MARK; + pending->unlink_forknum = MAIN_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +223,207 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + (pending->op & PDOP_DELETE) == 0 && + (pending->unlink_forknum == INIT_FORKNUM || + (pending->op & PDOP_SET_PERSISTENCE) != 0)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_UNLINK_MARK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop mark file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + (pending->op & PDOP_DELETE) == 0 && + (pending->unlink_forknum == INIT_FORKNUM || + (pending->op & PDOP_SET_PERSISTENCE) != 0)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +443,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +538,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +941,104 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op & PDOP_DELETE) + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; + } + + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PDOP_UNLINK_MARK) + { + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1208,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1222,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1303,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1400,65 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 3349bcfaa7..4e2bceffda 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -51,6 +51,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5085,6 +5086,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5205,45 +5370,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 56cd473f9f..bc5288de05 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1255,6 +1255,7 @@ sendDir(const char *path, int basepathlen, bool sizeonly, List *tablespaces, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1305,7 +1306,7 @@ sendDir(const char *path, int basepathlen, bool sizeonly, List *tablespaces, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 852138f9c9..50674fd027 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3100,6 +3101,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 06b57ae71f..bdf6916d63 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -342,8 +342,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3647,7 +3645,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..f52d2ac199 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,228 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -273,6 +409,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -280,9 +417,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -316,15 +455,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -367,7 +509,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -398,11 +540,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 1e12cfad8e..87a777b307 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,80 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pfree(path); + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1024,6 +1099,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1377,12 +1461,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..dd3496cf51 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index 708215614d..a23c03ca3e 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -88,7 +88,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -216,7 +217,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -230,6 +232,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* And remove the list entry */ pendingUnlinks = list_delete_first(pendingUnlinks); diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 59ebac7d6a..db6b658489 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..67f24890d6 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[10]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, 10, ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..12346ed7f6 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..106a5cf508 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 328473bdc9..485c58e5f1 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -167,6 +167,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..99620816b5 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..e1f97e9b89 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..201ecace8a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Mar_25_14_08_05_2021_730)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v6-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v13 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 545 +++++++++++++++++- src/backend/commands/tablecmds.c | 266 +++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 +++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 +++++++---- src/backend/storage/smgr/md.c | 93 ++- src/backend/storage/smgr/smgr.c | 32 + src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 + src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + src/test/recovery/t/027_persistence_change.pl | 244 ++++++++ 24 files changed, 1704 insertions(+), 182 deletions(-) create mode 100644 src/test/recovery/t/027_persistence_change.pl diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..d251f22207 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..b344bbe511 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index e7b0bc804d..b41186d6d8 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2197,6 +2197,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2447,6 +2450,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2772,6 +2778,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 1e1fbe957f..59f4c2eacf 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4564,6 +4565,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7824,6 +7833,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index c5ad28d71f..d6b30387e9 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup *pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -117,7 +136,8 @@ AddPendingSync(const RelFileNode *rnode) SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; SMgrRelation srel; BackendId backend; bool needs_wal; @@ -143,21 +163,41 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); /* Add the relation to the list of stuff to delete at abort */ - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { @@ -168,6 +208,203 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have entries for init-fork operations on this relation, that means + * that we have already registered pending delete entries to drop an + * init-fork preexisting since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* + * We are going to create an init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The mark file works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just register pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +424,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -255,6 +574,7 @@ RelationPreserveStorage(RelFileNode rnode, bool atCommit) prev->next = next; else pendingDeletes = next; + pfree(pending); /* prev does not change */ } @@ -673,6 +993,88 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * emitted before the commit record for the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert ((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -933,6 +1335,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1432,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 45e59e3d5c..41e77e1072 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -52,6 +52,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5329,6 +5330,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5459,47 +5641,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index ec0485705d..45e1a5d817 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1070,6 +1070,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1120,7 +1121,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index b4532948d3..dab74bf99a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3154,6 +3155,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 263057841d..8487ae1f02 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 0ae3fb6902..0137902bb2 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index b4bca7eed6..580b74839f 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,81 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1025,6 +1101,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1378,12 +1463,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0fcef4994b..110e64b0b2 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index d4083e8a56..9563940d45 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -89,7 +89,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -222,7 +223,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -236,6 +238,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 436df54120..dbc0da5da5 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..4945b111cc 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..584ebac391 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -41,6 +43,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..12346ed7f6 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..106a5cf508 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index cfce23ecbc..f5a7df87a4 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 34602ae006..2dc0357ad5 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..99620816b5 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..e1f97e9b89 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..201ecace8a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/test/recovery/t/027_persistence_change.pl b/src/test/recovery/t/027_persistence_change.pl new file mode 100644 index 0000000000..a45bacc9b2 --- /dev/null +++ b/src/test/recovery/t/027_persistence_change.pl @@ -0,0 +1,244 @@ + +# Copyright (c) 2021, PostgreSQL Global Development Group + +# Test relation persistence change +use strict; +use warnings; +use PostgreSQL::Test::Cluster; +use PostgreSQL::Test::Utils; +use Test::More; +use Test::More tests => 25; +use IPC::Run qw(pump finish timer); +use Config; + +my $data_unit = 2000; + +# Initialize primary node. +my $node = PostgreSQL::Test::Cluster->new('node'); +$node->init; +# we don't want checkpointing +$node->append_conf('postgresql.conf', qq( +checkpoint_timeout = '24h' +)); +$node->start; +create($node); + +my $relfilenodes1 = relfilenodes(); + +# correctly recover empty tables +$node->stop('immediate'); +$node->start; +insert($node, 0, $data_unit); + +# data persists after a crash +$node->stop('immediate'); +$node->start; +checkdataloss($data_unit, 'crash logged 1'); + +set_unlogged($node); +# SET UNLOGGED didn't change relfilenode +my $relfilenodes2 = relfilenodes(); +checkrelfilenodes($relfilenodes1, $relfilenodes2, 'logged->unlogged'); + +# data cleanly vanishes after a crash +$node->stop('immediate'); +$node->start; +checkdataloss(0, 'crash unlogged'); + +insert($node, 0, $data_unit); +set_logged($node); + +$node->stop('immediate'); +$node->start; +# SET LOGGED didn't change relfilenode and data survive a crash +my $relfilenodes3 = relfilenodes(); +checkrelfilenodes($relfilenodes2, $relfilenodes3, 'unlogged->logged'); +checkdataloss($data_unit, 'crash logged 2'); + +# unlogged insert -> graceful stop +set_unlogged($node); +insert($node, $data_unit, $data_unit, 0); +$node->stop; +$node->start; +checkdataloss($data_unit * 2, 'unlogged graceful restart'); + +# crash during transaction +set_logged($node); +$node->stop('immediate'); +$node->start; +insert($node, $data_unit * 2, $data_unit); + +my $h = insert($node, $data_unit * 3, $data_unit, 1); ## this is aborted +$node->stop('immediate'); + +# finishing $h stalls this case, just tear it off. +$h = undef; + +# check if indexes are working +$node->start; +# drop first half of data to reduce run time +$node->safe_psql('postgres', 'DELETE FROM t WHERE bt < ' . $data_unit * 2); +check($node, $data_unit * 2, $data_unit * 3 - 1, 'final check'); + +sub create +{ + my ($node) = @_; + + $node->psql('postgres', qq( + CREATE TABLE t (bt int, gin int[], gist point, hash int, + brin int, spgist point); + CREATE INDEX i_bt ON t USING btree (bt); + CREATE INDEX i_gin ON t USING gin (gin); + CREATE INDEX i_gist ON t USING gist (gist); + CREATE INDEX i_hash ON t USING hash (hash); + CREATE INDEX i_brin ON t USING brin (brin); + CREATE INDEX i_spgist ON t_spgist USING spgist (spgist);)); +} + + +sub insert +{ + my ($node, $st, $num, $interactive) = @_; + my $ed = $st + $num - 1; + my $query = qq(BEGIN; +INSERT INTO t + (SELECT i, ARRAY[i, i * 2], point(i, i * 2), i, i, point(i, i) + FROM generate_series($st, $ed) i); +); + + if ($interactive) + { + my $in = ''; + my $out = ''; + my $timer = timer(10); + + my $h = $node->interactive_psql('postgres', \$in, \$out, $timer); + like($out, qr/psql/, "print startup banner"); + + $timer->start(10); + $in .= $query . "SELECT 'END';\n"; + pump $h until ($out =~ /\nEND/ || $timer->is_expired); + ok(($out =~ /\nEND/ && !$timer->is_expired), "inserted"); + return $h + # the trasaction is not terminated + } + else + { + $node->psql('postgres', $query . "COMMIT;"); + return undef; + } +} + +sub check +{ + my ($node, $st, $ed, $head) = @_; + my $num_data = $ed - $st + 1; + + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO true; + SET enable_indexscan TO false; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE bt = i)), + $num_data, "$head: heap is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE bt = i)), + $num_data, "$head: btree is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE gin = ARRAY[i, i * 2];)), + $num_data, "$head: gin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE gist <@ box(point(i-0.5, i*2-0.5),point(i+0.5, i*2+0.5));)), + $num_data, "$head: gist is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE hash = i;)), + $num_data, "$head: hash is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE brin = i;)), + $num_data, "$head: brin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE spgist <@ box(point(i-0.5,i-0.5),point(i+0.5,i+0.5));)), + $num_data, "$head: spgist is not broken"); +} + +sub set_unlogged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t SET UNLOGGED; +)); +} + +sub set_logged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t SET LOGGED; +)); +} + +sub relfilenodes +{ + my $result = $node->safe_psql('postgres', qq{ + SELECT relname, relfilenode FROM pg_class + WHERE relname + IN ('t', 'i_bt','i_gin','i_gist','i_hash','i_brin','i_spgist');}); + + my %relfilenodes; + + foreach my $l (split(/\n/, $result)) + { + die "unexpected format: $l" if ($l !~ /^([^|]+)\|([0-9]+)$/); + $relfilenodes{$1} = $2; + } + + return \%relfilenodes; +} + +sub checkrelfilenodes +{ + my ($rnodes1, $rnodes2, $s) = @_; + + foreach my $n (keys %{$rnodes1}) + { + if ($n eq 'i_gist') + { + # persistence of GiST index is not changed in-place + isnt($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is changed: $n"); + } + else + { + # otherwise all relations are processed in-place + is($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is not changed: $n"); + } + } +} + +sub checkdataloss +{ + my ($expected, $s) = @_; + + is($node->safe_psql('postgres', "SELECT count(*) FROM t;"), $expected, + "$s: data in table t is in the expected state"); +} -- 2.27.0 ----Next_Part(Thu_Dec_23_15_33_35_2021_938)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v13-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v21 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 49 ++ src/backend/access/transam/README | 9 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlogrecovery.c | 18 + src/backend/catalog/storage.c | 548 +++++++++++++++++++++- src/backend/commands/tablecmds.c | 266 +++++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 85 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 340 ++++++++++---- src/backend/storage/smgr/md.c | 94 +++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 21 +- src/bin/pg_rewind/parsexlog.c | 22 + src/bin/pg_rewind/pg_rewind.c | 1 - src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + src/tools/pgindent/typedefs.list | 7 + 25 files changed, 1464 insertions(+), 181 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7547813254..225ffbafef 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,46 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action = "<none>"; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +95,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..2ecd8c8c7c 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -725,6 +725,15 @@ then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index 3596a7d734..f48d950895 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2199,6 +2199,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2449,6 +2452,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2774,6 +2780,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index 8d2395dae2..e7786a3851 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -40,6 +40,7 @@ #include "access/xlogrecovery.h" #include "access/xlogutils.h" #include "catalog/pg_control.h" +#include "catalog/storage.h" #include "commands/tablespace.h" #include "miscadmin.h" #include "pgstat.h" @@ -53,6 +54,7 @@ #include "storage/pmsignal.h" #include "storage/proc.h" #include "storage/procarray.h" +#include "storage/reinit.h" #include "storage/spin.h" #include "utils/builtins.h" #include "utils/guc.h" @@ -1746,6 +1748,14 @@ PerformWalRecovery(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { @@ -3026,6 +3036,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 9898701a43..3607177ffb 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + ForkNumber unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup * pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -123,6 +142,7 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence, SMgrRelation srel; BackendId backend; bool needs_wal; + PendingCleanup *pendingclean; Assert(!IsInParallelMode()); /* couldn't update pendingSyncHash */ @@ -145,7 +165,14 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence, return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -157,18 +184,31 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence, */ if (register_delete) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; } + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -178,6 +218,200 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init-fork exists since before the current transaction + * started. This function reverts that change just by removing the entry. + * See RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the commit-sentinel file */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * init fork for indexes needs further initialization. ambuildempty should + * do WAL-log and file sync by itself but otherwise we do that by + * ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion is canceled. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init fork is created in the current transaction. We remove + * both the init fork and mark file immediately in that case. Otherwise + * just register a pending-unlink for the existing init fork. See + * RelationCreateInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks are never loaded to shared buffer so no point in + * dropping buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -197,6 +431,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -710,6 +1026,95 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * part of the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + + if (pending->op & PCOP_UNLINK_FORK) + { + /* + * Unlink the fork file. Currently we use this only for + * init forks and we're sure that the init fork is not + * loaded on shared buffers. For RelationDropInitFork + * case, the function dropped that buffers. For + * RelationCreateInitFork case, PCOP_SET_PERSISTENCE(true) + * is set and the buffers have been dropped just before. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -970,6 +1375,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1058,6 +1472,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 51b4a00d50..71aaf3320a 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5374,6 +5375,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. We don't + * emit this if wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5504,47 +5686,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenode will have the right persistence set, and at the + * same time ensure that the original filenode's buffers will + * get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would + * possibly result with buffers for the original filenode + * having the wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 6884cad2c0..c67bae34f5 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1188,6 +1188,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1238,7 +1239,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index d73a40c1bc..01974b71d2 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3159,6 +3159,91 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert(!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(bufHdr->tag.forkNum != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 14b77f2861..2fc9f17c28 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index f053fe0495..fa4b1c0e6e 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlogrecovery.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then + * the smgr object before initializing the unlogged relation. This is + * safe as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while ((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays persistent. Don't drop the + * buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0; i < nrels; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; - unlogged_relation_entry ent; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed transaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path)); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 879f647dbc..692508ea98 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,82 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not create mark file \"%s\": %m", path)); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path)); + pfree(path); + + if (fd < 0) + return false; + + close(fd); + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1031,6 +1108,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1384,12 +1470,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index d71a557a35..0710e8b145 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -63,6 +63,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -84,6 +88,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -337,6 +343,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -664,6 +690,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index c695d816fc..ab11600724 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -91,7 +91,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -236,7 +237,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -245,6 +247,21 @@ SyncPostCheckpoint(void) * here. rmtree() also has to ignore ENOENT errors, to deal with * the possibility that we delete the file first. */ + if (errno != ENOENT) + ereport(WARNING, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path)); + } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, + path, + SMGR_MARK_UNCOMMITTED) + < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork file has been successfully removed. It's ok if the file + * does not exist. + */ if (errno != ENOENT) ereport(WARNING, (errcode_for_file_access(), diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 49966e7b7f..c3515e5546 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -416,6 +416,28 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore there. We'll see that the file don't exist in + * the target data dir, and copy them in from the source system. No + * need to do anything special here. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these, The file will be removed from the + * target, if it doesn't exist in the source system. The files are + * empty so we don't need to bother the content. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/bin/pg_rewind/pg_rewind.c b/src/bin/pg_rewind/pg_rewind.c index b39b5c1aac..9f7235b920 100644 --- a/src/bin/pg_rewind/pg_rewind.c +++ b/src/bin/pg_rewind/pg_rewind.c @@ -413,7 +413,6 @@ main(int argc, char **argv) if (showprogress) pg_log_info("reading source file list"); source->traverse_files(source, &process_source_file); - if (showprogress) pg_log_info("reading target file list"); traverse_datadir(datadir_target, &process_target_file); diff --git a/src/common/relpath.c b/src/common/relpath.c index 636c96efd3..1c19e16fea 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 844a023b2c..a685665fab 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -25,6 +25,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence, bool register_delete); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -43,6 +45,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 622de22b03..d83fc6876e 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a4b5dc853b..a864c91614 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index a6b657f0ba..b7db0b2922 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -210,6 +210,8 @@ extern void CreateAndCopyRelationData(RelFileNode src_rnode, extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 29209e2724..8bf746bf45 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 6e46d8d96a..18b27d366b 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -24,6 +24,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); extern void mdrelease(void); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -42,12 +46,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index bf2c10d443..e399aec0c7 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index 8e3ef92cda..43b33b6b8d 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 72fafb795b..181709039c 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -1890,6 +1890,7 @@ PatternInfoArray Pattern_Prefix_Status Pattern_Type PendingFsyncEntry +PendingMarkCleanup PendingRelDelete PendingRelSync PendingUnlinkEntry @@ -2535,6 +2536,7 @@ StdRdOptIndexCleanup StdRdOptions Step StopList +StorageMarks StrategyNumber StreamCtl StreamXidHash @@ -3502,6 +3504,7 @@ registered_buffer regmatch_t regoff_t regproc +relfile_entry relopt_bool relopt_enum relopt_enum_elt_def @@ -3555,6 +3558,7 @@ slist_iter slist_mutable_iter slist_node slock_t +smgr_mark_action socket_set spgBulkDeleteState spgChooseIn @@ -3755,8 +3759,11 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create +xl_smgr_mark xl_smgr_truncate +xl_smgr_unlink xl_standby_lock xl_standby_locks xl_tblspc_create_rec -- 2.27.0 ----Next_Part(Thu_Mar_31_18_33_18_2022_007)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="v21-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v17 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 545 +++++++++++++++++- src/backend/commands/tablecmds.c | 266 +++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 +++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 +++++++---- src/backend/storage/smgr/md.c | 94 ++- src/backend/storage/smgr/smgr.c | 32 + src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 + src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + src/test/recovery/t/027_persistence_change.pl | 263 +++++++++ 24 files changed, 1724 insertions(+), 182 deletions(-) create mode 100644 src/test/recovery/t/027_persistence_change.pl diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7547813254..2c674e5de0 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..b344bbe511 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index c9516e03fa..3c7010eb0f 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2197,6 +2197,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2447,6 +2450,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2772,6 +2778,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index c9d4cbf3ff..7cab6a0170 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4564,6 +4565,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7824,6 +7833,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 9b8075536a..92a9451e90 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup *pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -117,7 +136,8 @@ AddPendingSync(const RelFileNode *rnode) SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; SMgrRelation srel; BackendId backend; bool needs_wal; @@ -143,21 +163,41 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); /* Add the relation to the list of stuff to delete at abort */ - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { @@ -168,6 +208,203 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have entries for init-fork operations on this relation, that means + * that we have already registered pending delete entries to drop an + * init-fork preexisting since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* + * We are going to create an init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The mark file works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just register pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +424,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -255,6 +574,7 @@ RelationPreserveStorage(RelFileNode rnode, bool atCommit) prev->next = next; else pendingDeletes = next; + pfree(pending); /* prev does not change */ } @@ -673,6 +993,88 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * emitted before the commit record for the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert ((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -933,6 +1335,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1432,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 1f0654c2f5..9e673ba68f 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -52,6 +52,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5346,6 +5347,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5476,47 +5658,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 3afbbe7e02..3f16b5f58c 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1102,6 +1102,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1152,7 +1153,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index a2512e750c..6384b4efbe 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3154,6 +3155,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 14b77f2861..2fc9f17c28 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index f053fe0495..1124e95d0d 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index d26c915f90..007efe68a5 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,82 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + close(fd); + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1025,6 +1102,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1378,12 +1464,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index eb701dce57..4819b5c404 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index 11fa17ddea..ddc344dad2 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -89,7 +89,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -222,7 +223,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -236,6 +238,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 9143797458..b21d01d04a 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 636c96efd3..1c19e16fea 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 9ffc741913..d362d62ed2 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -41,6 +43,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 622de22b03..8139308634 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a4b5dc853b..a864c91614 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index dd01841c30..739b386216 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 29209e2724..8bf746bf45 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index ffffa40db7..046afdb5fb 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index bf2c10d443..e399aec0c7 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index 052e0b8426..48e69ab69b 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/test/recovery/t/027_persistence_change.pl b/src/test/recovery/t/027_persistence_change.pl new file mode 100644 index 0000000000..261c4cf943 --- /dev/null +++ b/src/test/recovery/t/027_persistence_change.pl @@ -0,0 +1,263 @@ + +# Copyright (c) 2021, PostgreSQL Global Development Group + +# Test relation persistence change +use strict; +use warnings; +use PostgreSQL::Test::Cluster; +use PostgreSQL::Test::Utils; +use Test::More; +use Test::More tests => 30; +use IPC::Run qw(pump finish timer); +use Config; + +my $data_unit = 2000; + +# Initialize primary node. +my $node = PostgreSQL::Test::Cluster->new('node'); +$node->init; +# we don't want checkpointing +$node->append_conf('postgresql.conf', qq( +checkpoint_timeout = '24h' +)); +$node->start; +create($node); + +my $relfilenodes1 = relfilenodes(); + +# correctly recover empty tables +$node->stop('immediate'); +$node->start; +insert($node, 0, $data_unit, 0); + +# data persists after a crash +$node->stop('immediate'); +$node->start; +checkdataloss($data_unit, 'crash logged 1'); + +set_unlogged($node); +# SET UNLOGGED shouldn't change relfilenode +my $relfilenodes2 = relfilenodes(); +checkrelfilenodes($relfilenodes1, $relfilenodes2, 'logged->unlogged'); + +# data cleanly vanishes after a crash +$node->stop('immediate'); +$node->start; +checkdataloss(0, 'crash unlogged'); + +insert($node, 0, $data_unit, 0); +set_logged($node); + +$node->stop('immediate'); +$node->start; +# SET LOGGED shouldn't change relfilenode and data should survive the crash +my $relfilenodes3 = relfilenodes(); +checkrelfilenodes($relfilenodes2, $relfilenodes3, 'unlogged->logged'); +checkdataloss($data_unit, 'crash logged 2'); + +# unlogged insert -> graceful stop +set_unlogged($node); +insert($node, $data_unit, $data_unit, 0); +$node->stop; +$node->start; +checkdataloss($data_unit * 2, 'unlogged graceful restart'); + +# crash during transaction +set_logged($node); +$node->stop('immediate'); +$node->start; +insert($node, $data_unit * 2, $data_unit, 0); + +my $h; + +# insert(,,,1) requires IO::Pty. Skip the test if the module is not +# available, but do the insert to make the expected situation for the +# later tests. +eval { require IO::Pty; }; +if ($@) +{ + insert($node, $data_unit * 3, $data_unit, 0); + ok (1, 'SKIPPED: IO::Pty is needed'); + ok (1, 'SKIPPED: IO::Pty is needed'); +} +else +{ + $h = insert($node, $data_unit * 3, $data_unit, 1); ## this is aborted +} + +$node->stop('immediate'); + +# finishing $h stalls this case, just tear it off. +$h = undef; + +# check if indexes are working +$node->start; +# drop first half of data to reduce run time +$node->safe_psql('postgres', 'DELETE FROM t WHERE bt < ' . $data_unit * 2); +check($node, $data_unit * 2, $data_unit * 3 - 1, 'final check'); + +sub create +{ + my ($node) = @_; + + $node->psql('postgres', qq( + CREATE TABLE t (bt int, gin int[], gist point, hash int, + brin int, spgist point); + CREATE INDEX i_bt ON t USING btree (bt); + CREATE INDEX i_gin ON t USING gin (gin); + CREATE INDEX i_gist ON t USING gist (gist); + CREATE INDEX i_hash ON t USING hash (hash); + CREATE INDEX i_brin ON t USING brin (brin); + CREATE INDEX i_spgist ON t USING spgist (spgist);)); +} + + +sub insert +{ + my ($node, $st, $num, $interactive) = @_; + my $ed = $st + $num - 1; + my $query = qq(BEGIN; +INSERT INTO t + (SELECT i, ARRAY[i, i * 2], point(i, i * 2), i, i, point(i, i) + FROM generate_series($st, $ed) i); +); + + if ($interactive) + { + my $in = ''; + my $out = ''; + my $timer = timer(10); + + my $h = $node->interactive_psql('postgres', \$in, \$out, $timer); + like($out, qr/psql/, "print startup banner"); + + $in .= "$query\n"; + pump $h until ($out =~ /[\n\r]+INSERT 0 $num[\n\r]+/ || + $timer->is_expired); + ok(($out =~ /[\n\r]+INSERT 0 $num[\n\r]+/), "inserted-$st-$num"); + return $h + # the trasaction is not terminated + } + else + { + $node->psql('postgres', $query . "COMMIT;"); + return undef; + } +} + +sub check +{ + my ($node, $st, $ed, $head) = @_; + my $num_data = $ed - $st + 1; + + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO true; + SET enable_indexscan TO false; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE bt = i)), + $num_data, "$head: heap is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE bt = i)), + $num_data, "$head: btree is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE gin = ARRAY[i, i * 2];)), + $num_data, "$head: gin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE gist <@ box(point(i-0.5, i*2-0.5),point(i+0.5, i*2+0.5));)), + $num_data, "$head: gist is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE hash = i;)), + $num_data, "$head: hash is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE brin = i;)), + $num_data, "$head: brin is not broken"); + is($node->safe_psql('postgres', qq( + SET enable_seqscan TO false; + SET enable_indexscan TO true; + SELECT COUNT(*) FROM t, generate_series($st, $ed) i + WHERE spgist <@ box(point(i-0.5,i-0.5),point(i+0.5,i+0.5));)), + $num_data, "$head: spgist is not broken"); +} + +sub set_unlogged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t SET UNLOGGED; +)); +} + +sub set_logged +{ + my ($node) = @_; + + $node->psql('postgres', qq( + ALTER TABLE t SET LOGGED; +)); +} + +sub relfilenodes +{ + my $result = $node->safe_psql('postgres', qq{ + SELECT relname, relfilenode FROM pg_class + WHERE relname + IN ('t', 'i_bt','i_gin','i_gist','i_hash','i_brin','i_spgist');}); + + my %relfilenodes; + + foreach my $l (split(/\n/, $result)) + { + die "unexpected format: $l" if ($l !~ /^([^|]+)\|([0-9]+)$/); + $relfilenodes{$1} = $2; + } + + # the number must correspond to the in list above + is (scalar %relfilenodes, 7, "number of relations is correct"); + + return \%relfilenodes; +} + +sub checkrelfilenodes +{ + my ($rnodes1, $rnodes2, $s) = @_; + + foreach my $n (keys %{$rnodes1}) + { + if ($n eq 'i_gist') + { + # persistence of GiST index is not changed in-place + isnt($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is changed: $n"); + } + else + { + # otherwise all relations are processed in-place + is($rnodes1->{$n}, $rnodes2->{$n}, + "$s: relfilenode is not changed: $n"); + } + } +} + +sub checkdataloss +{ + my ($expected, $s) = @_; + + is($node->safe_psql('postgres', "SELECT count(*) FROM t;"), $expected, + "$s: data in table t is in the expected state"); +} -- 2.27.0 ----Next_Part(Wed_Jan_19_09_39_07_2022_989)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v17-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v18 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 49 ++ src/backend/access/transam/README | 9 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlogrecovery.c | 18 + src/backend/catalog/storage.c | 548 +++++++++++++++++++++- src/backend/commands/tablecmds.c | 266 +++++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 86 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 ++++++++++---- src/backend/storage/smgr/md.c | 94 +++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 22 + src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + 23 files changed, 1459 insertions(+), 182 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7547813254..f8908e2c0a 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,46 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +95,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..2ecd8c8c7c 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -725,6 +725,15 @@ then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index adf763a8ea..559666b802 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2198,6 +2198,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2448,6 +2451,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2773,6 +2779,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index f9f212680b..2923b8ef8c 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -40,6 +40,7 @@ #include "access/xlogrecovery.h" #include "access/xlogutils.h" #include "catalog/pg_control.h" +#include "catalog/storage.h" #include "commands/tablespace.h" #include "miscadmin.h" #include "pgstat.h" @@ -53,6 +54,7 @@ #include "storage/pmsignal.h" #include "storage/proc.h" #include "storage/procarray.h" +#include "storage/reinit.h" #include "storage/spin.h" #include "utils/builtins.h" #include "utils/guc.h" @@ -1746,6 +1748,14 @@ PerformWalRecovery(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { @@ -3022,6 +3032,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 9b8075536a..cd1445713a 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup *pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -117,7 +136,8 @@ AddPendingSync(const RelFileNode *rnode) SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; SMgrRelation srel; BackendId backend; bool needs_wal; @@ -143,21 +163,41 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); /* Add the relation to the list of stuff to delete at abort */ - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { @@ -168,6 +208,200 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init-fork exists since before the current transaction + * started. This function reverts that change just by removing the entry. + * See RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the commit-sentinel file */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * init fork for indexes needs further initialization. ambuildempty should + * do WAL-log and file sync by itself but otherwise we do that by + * ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion is canceled. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init fork is created in the current transaction. We remove + * both the init fork and mark file immediately in that case. Otherwise + * just register a pending-unlink for the existing init fork. See + * RelationCreateInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks are never loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +421,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -673,6 +989,95 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * part of the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert ((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + + if (pending->op & PCOP_UNLINK_FORK) + { + /* + * Unlink the fork file. Currently we use this only for + * init forks and we're sure that the init fork is not + * loaded on shared buffers. For RelationDropInitFork + * case, the function dropped that buffers. For + * RelationCreateInitFork case, PCOP_SET_PERSISTENCE(true) + * is set and the buffers have been dropped just before. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -933,6 +1338,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1435,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 3e83f375b5..9e5b77e94a 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -53,6 +53,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5347,6 +5348,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5477,47 +5659,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 0bf28b55d7..17185f4e55 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1209,6 +1209,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1259,7 +1260,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index f5459c68f8..6cd010429a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -38,6 +38,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3155,6 +3156,91 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(bufHdr->tag.forkNum != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 14b77f2861..2fc9f17c28 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index f053fe0495..f28f55baa6 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlogrecovery.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 879f647dbc..4d44bdd78b 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,82 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + close(fd); + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1031,6 +1108,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1384,12 +1470,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index d71a557a35..0710e8b145 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -63,6 +63,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -84,6 +88,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -337,6 +343,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -664,6 +690,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index e161d57761..f5ded7cb34 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -90,7 +90,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -223,7 +224,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -237,6 +239,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 56df08c64f..f1382d4c4f 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,28 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore there. We'll see that the file don't exist in + * the target data dir, and copy them in from the source system. No + * need to do anything special here. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these, The file will be removed from the + * target, if it doesn't exist in the source system. The files are + * empty so we don't need to bother the content. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 636c96efd3..1c19e16fea 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 9ffc741913..d362d62ed2 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -41,6 +43,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 622de22b03..8139308634 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a4b5dc853b..a864c91614 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index dd01841c30..739b386216 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 29209e2724..8bf746bf45 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 6e46d8d96a..ef5fdaf4f8 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -24,6 +24,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); extern void mdrelease(void); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -42,12 +46,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index bf2c10d443..e399aec0c7 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index 8e3ef92cda..022654b7b2 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Mar__1_14_14_13_2022_434)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v18-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v20 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 49 ++ src/backend/access/transam/README | 9 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlogrecovery.c | 18 + src/backend/catalog/storage.c | 548 +++++++++++++++++++++- src/backend/commands/tablecmds.c | 266 +++++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 85 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 340 ++++++++++---- src/backend/storage/smgr/md.c | 94 +++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 21 +- src/bin/pg_rewind/parsexlog.c | 22 + src/bin/pg_rewind/pg_rewind.c | 1 - src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + src/tools/pgindent/typedefs.list | 7 + 25 files changed, 1464 insertions(+), 181 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7547813254..225ffbafef 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,46 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action = "<none>"; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +95,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..2ecd8c8c7c 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -725,6 +725,15 @@ then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index 3596a7d734..f48d950895 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2199,6 +2199,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2449,6 +2452,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2774,6 +2780,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index 8d2395dae2..e7786a3851 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -40,6 +40,7 @@ #include "access/xlogrecovery.h" #include "access/xlogutils.h" #include "catalog/pg_control.h" +#include "catalog/storage.h" #include "commands/tablespace.h" #include "miscadmin.h" #include "pgstat.h" @@ -53,6 +54,7 @@ #include "storage/pmsignal.h" #include "storage/proc.h" #include "storage/procarray.h" +#include "storage/reinit.h" #include "storage/spin.h" #include "utils/builtins.h" #include "utils/guc.h" @@ -1746,6 +1748,14 @@ PerformWalRecovery(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { @@ -3026,6 +3036,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 9898701a43..ab8ec34c3d 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + ForkNumber unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup * pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -145,7 +164,14 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence, return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -157,16 +183,30 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence, */ if (register_delete) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; } if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) @@ -178,6 +218,200 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init-fork exists since before the current transaction + * started. This function reverts that change just by removing the entry. + * See RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the commit-sentinel file */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * init fork for indexes needs further initialization. ambuildempty should + * do WAL-log and file sync by itself but otherwise we do that by + * ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion is canceled. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init fork is created in the current transaction. We remove + * both the init fork and mark file immediately in that case. Otherwise + * just register a pending-unlink for the existing init fork. See + * RelationCreateInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks are never loaded to shared buffer so no point in + * dropping buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -197,6 +431,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -710,6 +1026,95 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * part of the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + + if (pending->op & PCOP_UNLINK_FORK) + { + /* + * Unlink the fork file. Currently we use this only for + * init forks and we're sure that the init fork is not + * loaded on shared buffers. For RelationDropInitFork + * case, the function dropped that buffers. For + * RelationCreateInitFork case, PCOP_SET_PERSISTENCE(true) + * is set and the buffers have been dropped just before. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -970,6 +1375,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1058,6 +1472,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 51b4a00d50..71aaf3320a 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5374,6 +5375,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. We don't + * emit this if wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5504,47 +5686,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenode will have the right persistence set, and at the + * same time ensure that the original filenode's buffers will + * get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would + * possibly result with buffers for the original filenode + * having the wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 6884cad2c0..c67bae34f5 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1188,6 +1188,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1238,7 +1239,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index d73a40c1bc..01974b71d2 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3159,6 +3159,91 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert(!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(bufHdr->tag.forkNum != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 14b77f2861..2fc9f17c28 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index f053fe0495..fa4b1c0e6e 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlogrecovery.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then + * the smgr object before initializing the unlogged relation. This is + * safe as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while ((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays persistent. Don't drop the + * buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0; i < nrels; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; - unlogged_relation_entry ent; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed transaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path)); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 879f647dbc..692508ea98 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,82 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not create mark file \"%s\": %m", path)); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path)); + pfree(path); + + if (fd < 0) + return false; + + close(fd); + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1031,6 +1108,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1384,12 +1470,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index d71a557a35..0710e8b145 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -63,6 +63,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -84,6 +88,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -337,6 +343,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -664,6 +690,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index c695d816fc..ab11600724 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -91,7 +91,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -236,7 +237,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -245,6 +247,21 @@ SyncPostCheckpoint(void) * here. rmtree() also has to ignore ENOENT errors, to deal with * the possibility that we delete the file first. */ + if (errno != ENOENT) + ereport(WARNING, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path)); + } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, + path, + SMGR_MARK_UNCOMMITTED) + < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork file has been successfully removed. It's ok if the file + * does not exist. + */ if (errno != ENOENT) ereport(WARNING, (errcode_for_file_access(), diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 49966e7b7f..c3515e5546 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -416,6 +416,28 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore there. We'll see that the file don't exist in + * the target data dir, and copy them in from the source system. No + * need to do anything special here. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these, The file will be removed from the + * target, if it doesn't exist in the source system. The files are + * empty so we don't need to bother the content. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/bin/pg_rewind/pg_rewind.c b/src/bin/pg_rewind/pg_rewind.c index b39b5c1aac..9f7235b920 100644 --- a/src/bin/pg_rewind/pg_rewind.c +++ b/src/bin/pg_rewind/pg_rewind.c @@ -413,7 +413,6 @@ main(int argc, char **argv) if (showprogress) pg_log_info("reading source file list"); source->traverse_files(source, &process_source_file); - if (showprogress) pg_log_info("reading target file list"); traverse_datadir(datadir_target, &process_target_file); diff --git a/src/common/relpath.c b/src/common/relpath.c index 636c96efd3..1c19e16fea 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 844a023b2c..a685665fab 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -25,6 +25,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence, bool register_delete); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -43,6 +45,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 622de22b03..d83fc6876e 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a4b5dc853b..a864c91614 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index a6b657f0ba..b7db0b2922 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -210,6 +210,8 @@ extern void CreateAndCopyRelationData(RelFileNode src_rnode, extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 29209e2724..8bf746bf45 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 6e46d8d96a..18b27d366b 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -24,6 +24,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); extern void mdrelease(void); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -42,12 +46,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index bf2c10d443..e399aec0c7 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index 8e3ef92cda..43b33b6b8d 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 72fafb795b..181709039c 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -1890,6 +1890,7 @@ PatternInfoArray Pattern_Prefix_Status Pattern_Type PendingFsyncEntry +PendingMarkCleanup PendingRelDelete PendingRelSync PendingUnlinkEntry @@ -2535,6 +2536,7 @@ StdRdOptIndexCleanup StdRdOptions Step StopList +StorageMarks StrategyNumber StreamCtl StreamXidHash @@ -3502,6 +3504,7 @@ registered_buffer regmatch_t regoff_t regproc +relfile_entry relopt_bool relopt_enum relopt_enum_elt_def @@ -3555,6 +3558,7 @@ slist_iter slist_mutable_iter slist_node slock_t +smgr_mark_action socket_set spgBulkDeleteState spgChooseIn @@ -3755,8 +3759,11 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create +xl_smgr_mark xl_smgr_truncate +xl_smgr_unlink xl_standby_lock xl_standby_locks xl_tblspc_create_rec -- 2.27.0 ----Next_Part(Thu_Mar_31_13_58_45_2022_151)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="v20-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/catalog/storage.c | 355 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 217 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 206 ++++++++------ src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 16 ++ src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 4 + src/include/storage/smgr.h | 1 + 12 files changed, 784 insertions(+), 142 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index a7c0cb1bc3..097dacfee6 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index d538f25726..0f1649758f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -57,9 +58,19 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ + +/* This is bit-map, not ordianal numbers */ +#define PDOP_DELETE 0x00 +#define PDOP_UNLINK_FORK 0x01 +#define PDOP_SET_PERSISTENCE 0x02 + + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -153,6 +164,7 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -168,6 +180,209 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* We don't have existing init fork, create it. */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* + * We have created the init fork. If server crashes before the current + * transaction ends the init fork left alone corrupts data while recovery. + * The inittmp fork works as the sentinel to identify that situaton. + */ + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit*/ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + smgrclose(srel); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +402,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +453,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -606,43 +860,68 @@ smgrDoPendingDeletes(bool isCommit) prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + srel = smgropen(pending->relnode, pending->backend); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + BlockNumber block = 0; + RelFileNodeBackend rbnode; + + rbnode.node = pending->relnode; + rbnode.backend = InvalidBackendId; + + DropRelFileNodeBuffers(rbnode, &pending->unlink_forknum, 1, + &block); + smgrclose(srel); + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } @@ -824,7 +1103,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1117,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1198,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrclose(reln); + smgrunlink(reln, xlrec->forkNum, true); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1295,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index e3cfaf8b07..29f786142a 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -4916,6 +4916,142 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, tab->afterStmts = list_concat(tab->afterStmts, afterStmts); return newcmd; +} + +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, lockmode); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * alredy flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recovery the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); + + + + } /* @@ -5038,45 +5174,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index ad0d1a9abc..ddd0133cdf 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3033,6 +3034,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when swithing to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 0c2094f766..6524262a74 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,6 +31,7 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { char oid[OIDCHARS + 1]; + bool dirty; } unlogged_relation_entry; /* @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,62 +163,73 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(unlogged_relation_entry); + ctl.entrysize = sizeof(unlogged_relation_entry); + hash = hash_create("unlogged hash", 32, &ctl, HASH_ELEM); + + /* Scan the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + ForkNumber forkNum; + int oidchars; + bool found; + unlogged_relation_entry key; + unlogged_relation_entry *ent; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Also skip it unless this is the init fork. */ + if (forkNum != INIT_FORKNUM && forkNum != INITTMP_FORKNUM) + continue; + + /* + * Put the OID portion of the name into the hash table, if it + * isn't already. + */ + memset(key.oid, 0, sizeof(key.oid)); + memcpy(key.oid, de->d_name, oidchars); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = 0; + + /* + * If we have the inittmp fork, the transaction that created the + * corresponding init file was not committed nor aborted. Mark this + * init fork as dirty so that we can clean up them properly. + */ + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* + * If we didn't find any init forks, there's no point in continuing; + * we can bail out now. + */ + if (hash_get_num_entries(hash) == 0) + { + hash_destroy(hash); + return; + } + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - memset(&ctl, 0, sizeof(ctl)); - ctl.keysize = sizeof(unlogged_relation_entry); - ctl.entrysize = sizeof(unlogged_relation_entry); - hash = hash_create("unlogged hash", 32, &ctl, HASH_ELEM); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - memset(ent.oid, 0, sizeof(ent.oid)); - memcpy(ent.oid, de->d_name, oidchars); - hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ @@ -224,39 +238,48 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) { ForkNumber forkNum; int oidchars; - bool found; - unlogged_relation_entry ent; + unlogged_relation_entry key; + unlogged_relation_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. */ - memset(ent.oid, 0, sizeof(ent.oid)); - memcpy(ent.oid, de->d_name, oidchars); - hash_search(hash, &ent, HASH_FIND, &found); + memset(key.oid, 0, sizeof(key.oid)); + memcpy(key.oid, de->d_name, oidchars); + ent = hash_search(hash, &key, HASH_FIND, NULL); - /* If so, nuke it! */ - if (found) + /* Don't remove files if corresponding init fork is not found */ + if (!ent) + continue; + + if (!ent->dirty) + { + /* Don't remove clean init file */ + if (forkNum == INIT_FORKNUM) + continue; + }else { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); - else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + /* Remove dirty init file, together with inittmp file */ + if (forkNum != INIT_FORKNUM && forkNum != INITTMP_FORKNUM) + continue; } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); } /* Cleanup is complete. */ @@ -273,6 +296,9 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ if ((op & UNLOGGED_RELATION_INIT) != 0) { + unlogged_relation_entry key; + unlogged_relation_entry *ent; + /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) @@ -288,6 +314,38 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + memset(key.oid, 0, sizeof(key.oid)); + memcpy(key.oid, de->d_name, oidchars); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* Don't init file that doesn't have the init fork. */ + if (!ent) + continue; + + if (ent->dirty && + (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM)) + { + /* + * The init file is dirty. The files has been removed once at + * cleanup time but recovery can create them again. Remove both + * INIT and INITTMP files. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index dcc09df0c7..5eb9e97b3d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -645,6 +645,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index ad733d1363..2a5e5fa990 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 30c38e0ca6..c2259cd7e3 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 7b21cab2e0..d48b5288ce 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -29,6 +29,8 @@ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +38,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +65,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index 869cabcc0d..f6e1a74a38 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ee91b8fa26..e2496ed1c8 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -168,6 +168,8 @@ extern PGDLLIMPORT int32 *LocalRefCount; */ #define BufferGetPage(buffer) ((Page)BufferGetBlock(buffer)) +struct SmgrRelationData; + /* * prototypes for functions in bufmgr.c */ @@ -205,6 +207,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index f28a842401..5d74631006 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.18.4 ----Next_Part(Fri_Nov_13_13_22_01_2020_316)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v11 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 +++ src/backend/access/transam/README | 8 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 545 ++++++++++++++++++++++++- src/backend/commands/tablecmds.c | 265 ++++++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 344 +++++++++++----- src/backend/storage/smgr/md.c | 93 ++++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 ++ src/common/relpath.c | 47 ++- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + 23 files changed, 1459 insertions(+), 182 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..d251f22207 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..b344bbe511 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index e7b0bc804d..b41186d6d8 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2197,6 +2197,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2447,6 +2450,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2772,6 +2778,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 1e1fbe957f..59f4c2eacf 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4564,6 +4565,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7824,6 +7833,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index c5ad28d71f..d6b30387e9 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileNode rnode; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup *pendingCleanups = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; @@ -117,7 +136,8 @@ AddPendingSync(const RelFileNode *rnode) SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; + PendingCleanup *pendingclean; SMgrRelation srel; BackendId backend; bool needs_wal; @@ -143,21 +163,41 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); /* Add the relation to the list of stuff to delete at abort */ - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->relnode = rnode; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->relnode = rnode; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->relnode = rnode; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { @@ -168,6 +208,203 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have entries for init-fork operations on this relation, that means + * that we have already registered pending delete entries to drop an + * init-fork preexisting since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* + * We are going to create an init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The mark file works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just register pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +424,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -255,6 +574,7 @@ RelationPreserveStorage(RelFileNode rnode, bool atCommit) prev->next = next; else pendingDeletes = next; + pfree(pending); /* prev does not change */ } @@ -673,6 +993,88 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * emitted before the commit record for the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->relnode, pending->backend); + + Assert ((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->relnode, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -933,6 +1335,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1432,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileNodeEquals(xlrec->rnode, pending->relnode) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert (pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->relnode = xlrec->rnode; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index bf42587e38..451ed9adb1 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -52,6 +52,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5329,6 +5330,186 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take this way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + * We don't emit this fhile wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rnode = r->rd_node; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5459,47 +5640,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index ec0485705d..45e1a5d817 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1070,6 +1070,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1120,7 +1121,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index b4532948d3..dab74bf99a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3154,6 +3155,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 263057841d..8487ae1f02 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 0ae3fb6902..0137902bb2 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index b4bca7eed6..580b74839f 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,81 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1025,6 +1101,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1378,12 +1463,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0fcef4994b..110e64b0b2 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index d4083e8a56..9563940d45 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -89,7 +89,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -222,7 +223,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -236,6 +238,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 436df54120..dbc0da5da5 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..4945b111cc 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..584ebac391 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -41,6 +43,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..12346ed7f6 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..106a5cf508 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index cfce23ecbc..f5a7df87a4 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 34602ae006..2dc0357ad5 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..99620816b5 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..e1f97e9b89 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..201ecace8a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Wed_Dec_22_15_13_27_2021_833)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v11-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v3 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 10 + src/backend/catalog/storage.c | 420 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 ++++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 ++++++ src/backend/storage/file/reinit.c | 162 ++++++---- src/backend/storage/smgr/md.c | 4 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 3 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 5 +- src/include/storage/bufmgr.h | 2 + src/include/storage/smgr.h | 1 + 14 files changed, 854 insertions(+), 140 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..51616b2458 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The INITTMP fork file +-------------------------------- + +An INITTMP fork is created when new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. The file is +removed at transaction end but is left when the process crashes before +the transaction ends. In contrast to 4 above, failure to remove an +INITTMP file will lead to data loss, in which case the server will +shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..bd9680583b 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The inittmp fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,215 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending sync entries to drop preexisting + * init fork since before the current transaction started. This function + * reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The inittmp fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, INITTMP_FORKNUM, false); + log_smgrcreate(&rnode, INITTMP_FORKNUM); + smgrimmedsync(srel, INITTMP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop inittmp fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INITTMP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and inittmp forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/INITTMP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, INITTMP_FORKNUM); + smgrunlink(srel, INITTMP_FORKNUM, false); + smgrclose(srel); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +446,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +497,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +900,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; + prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == INITTMP_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + false); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1160,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1174,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1255,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1352,15 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..adcb54b0fa 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -31,7 +31,8 @@ static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool dirty; /* to be removed */ +} relfile_entry; /* * Reset unlogged relations from before the last restart. @@ -151,6 +152,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); @@ -160,88 +163,86 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("inittmp hash", 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect inttmp forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + /* Record init and inittmp forks */ + if (forkNum == INIT_FORKNUM || forkNum == INITTMP_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode. If it has INITTMP fork, the all files + * needs to be cleaned up. Otherwise the relfilenode is cleaned up + * according to the unloggedness. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + ent->dirty = false; + + if (forkNum == INITTMP_FORKNUM) + ent->dirty = true; + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - /* * Now, make a second pass and remove anything that matches. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) { + /* so, nuke it! */ snprintf(rm_path, sizeof(rm_path), "%s/%s", dbspacedirname, de->d_name); if (unlink(rm_path) < 0) @@ -256,7 +257,6 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* @@ -277,12 +277,42 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; char dstpath[MAXPGPATH]; + Oid key; + relfile_entry *ent; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; + /* + * See whether the OID portion of the name shows up in the hash + * table. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + /* we don't remove clean init file */ + if (ent && (ent->dirty || forkNum != INIT_FORKNUM)) + { + /* + * The file is dirty. It shoudl have been removed once at + * cleanup time but recovery can create them again. Remove + * them. + */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + else + elog(DEBUG2, "unlinked file \"%s\"", rm_path); + + continue; + } + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -351,6 +381,8 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ fsync_fname(dbspacedirname, true); } + + hash_destroy(hash); } /* diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..416fd859e6 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing inittmp fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != INITTMP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..2954cd9c24 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,8 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "itmp" /* INITTMP_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..4305bdbe96 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,8 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + INITTMP_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +53,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM INITTMP_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Fri_Jan__8_14_47_05_2021_579)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v3-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v8 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 52 +++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 539 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 245 +++++++++-- src/backend/replication/basebackup.c | 3 +- src/backend/storage/buffer/bufmgr.c | 88 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 345 +++++++++++----- src/backend/storage/smgr/md.c | 93 ++++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 20 +- src/bin/pg_rewind/parsexlog.c | 24 ++ src/common/relpath.c | 47 ++- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 42 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + 22 files changed, 1401 insertions(+), 207 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..d251f22207 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,49 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rnode.dbNode, + xlrec->rnode.spcNode, + xlrec->rnode.relNode, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + default: + action = "<unknown action>"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +98,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..b344bbe511 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The Smgr MARK files +-------------------------------- + +An smgr mark file is created when a new relation file is created to +mark the relfilenode needs to be cleaned up at recovery time. In +contrast to the four actions above, failure to remove smgr mark files +will lead to data loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 1e1fbe957f..59f4c2eacf 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4564,6 +4565,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7824,6 +7833,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index c5ad28d71f..f2bcc12958 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,18 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (1 << 0) +#define PDOP_UNLINK_FORK (1 << 1) +#define PDOP_UNLINK_MARK (1 << 2) +#define PDOP_SET_PERSISTENCE (1 << 3) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +86,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,22 +172,48 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The SMGR_MARK_UNCOMMITED mark file works as + * the signal of that situation. + */ srel = smgropen(rnode, backend); + log_smgrcreatemark(&rnode, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) log_smgrcreate(&srel->smgr_rnode.node, MAIN_FORKNUM); - /* Add the relation to the list of stuff to delete at abort */ + /* + * Add the relation to the list of stuff to delete at abort. We don't + * remove the mark file at commit. It needs to persists until the main fork + * file is actually deleted. See SyncPostCheckpoint. + */ pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_MARK; + pending->unlink_forknum = MAIN_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +223,226 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have entries for init-fork operations on this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init-fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + + /* + * We don't touch unrelated entries. Although init-fork related entries + * are not useful if the relation is created or dropped in this + * transaction, we don't bother to avoid registering entries for such + * relations here. + */ + if (!RelFileNodeEquals(rnode, pending->relnode) || + (pending->op & PDOP_DELETE) != 0 || + pending->unlink_forknum != INIT_FORKNUM) + { + prev = pending; + continue; + } + + /* make sure the entry is what we're expecting here */ + Assert(((pending->op & (PDOP_UNLINK_FORK|PDOP_UNLINK_MARK)) != 0 && + pending->unlink_forknum == INIT_FORKNUM) || + (pending->op & PDOP_SET_PERSISTENCE) != 0); + + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + + if (!create) + return; + + /* + * We are going to create an init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The mark file works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + log_smgrcreatemark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_UNLINK_MARK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop mark file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have entries for init-fork operations of this relation, that means + * that we have created the init fork in the current transaction. We + * remove the init fork and mark file immediately in that case. Otherwise + * just register pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + + /* + * We don't touch unrelated entries. Although init-fork related entries + * are not useful if the relation is created or dropped in this + * transaction, we don't bother to avoid registering entries for such + * relations here. + */ + if (!RelFileNodeEquals(rnode, pending->relnode) || + (pending->op & PDOP_DELETE) != 0 || + pending->unlink_forknum != INIT_FORKNUM)) + { + prev = pending; + continue; + } + + /* make sure the entry is what we're expecting here */ + Assert(((pending->op & (PDOP_UNLINK_FORK|PDOP_UNLINK_MARK)) != 0 && + pending->unlink_forknum == INIT_FORKNUM) || + (pending->op & PDOP_SET_PERSISTENCE) != 0); + + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT forks never be loaded to shared buffer so no point in dropping + * buffers for such files. + */ + log_smgrunlinkmark(&rnode, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +462,88 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +557,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -618,59 +976,104 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op & PDOP_DELETE) + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; + } + + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PDOP_UNLINK_MARK) + { + if (!InRecovery) + log_smgrunlinkmark(&pending->relnode, + pending->unlink_forknum, + pending->unlink_mark); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -840,7 +1243,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -853,7 +1257,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -933,6 +1338,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1021,6 +1435,65 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + bool created = false; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index bf42587e38..afc77f0d98 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -52,6 +52,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5329,6 +5330,166 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistnce change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother to allow in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5459,47 +5620,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index ec0485705d..45e1a5d817 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1070,6 +1070,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1120,7 +1121,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index b4532948d3..dab74bf99a 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlogutils.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3154,6 +3155,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 263057841d..8487ae1f02 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3759,7 +3757,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 0ae3fb6902..f8458a1e1e 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,228 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the relfilenode is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + StorageMarks mark; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +420,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +428,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +466,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +520,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +551,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index b4bca7eed6..580b74839f 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -139,7 +139,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -169,6 +170,81 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path =markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rnode.node.spcNode, + reln->smgr_rnode.node.dbNode, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not crete mark file \"%s\": %m", path))); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rnode, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path))); + pfree(path); + + if (fd < 0) + return false; + + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1025,6 +1101,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1378,12 +1463,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rnode, MAIN_FORKNUM); + p = GetRelationPath(ftag->rnode.dbNode, ftag->rnode.spcNode, + ftag->rnode.relNode, InvalidBackendId, MAIN_FORKNUM, + mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0fcef4994b..110e64b0b2 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -335,6 +341,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -662,6 +688,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index d4083e8a56..9563940d45 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -89,7 +89,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -222,7 +223,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -236,6 +238,20 @@ SyncPostCheckpoint(void) (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag( + &entry->tag, path, + SMGR_MARK_UNCOMMITTED) < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork files has been successfully removed. It's ok if the file + * does not exist. + */ + if (errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); + } /* Mark the list entry as canceled, just in case */ entry->canceled = true; diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 436df54120..dbc0da5da5 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -407,6 +407,30 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. When we compare the sizes later on, + * we'll notice that they differ, and copy the missing tail from + * source system. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..67f24890d6 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbNode, Oid spcNode) */ char * GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[10]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, 10, ".%c", mark); if (spcNode == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, Assert(dbNode == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNode, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNode, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNode); + path = psprintf("global/%u%s", relNode, markstr); } else if (spcNode == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbNode, relNode); + path = psprintf("base/%u/%u%s", + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbNode, backendId, relNode); + path = psprintf("base/%u/t%d_%u%s", + dbNode, backendId, relNode, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, relNode); + dbNode, relNode, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcNode, TABLESPACE_VERSION_DIRECTORY, dbNode, backendId, relNode, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcNode, TABLESPACE_VERSION_DIRECTORY, - dbNode, backendId, relNode); + dbNode, backendId, relNode, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..12346ed7f6 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilenode.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileNode rnode; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,12 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileNode *rnode, ForkNumber forkNum, + StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..106a5cf508 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbNode, Oid spcNode); extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, /* First argument is a RelFileNode */ #define relpathbackend(rnode, backend, forknum) \ GetRelationPath((rnode).dbNode, (rnode).spcNode, (rnode).relNode, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileNode */ #define relpathperm(rnode, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbNode, Oid spcNode, Oid relNode, #define relpath(rnode, forknum) \ relpathbackend((rnode).node, (rnode).backend, forknum) +/* First argument is a RelFileNodeBackend */ +#define markpath(rnode, forknum, mark) \ + GetRelationPath((rnode).node.dbNode, (rnode).node.spcNode, \ + (rnode).node.relNode, \ + (rnode).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index cfce23ecbc..f5a7df87a4 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -206,6 +206,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 34602ae006..2dc0357ad5 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern int durable_rename_excl(const char *oldfile, const char *newfile, int loglevel); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..99620816b5 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileNodeBackend rnode, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..e1f97e9b89 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..201ecace8a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilenode.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -85,7 +97,12 @@ extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln); extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Mon_Dec_20_15_28_23_2021_618)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v8-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v4 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 322 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1034 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 71b5852224..b730b4417c 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3032,6 +3033,93 @@ DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..b07709bc4f 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,232 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + RelFileNodeBackend *rnodes; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + rnodes = palloc(sizeof(RelFileNodeBackend) * nrels); + + for (i = 0 ; i < nrels ; i++) + rnodes[i] = srels[i]->smgr_rnode; + + DropRelFileNodesAllBuffers(rnodes, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 0f31ff3822..4102d3d59c 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -644,6 +644,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index ff6cd0fc54..d9752a8317 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(RelFileNodeBackend rnode, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(RelFileNodeBackend *rnodes, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index ebf4a199dc..8be17d9afc 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Tue_Jan_12_18_58_08_2021_481)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v4-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v5 1/2] In-place table persistence change @ 2020-11-11 12:51 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2020-11-11 12:51 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 23 ++ src/backend/access/transam/README | 8 + src/backend/access/transam/xlog.c | 17 + src/backend/catalog/storage.c | 436 +++++++++++++++++++++++-- src/backend/commands/tablecmds.c | 246 +++++++++++--- src/backend/storage/buffer/bufmgr.c | 88 +++++ src/backend/storage/file/reinit.c | 316 ++++++++++++------ src/backend/storage/smgr/md.c | 13 +- src/backend/storage/smgr/smgr.c | 6 + src/common/relpath.c | 4 +- src/include/catalog/storage.h | 2 + src/include/catalog/storage_xlog.h | 22 +- src/include/common/relpath.h | 6 +- src/include/storage/bufmgr.h | 2 + src/include/storage/md.h | 2 + src/include/storage/reinit.h | 3 +- src/include/storage/smgr.h | 1 + 17 files changed, 1028 insertions(+), 167 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index 7755553d57..2c109b8ca4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,23 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rnode, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rnode, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +72,12 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 1edc8180c1..547107a771 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,14 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +The CLEANUP fork file +-------------------------------- + +An CLEANUP fork is created when a new relation file is created to mark +the relfilenode needs to be cleaned up at recovery time. In contrast +to 4 above, failure to remove an CLEANUP fork file will lead to data +loss, in which case the server will shut down. + Skipping WAL for New RelFileNode -------------------------------- diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b18257c198..6dcbcbe387 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -40,6 +40,7 @@ #include "catalog/catversion.h" #include "catalog/pg_control.h" #include "catalog/pg_database.h" +#include "catalog/storage.h" #include "commands/progress.h" #include "commands/tablespace.h" #include "common/controldata_utils.h" @@ -4442,6 +4443,14 @@ ReadRecord(XLogReaderState *xlogreader, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; @@ -7455,6 +7464,14 @@ StartupXLOG(void) } } + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + /* Allow resource managers to do any required cleanup. */ for (rmid = 0; rmid <= RM_MAX_ID; rmid++) { diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index cba7a9ada0..c54d70747f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -27,6 +28,7 @@ #include "access/xlogutils.h" #include "catalog/storage.h" #include "catalog/storage_xlog.h" +#include "common/hashfn.h" #include "miscadmin.h" #include "storage/freespace.h" #include "storage/smgr.h" @@ -57,9 +59,16 @@ int wal_skip_threshold = 2048; /* in kilobytes */ * but I'm being paranoid. */ +#define PDOP_DELETE (0) +#define PDOP_UNLINK_FORK (1 << 0) +#define PDOP_SET_PERSISTENCE (1 << 1) + typedef struct PendingRelDelete { RelFileNode relnode; /* relation that may need to be deleted */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + int unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ int nestLevel; /* xact nesting level of request */ @@ -75,6 +84,24 @@ typedef struct PendingRelSync static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ HTAB *pendingSyncHash = NULL; +typedef struct SRelHashEntry +{ + SMgrRelation srel; + char status; /* for simplehash use */ +} SRelHashEntry; + +/* define hashtable for workarea for pending deletes */ +#define SH_PREFIX srelhash +#define SH_ELEMENT_TYPE SRelHashEntry +#define SH_KEY_TYPE SMgrRelation +#define SH_KEY srel +#define SH_HASH_KEY(tb, key) \ + hash_bytes((unsigned char *)&key, sizeof(SMgrRelation)) +#define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(SMgrRelation)) == 0) +#define SH_SCOPE static inline +#define SH_DEFINE +#define SH_DECLARE +#include "lib/simplehash.h" /* * AddPendingSync @@ -143,7 +170,17 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up but there's no + * clue to the orphan files. The cleanup fork works as the sentinel to + * identify that situation. + */ srel = smgropen(rnode, backend); + smgrcreate(srel, CLEANUP2_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP2_FORKNUM); + smgrimmedsync(srel, CLEANUP2_FORKNUM); + smgrcreate(srel, MAIN_FORKNUM, false); if (needs_wal) @@ -153,12 +190,25 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rnode; + pending->op = PDOP_DELETE; pending->backend = backend; pending->atCommit = false; /* delete if abort */ pending->nestLevel = GetCurrentTransactionNestLevel(); pending->next = pendingDeletes; pendingDeletes = pending; + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP2_FORKNUM; + pending->backend = backend; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) { Assert(backend == InvalidBackendId); @@ -168,6 +218,218 @@ RelationCreateStorage(RelFileNode rnode, char relpersistence) return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + SMgrRelation srel; + PendingRelDelete *prev; + PendingRelDelete *next; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, false, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have already registered pending delete entries to drop + * preexisting init fork since before the current transaction started. This + * function reverts that change just by removing the entries. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + create = false; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (!create) + return; + + /* + * We are going to create the init fork. If server crashes before the + * current transaction ends the init fork left alone corrupts data while + * recovery. The cleanup fork works as the sentinel to identify that + * situation. + */ + srel = smgropen(rnode, InvalidBackendId); + smgrcreate(srel, CLEANUP_FORKNUM, false); + log_smgrcreate(&rnode, CLEANUP_FORKNUM); + smgrimmedsync(srel, CLEANUP_FORKNUM); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * index-init fork needs further initialization. ambuildempty shoud do + * WAL-log and file sync by itself but otherwise we do that by myself. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rnode, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop this init fork file at abort and revert persistence */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK | PDOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* drop cleanup fork at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = CLEANUP_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion doesn't happen. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileNode rnode = rel->rd_node; + PendingRelDelete *pending; + PendingRelDelete *prev; + PendingRelDelete *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(rel->rd_smgr, true, false); + + /* + * If we have entries for init-fork operation of this relation, that means + * that we have created the init fork in the current transaction. We + * immediately remove the init and cleanup forks immediately in that case. + * Otherwise just reister pending-delete for the existing init fork. + */ + prev = NULL; + for (pending = pendingDeletes; pending != NULL; pending = next) + { + next = pending->next; + if (RelFileNodeEquals(rnode, pending->relnode) && + pending->op != PDOP_DELETE && + ((pending->op & PDOP_UNLINK_FORK) != 0 && + pending->unlink_forknum == CLEANUP_FORKNUM)) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingDeletes = next; + pfree(pending); + + inxact_created = true; + } + else + { + /* unrelated entry, don't touch it */ + prev = pending; + } + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rnode, InvalidBackendId); + + /* + * INIT/CLEANUP forks never be loaded to shared buffer so no point in + * dropping buffers for these files. + */ + log_smgrunlink(&rnode, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + log_smgrunlink(&rnode, CLEANUP_FORKNUM); + smgrunlink(srel, CLEANUP_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -187,6 +449,44 @@ log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rnode = *rnode; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileNode *rnode, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rnode = *rnode; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -200,6 +500,7 @@ RelationDropStorage(Relation rel) pending = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); pending->relnode = rel->rd_node; + pending->op = PDOP_DELETE; pending->backend = rel->rd_backend; pending->atCommit = true; /* delete if commit */ pending->nestLevel = GetCurrentTransactionNestLevel(); @@ -602,59 +903,97 @@ smgrDoPendingDeletes(bool isCommit) int nrels = 0, maxrels = 0; SMgrRelation *srels = NULL; + srelhash_hash *close_srels = NULL; + bool found; prev = NULL; for (pending = pendingDeletes; pending != NULL; pending = next) { + SMgrRelation srel; + next = pending->next; if (pending->nestLevel < nestLevel) { /* outer-level entries should not be processed yet */ prev = pending; + continue; } + + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; else + pendingDeletes = next; + + if (pending->atCommit != isCommit) { - /* unlink list entry first, so we don't retry on failure */ - if (prev) - prev->next = next; - else - pendingDeletes = next; - /* do deletion if called for */ - if (pending->atCommit == isCommit) - { - SMgrRelation srel; - - srel = smgropen(pending->relnode, pending->backend); - - /* allocate the initial array, or extend it, if needed */ - if (maxrels == 0) - { - maxrels = 8; - srels = palloc(sizeof(SMgrRelation) * maxrels); - } - else if (maxrels <= nrels) - { - maxrels *= 2; - srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); - } - - srels[nrels++] = srel; - } /* must explicitly free the list entry */ pfree(pending); /* prev does not change */ + continue; + } + + if (close_srels == NULL) + close_srels = srelhash_create(CurrentMemoryContext, 32, NULL); + + srel = smgropen(pending->relnode, pending->backend); + + /* Uniquify the smgr relations */ + srelhash_insert(close_srels, srel, &found); + + if (pending->op != PDOP_DELETE) + { + if (pending->op & PDOP_UNLINK_FORK) + { + /* other forks needs to drop buffers */ + Assert(pending->unlink_forknum == INIT_FORKNUM || + pending->unlink_forknum == CLEANUP_FORKNUM || + pending->unlink_forknum == CLEANUP2_FORKNUM); + + log_smgrunlink(&pending->relnode, pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + + } + + if (pending->op & PDOP_SET_PERSISTENCE) + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + else + { + /* allocate the initial array, or extend it, if needed */ + if (maxrels == 0) + { + maxrels = 8; + srels = palloc(sizeof(SMgrRelation) * maxrels); + } + else if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + srels[nrels++] = srel; } } if (nrels > 0) { smgrdounlinkall(srels, nrels, false); - - for (int i = 0; i < nrels; i++) - smgrclose(srels[i]); - pfree(srels); } + + if (close_srels) + { + srelhash_iterator i; + SRelHashEntry *ent; + + /* close smgr relatoins */ + srelhash_start_iterate(close_srels, &i); + while ((ent = srelhash_iterate(close_srels, &i)) != NULL) + smgrclose(ent->srel); + srelhash_destroy(close_srels); + } } /* @@ -824,7 +1163,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId + && pending->op == PDOP_DELETE) nrels++; } if (nrels == 0) @@ -837,7 +1177,8 @@ smgrGetPendingDeletes(bool forCommit, RelFileNode **ptr) for (pending = pendingDeletes; pending != NULL; pending = pending->next) { if (pending->nestLevel >= nestLevel && pending->atCommit == forCommit - && pending->backend == InvalidBackendId) + && pending->backend == InvalidBackendId && + pending->op == PDOP_DELETE) { *rptr = pending->relnode; rptr++; @@ -917,6 +1258,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rnode, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1005,6 +1355,28 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingRelDelete *pending; + + reln = smgropen(xlrec->rnode, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* revert buffer-persistence changes at abort */ + pending = (PendingRelDelete *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); + pending->relnode = xlrec->rnode; + pending->op = PDOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingDeletes; + pendingDeletes = pending; + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 993da56d43..37a15d31ee 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -50,6 +50,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -4917,6 +4918,170 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform im-place persistnce change"); + + RelationOpenSmgr(rel); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + RelationOpenSmgr(toastrel); + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach (lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * Some access methods do not accept in-place persistence change. For + * example, GiST uses page LSNs to figure out whether a block has + * changed, where UNLOGGED GiST indexes use fake LSNs that are + * incompatible with real LSNs used for LOGGED ones. + * + * XXXX: We don't bother allowing in-place persistence change for index + * methods other than btree for now. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + r->rd_rel->relam != BTREE_AM_OID) + { + int reindex_flags; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, 0); + + continue; + } + + RelationOpenSmgr(r); + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0 ; i < INIT_FORKNUM ; i++) + { + if (smgrexists(r->rd_smgr, i)) + smgrimmedsync(r->rd_smgr, i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + + for (fork = 0; fork < INIT_FORKNUM ; fork++) + { + if (smgrexists(r->rd_smgr, fork)) + log_newpage_range(r, fork, + 0, smgrnblocks(r->rd_smgr, fork), false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5037,45 +5202,52 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenode will - * have the right persistence set, and at the same time ensure - * that the original filenode's buffers will get read in with the - * correct setting (i.e. the original one). Otherwise a rollback - * after the rewrite would possibly result with buffers for the - * original filenode having the wrong persistence setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, - lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new relfilenode + * will have the right persistence set, and at the same time + * ensure that the original filenode's buffers will get read in + * with the correct setting (i.e. the original one). Otherwise + * a rollback after the rewrite would possibly result with + * buffers for the original filenode having the wrong + * persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that can't + * be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, persistence, + lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting this + * code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + } } else { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 561c212092..eacbdc6447 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -37,6 +37,7 @@ #include "access/xlog.h" #include "catalog/catalog.h" #include "catalog/storage.h" +#include "catalog/storage_xlog.h" #include "executor/instrument.h" #include "lib/binaryheap.h" #include "miscadmin.h" @@ -3094,6 +3095,93 @@ DropRelFileNodeBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelFileNodeBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileNodeBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileNodeBackend rnode = srel->smgr_rnode; + + Assert (!RelFileNodeBackendIsTemp(rnode)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rnode.node, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileNodeEquals(bufHdr->tag.rnode, rnode.node)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* init fork is always BM_PERMANENT. See BufferAlloc */ + if (bufHdr->tag.forkNum != INIT_FORKNUM) + buf_state &= ~BM_PERMANENT; + + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelFileNodesAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 40c758d789..0eac1956cc 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,50 @@ #include <unistd.h> +#include "access/xlog.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. * + * If CLEANUP_FORKNUM (clup) is present, we remove the init fork of the same + * relation along with the clup fork. + * + * If CLEANUP2_FORKNUM (cln2) is present we remove the whole relation along + * with the cln2 fork. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. + * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. */ @@ -68,7 +89,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -77,13 +98,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + Assert(tspid != 0); + + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -99,7 +126,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -126,6 +154,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -136,7 +166,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) snprintf(dbspace_path, sizeof(dbspace_path), "%s/%s", tsdirname, de->d_name); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -146,125 +179,226 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create a ton of unlogged relations + * in the same database & tablespace, so we'd better use a hash table + * rather than an array or linked list to keep track of which files + * need to be reset. Otherwise, this cleanup operation would be + * O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relfilenode cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT and CLEANUP forks in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum)) + continue; + + if (forkNum == INIT_FORKNUM || + forkNum == CLEANUP_FORKNUM || forkNum == CLEANUP2_FORKNUM) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relfilenode information. If it has the CLEANUP fork, + * the relfilenode is in dirty state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == CLEANUP_FORKNUM) + ent->dirty_init = true; + else if (forkNum == CLEANUP2_FORKNUM) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then the + * smgr object before initializing the unlogged relation. This is safe + * as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileNodeBackend rel; + + /* + * The relation is persistent and stays remain persistent. Don't + * drop the buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = ent->reloid; + + srels[nrels++] = smgropen(rel.node, InvalidBackendId); + } + + DropRelFileNodesAllBuffers(srels, nrels); + + for (i = 0 ; i < nrels ; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; + ForkNumber forkNum; + int oidchars; + Oid key; + relfile_entry *ent; + RelFileNodeBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, &forkNum)) continue; - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) - continue; - /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed trasaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM && forkNum != CLEANUP_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relfilenode + */ + if (forkNum == INIT_FORKNUM) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path))); + + rel.backend = InvalidBackendId; + rel.node.spcNode = tspid; + rel.node.dbNode = dbid; + rel.node.relNode = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 0643d714fb..6b37195c52 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -338,8 +338,10 @@ mdunlinkfork(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (ret == 0 || errno != ENOENT) { ret = unlink(path); + + /* failure of removing cleanup fork leads to a data loss. */ if (ret < 0 && errno != ENOENT) - ereport(WARNING, + ereport((forkNum != CLEANUP_FORKNUM ? WARNING : ERROR), (errcode_for_file_access(), errmsg("could not remove file \"%s\": %m", path))); } @@ -1024,6 +1026,15 @@ register_forget_request(RelFileNodeBackend rnode, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, ForkNumber forknum) +{ + register_forget_request(rnode, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index 4dc24649df..96480e321d 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -662,6 +662,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rnode, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/common/relpath.c b/src/common/relpath.c index 1f5c426ec0..479dcc248e 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -34,7 +34,9 @@ const char *const forkNames[] = { "main", /* MAIN_FORKNUM */ "fsm", /* FSM_FORKNUM */ "vm", /* VISIBILITYMAP_FORKNUM */ - "init" /* INIT_FORKNUM */ + "init", /* INIT_FORKNUM */ + "clup", /* CLEANUP_FORKNUM */ + "cln2" /* CLEANUP2_FORKNUM */ }; StaticAssertDecl(lengthof(forkNames) == (MAX_FORKNUM + 1), diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 0ab32b44e9..382623159c 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -23,6 +23,8 @@ extern int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileNode rnode, char relpersistence); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileNode rnode, bool atCommit); extern void RelationPreTruncate(Relation rel); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index f0814f1458..0fd0832a8b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -22,13 +22,17 @@ /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation, deletion and persistence change + * here. logging of deletion actions is mainly handled by xact.c, because it is + * part of transaction commit, but we log deletions happens outside of a + * transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -36,6 +40,18 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileNode rnode; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef struct xl_smgr_bufpersistence +{ + RelFileNode rnode; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +67,8 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileNode *rnode, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileNode *rnode, bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index a44be11ca0..040070aa2b 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -43,7 +43,9 @@ typedef enum ForkNumber MAIN_FORKNUM = 0, FSM_FORKNUM, VISIBILITYMAP_FORKNUM, - INIT_FORKNUM + INIT_FORKNUM, + CLEANUP_FORKNUM, + CLEANUP2_FORKNUM /* * NOTE: if you add a new fork, change MAX_FORKNUM and possibly @@ -52,7 +54,7 @@ typedef enum ForkNumber */ } ForkNumber; -#define MAX_FORKNUM INIT_FORKNUM +#define MAX_FORKNUM CLEANUP2_FORKNUM #define FORKNAMECHARS 4 /* max chars for a fork name */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index fb00fda6a7..ccb0a388f6 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -205,6 +205,8 @@ extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels) extern void FlushDatabaseBuffers(Oid dbid); extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum, int nforks, BlockNumber *firstDelBlock); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes); extern void DropDatabaseBuffers(Oid dbid); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 752b440864..3cbbbf2edd 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -41,6 +41,8 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileNodeBackend rnode, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileNode *delrels, int ndelrels, bool isRedo); diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index fad1e5c473..b969ba8e86 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -23,6 +23,7 @@ extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, ForkNumber *fork); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a6fbf7b6a6..1ac3e4a74a 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -86,6 +86,7 @@ extern void smgrclose(SMgrRelation reln); extern void smgrcloseall(void); extern void smgrclosenode(RelFileNodeBackend rnode); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, -- 2.27.0 ----Next_Part(Thu_Jan_14_17_32_17_2021_289)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="v5-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LOG.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v24 1/2] In-place table persistence change @ 2022-07-19 04:23 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2022-07-19 04:23 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 49 ++ src/backend/access/transam/README | 10 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlogrecovery.c | 18 + src/backend/backup/basebackup.c | 9 +- src/backend/catalog/storage.c | 561 +++++++++++++++++++++- src/backend/commands/tablecmds.c | 267 ++++++++-- src/backend/storage/buffer/bufmgr.c | 85 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 313 ++++++++---- src/backend/storage/smgr/md.c | 95 +++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 21 +- src/bin/pg_rewind/parsexlog.c | 22 + src/bin/pg_rewind/pg_rewind.c | 1 - src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 43 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 8 +- src/include/storage/smgr.h | 17 + src/tools/pgindent/typedefs.list | 7 + 25 files changed, 1471 insertions(+), 168 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index e0ee8a078a..2f92c06f70 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,46 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rlocator, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rlocator.dbOid, + xlrec->rlocator.spcOid, + xlrec->rlocator.relNumber, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action = "<none>"; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +95,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 91c2578f7a..37527e16ca 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -725,6 +725,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +================================ +The Smgr MARK files +-------------------------------- + +An smgr mark file is an empty file that is created when a new relation +storage file is created to signal that the storage file needs to be +cleaned up at recovery time. In contrast to the four actions above, +failure to remove smgr mark files will lead to data loss, in which +case the server will shut down. + Skipping WAL for New RelFileLocator -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index 2bb975943c..03e4bcec34 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2213,6 +2213,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2463,6 +2466,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2788,6 +2794,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index 1026ce5dcf..aac303934c 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -42,6 +42,7 @@ #include "access/xlogutils.h" #include "backup/basebackup.h" #include "catalog/pg_control.h" +#include "catalog/storage.h" #include "commands/tablespace.h" #include "common/file_utils.h" #include "miscadmin.h" @@ -56,6 +57,7 @@ #include "storage/pmsignal.h" #include "storage/proc.h" #include "storage/procarray.h" +#include "storage/reinit.h" #include "storage/spin.h" #include "utils/builtins.h" #include "utils/datetime.h" @@ -1777,6 +1779,14 @@ PerformWalRecovery(void) RmgrCleanup(); + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + ereport(LOG, (errmsg("redo done at %X/%X system usage: %s", LSN_FORMAT_ARGS(xlogreader->ReadRecPtr), @@ -3113,6 +3123,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; diff --git a/src/backend/backup/basebackup.c b/src/backend/backup/basebackup.c index 1434bcdd85..4e0c6c4e98 100644 --- a/src/backend/backup/basebackup.c +++ b/src/backend/backup/basebackup.c @@ -1190,6 +1190,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, ForkNumber relForkNum; /* Type of fork if file is a relation */ int relnumchars; /* Chars in filename that are the * relnumber */ + StorageMarks mark; /* marker file sign */ /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1240,7 +1241,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relnumchars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) @@ -1447,6 +1448,7 @@ is_checksummed_file(const char *fullpath, const char *filename) strncmp(fullpath, "/", 1) == 0) { int excludeIdx; + char *p; /* Compare file against noChecksumFiles skip list */ for (excludeIdx = 0; noChecksumFiles[excludeIdx].name != NULL; excludeIdx++) @@ -1460,6 +1462,11 @@ is_checksummed_file(const char *fullpath, const char *filename) return false; } + /* exclude mark files */ + p = strchr(filename, '.'); + if (p && isalpha(p[1]) && p[2] == 0) + return false; + return true; } else diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 38bbe32550..2c6472cfd5 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileLocator rlocator; /* relation that need a cleanup */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + ForkNumber unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileLocator rlocator; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup * pendingCleanups = NULL; /* head of linked list */ static HTAB *pendingSyncHash = NULL; @@ -123,6 +142,7 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, SMgrRelation srel; BackendId backend; bool needs_wal; + PendingCleanup *pendingclean; Assert(!IsInParallelMode()); /* couldn't update pendingSyncHash */ @@ -145,9 +165,23 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + * We don't need this during WAL-loggged CREATE DATABASE. See + * CreateAndCopyRelationData for detail. + */ srel = smgropen(rlocator, backend); + + if (register_delete) + { + log_smgrcreatemark(&rlocator, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + } + smgrcreate(srel, MAIN_FORKNUM, false); - + if (needs_wal) log_smgrcreate(&srel->smgr_rlocator.locator, MAIN_FORKNUM); @@ -157,16 +191,29 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, */ if (register_delete) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->rlocator = rlocator; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->rlocator = rlocator; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->rlocator = rlocator; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; } if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) @@ -178,6 +225,204 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init-fork exists since before the current transaction + * started. This function reverts that change just by removing the entry. + * See RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the commit-sentinel file */ + srel = smgropen(rlocator, InvalidBackendId); + log_smgrcreatemark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * init fork for indexes needs further initialization. ambuildempty should + * do WAL-log and file sync by itself but otherwise we do that by + * ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion is canceled. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init fork is created in the current transaction. We remove + * both the init fork and mark file immediately in that case. Otherwise + * register an at-commit pending-unlink for the existing init fork. See + * RelationCreateInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + + /* + * Some AMs initializes INIT fork via buffer manager. Drop all buffers + * for the INIT fork then unlink the INIT fork along with the mark + * file. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + log_smgrunlinkmark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -197,6 +442,88 @@ log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileLocator *rlocator, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator *rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = *rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -711,6 +1038,95 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * part of the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->rlocator, pending->backend); + + Assert((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + + if (pending->op & PCOP_UNLINK_FORK) + { + /* + * Unlink the fork file. Currently we use this only for + * init forks and we're sure that the init fork is not + * loaded on shared buffers. For RelationDropInitFork + * case, the function dropped that buffers. For + * RelationCreateInitFork case, PCOP_SET_PERSISTENCE(true) + * is set and the buffers have been dropped just before. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->rlocator, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->rlocator, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->rlocator, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -975,6 +1391,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rlocator, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1067,6 +1492,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 1b8e6d5729..1779959410 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5407,6 +5408,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take the hard way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. We don't + * emit this if wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5537,48 +5719,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 5b0e531f97..796cb139c3 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3172,6 +3172,91 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileLocatorBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(bufHdr->tag.forkNum != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 073dab2be5..c5f577a694 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -347,8 +347,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3670,7 +3668,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index c3faa68126..ea5d6bbba1 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlogrecovery.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { RelFileNumber relnumber; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,228 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ - if ((op & UNLOGGED_RELATION_CLEANUP) != 0) + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(RelFileNumber); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("unlogged relation RelFileNumbers", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { - HTAB *hash; - HASHCTL ctl; + ForkNumber forkNum; + int relnumchars; + StorageMarks mark; - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(RelFileNumber); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation RelFileNumbers", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &relnumchars, + &forkNum, &mark)) + continue; - /* Scan the directory. */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) { - ForkNumber forkNum; - int relnumchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &relnumchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; + RelFileNumbers key; + relfile_entry *ent; + bool found; /* * Put the RELFILENUMBER portion of the name into the hash table, - * if it isn't already. + * if it isn't already. If it has SMGR_MARK_UNCOMMITTED mark + * files, the storage file is in dirty state, where clean up is + * needed. */ - ent.relnumber = atorelnumber(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); + key = atorelnumber(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } } + } - /* Done with the first pass. */ - FreeDir(dbspace_dir); + /* Done with the first pass. */ + FreeDir(dbspace_dir); + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then + * the smgr object before initializing the unlogged relation. This is + * safe as far as no other backends have accessed the relation before + * starting archive recovery. */ - if (hash_get_num_entries(hash) == 0) + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while ((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) { - hash_destroy(hash); - return; + RelFileLocatorBackend rel; + + /* + * The relation is persistent and stays persistent. Don't drop the + * buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.locator.spcOid = tspid; + rel.locator.dbOid = dbid; + rel.locator.relNumber = ent->reloid; + + srels[nrels++] = smgropen(rel.locator, InvalidBackendId); } - /* - * Now, make a second pass and remove anything that matches. - */ + DropRelationsAllBuffers(srels, nrels); + + for (i = 0; i < nrels; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ + if ((op & UNLOGGED_RELATION_CLEANUP) != 0) + { dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int relnumchars; - unlogged_relation_entry ent; + RelFileNumber key; + relfile_entry *ent; + RelFileLocatorBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &relnumchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the RELFILENUMBER portion of the name shows up in * the hash table. If so, nuke it! */ - ent.relnumber = atorelnumber(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atorelnumber(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed transaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relation files. + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path)); + + rel.backend = InvalidBackendId; + rel.locator.spcOid = tspid; + rel.locator.dbOid = dbid; + rel.locator.relNumber = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +420,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int relnumchars; char relnumbuf[RELNUMBERCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +428,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &relnumchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +466,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int relnumchars; char relnumbuf[RELNUMBERCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &relnumchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +520,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *relnumchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +551,19 @@ parse_filename_for_nontemp_relation(const char *name, int *relnumchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index bed47f07d7..d628461970 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -141,7 +141,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -173,6 +174,82 @@ mdexists(SMgrRelation reln, ForkNumber forknum) return (mdopenfork(reln, forknum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rlocator.locator.spcOid, + reln->smgr_rlocator.locator.dbOid, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not create mark file \"%s\": %m", path)); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path)); + pfree(path); + + if (fd < 0) + return false; + + close(fd); + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1056,6 +1133,16 @@ register_forget_request(RelFileLocatorBackend rlocator, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileLocatorBackend rlocator, + ForkNumber forknum) +{ + register_forget_request(rlocator, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1414,12 +1501,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rlocator, MAIN_FORKNUM); + p = GetRelationPath(ftag->rlocator.dbOid, ftag->rlocator.spcOid, + ftag->rlocator.relNumber,InvalidBackendId, + MAIN_FORKNUM, mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index ed46ac3f44..c7353d3dcf 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -371,6 +377,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -693,6 +719,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rlocator, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index 9d6a9e9109..74357efb1c 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -91,7 +91,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -235,7 +236,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -244,6 +246,21 @@ SyncPostCheckpoint(void) * here. rmtree() also has to ignore ENOENT errors, to deal with * the possibility that we delete the file first. */ + if (errno != ENOENT) + ereport(WARNING, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path)); + } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, + path, + SMGR_MARK_UNCOMMITTED) + < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork file has been successfully removed. It's ok if the file + * does not exist. + */ if (errno != ENOENT) ereport(WARNING, (errcode_for_file_access(), diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 53f011a2fe..4da62f71ce 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -418,6 +418,28 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore there. We'll see that the file don't exist in + * the target data dir, and copy them in from the source system. No + * need to do anything special here. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these, The file will be removed from the + * target, if it doesn't exist in the source system. The files are + * empty so we don't need to bother the content. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/bin/pg_rewind/pg_rewind.c b/src/bin/pg_rewind/pg_rewind.c index 3cd77c09b1..f87dfda8fe 100644 --- a/src/bin/pg_rewind/pg_rewind.c +++ b/src/bin/pg_rewind/pg_rewind.c @@ -419,7 +419,6 @@ main(int argc, char **argv) if (showprogress) pg_log_info("reading source file list"); source->traverse_files(source, &process_source_file); - if (showprogress) pg_log_info("reading target file list"); traverse_datadir(datadir_target, &process_target_file); diff --git a/src/common/relpath.c b/src/common/relpath.c index d0d83e593b..f4f7435cf9 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbOid, Oid spcOid) */ char * GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcOid == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, Assert(dbOid == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/" UINT64_FORMAT "_%s", - relNumber, forkNames[forkNumber]); + path = psprintf("global/" UINT64_FORMAT "_%s%s", + relNumber, forkNames[forkNumber], markstr); else - path = psprintf("global/" UINT64_FORMAT, relNumber); + path = psprintf("global/" UINT64_FORMAT "%s", relNumber, markstr); } else if (spcOid == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/" UINT64_FORMAT "_%s", + path = psprintf("base/%u/" UINT64_FORMAT "_%s%s", dbOid, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/" UINT64_FORMAT, - dbOid, relNumber); + path = psprintf("base/%u/" UINT64_FORMAT "%s", + dbOid, relNumber, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_" UINT64_FORMAT "_%s", + path = psprintf("base/%u/t%d_" UINT64_FORMAT "_%s%s", dbOid, backendId, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_" UINT64_FORMAT, - dbOid, backendId, relNumber); + path = psprintf("base/%u/t%d_" UINT64_FORMAT "%s", + dbOid, backendId, relNumber, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/" UINT64_FORMAT "_%s", + path = psprintf("pg_tblspc/%u/%s/%u/" UINT64_FORMAT "_%s%s", spcOid, TABLESPACE_VERSION_DIRECTORY, dbOid, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/" UINT64_FORMAT, + path = psprintf("pg_tblspc/%u/%s/%u/" UINT64_FORMAT "%s", spcOid, TABLESPACE_VERSION_DIRECTORY, - dbOid, relNumber); + dbOid, relNumber, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_" UINT64_FORMAT "_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_" UINT64_FORMAT "_%s%s", spcOid, TABLESPACE_VERSION_DIRECTORY, dbOid, backendId, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_" UINT64_FORMAT, + path = psprintf("pg_tblspc/%u/%s/%u/t%d_" UINT64_FORMAT "%s", spcOid, TABLESPACE_VERSION_DIRECTORY, - dbOid, backendId, relNumber); + dbOid, backendId, relNumber, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 9964c312aa..ee4179699a 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -25,6 +25,8 @@ extern PGDLLIMPORT int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileLocator rlocator, char relpersistence, bool register_delete); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileLocator rlocator, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -43,6 +45,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileLocator **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 44a5e2043b..9f48fb5e6f 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilelocator.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileLocator rlocator; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileLocator rlocator; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,13 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileLocator *rlocator, + ForkNumber forkNum, StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileLocator *rlocator, + ForkNumber forkNum, StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileLocator *rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index 2d3b52fe0b..bcbd66ead3 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -77,7 +77,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbOid, Oid spcOid); extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -87,7 +87,7 @@ extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, /* First argument is a RelFileLocator */ #define relpathbackend(rlocator, backend, forknum) \ GetRelationPath((rlocator).dbOid, (rlocator).spcOid, (rlocator).relNumber, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileLocator */ #define relpathperm(rlocator, forknum) \ @@ -97,4 +97,9 @@ extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, #define relpath(rlocator, forknum) \ relpathbackend((rlocator).locator, (rlocator).backend, forknum) +/* First argument is a RelFileLocatorBackend */ +#define markpath(rlocator, forknum, mark) \ + GetRelationPath((rlocator).locator.dbOid, (rlocator).locator.spcOid, \ + (rlocator).locator.relNumber, \ + (rlocator).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index 6f4dfa0960..5b753b768b 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -142,6 +142,8 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 5a48fccd9c..b2773ae8bb 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int elevel); extern int durable_unlink(const char *fname, int elevel); extern void SyncDataDirectory(void); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 10aa1b0109..294a09444c 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileLocatorBackend rlocator, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index b990d28d38..cacd5e7d2c 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,14 +16,16 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); extern bool parse_filename_for_nontemp_relation(const char *name, int *relnumchars, - ForkNumber *fork); + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a07715356b..75b9d41e4b 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilelocator.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -87,7 +99,12 @@ extern void smgrcloseall(void); extern void smgrcloserellocator(RelFileLocatorBackend rlocator); extern void smgrrelease(SMgrRelation reln); extern void smgrreleaseall(void); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 97c9bc1861..2fdd221408 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -1970,6 +1970,7 @@ PatternInfoArray Pattern_Prefix_Status Pattern_Type PendingFsyncEntry +PendingMarkCleanup PendingRelDelete PendingRelSync PendingUnlinkEntry @@ -2600,6 +2601,7 @@ StdRdOptIndexCleanup StdRdOptions Step StopList +StorageMarks StrategyNumber StreamCtl String @@ -3607,6 +3609,7 @@ registered_buffer regmatch_t regoff_t regproc +relfile_entry relopt_bool relopt_enum relopt_enum_elt_def @@ -3660,6 +3663,7 @@ slist_iter slist_mutable_iter slist_node slock_t +smgr_mark_action socket_set socklen_t spgBulkDeleteState @@ -3858,8 +3862,11 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create +xl_smgr_mark xl_smgr_truncate +xl_smgr_unlink xl_standby_lock xl_standby_locks xl_tblspc_create_rec -- 2.31.1 ----Next_Part(Wed_Sep_28_17_21_19_2022_527)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="v24-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v23 1/2] In-place table persistence change @ 2022-07-19 04:23 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2022-07-19 04:23 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 49 ++ src/backend/access/transam/README | 10 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlogrecovery.c | 18 + src/backend/catalog/storage.c | 561 +++++++++++++++++++++- src/backend/commands/tablecmds.c | 267 ++++++++-- src/backend/replication/basebackup.c | 9 +- src/backend/storage/buffer/bufmgr.c | 85 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 340 +++++++++---- src/backend/storage/smgr/md.c | 95 +++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 21 +- src/bin/pg_rewind/parsexlog.c | 22 + src/bin/pg_rewind/pg_rewind.c | 1 - src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 43 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 10 +- src/include/storage/smgr.h | 17 + src/tools/pgindent/typedefs.list | 7 + 25 files changed, 1485 insertions(+), 183 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index e0ee8a078a..2f92c06f70 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,46 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rlocator, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rlocator.dbOid, + xlrec->rlocator.spcOid, + xlrec->rlocator.relNumber, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action = "<none>"; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +95,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 734c39a4d0..f08bd7f42d 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -724,6 +724,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +================================ +The Smgr MARK files +-------------------------------- + +An smgr mark file is an empty file that is created when a new relation +storage file is created to signal that the storage file needs to be +cleaned up at recovery time. In contrast to the four actions above, +failure to remove smgr mark files will lead to data loss, in which +case the server will shut down. + Skipping WAL for New RelFileLocator -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index 116de1175b..a1d97150dd 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2213,6 +2213,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2463,6 +2466,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2788,6 +2794,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index 5d6f1b5e46..42b4f6b5c8 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -41,6 +41,7 @@ #include "access/xlogrecovery.h" #include "access/xlogutils.h" #include "catalog/pg_control.h" +#include "catalog/storage.h" #include "commands/tablespace.h" #include "miscadmin.h" #include "pgstat.h" @@ -54,6 +55,7 @@ #include "storage/pmsignal.h" #include "storage/proc.h" #include "storage/procarray.h" +#include "storage/reinit.h" #include "storage/spin.h" #include "utils/builtins.h" #include "utils/guc.h" @@ -1773,6 +1775,14 @@ PerformWalRecovery(void) RmgrCleanup(); + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + ereport(LOG, (errmsg("redo done at %X/%X system usage: %s", LSN_FORMAT_ARGS(xlogreader->ReadRecPtr), @@ -3052,6 +3062,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index d708af19ed..317e44acfd 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileLocator rlocator; /* relation that need a cleanup */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + ForkNumber unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileLocator rlocator; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup * pendingCleanups = NULL; /* head of linked list */ static HTAB *pendingSyncHash = NULL; @@ -123,6 +142,7 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, SMgrRelation srel; BackendId backend; bool needs_wal; + PendingCleanup *pendingclean; Assert(!IsInParallelMode()); /* couldn't update pendingSyncHash */ @@ -145,9 +165,23 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + * We don't need this during WAL-loggged CREATE DATABASE. See + * CreateAndCopyRelationData for detail. + */ srel = smgropen(rlocator, backend); + + if (register_delete) + { + log_smgrcreatemark(&rlocator, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + } + smgrcreate(srel, MAIN_FORKNUM, false); - + if (needs_wal) log_smgrcreate(&srel->smgr_rlocator.locator, MAIN_FORKNUM); @@ -157,16 +191,29 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, */ if (register_delete) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->rlocator = rlocator; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->rlocator = rlocator; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->rlocator = rlocator; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; } if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) @@ -178,6 +225,204 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init-fork exists since before the current transaction + * started. This function reverts that change just by removing the entry. + * See RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the commit-sentinel file */ + srel = smgropen(rlocator, InvalidBackendId); + log_smgrcreatemark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * init fork for indexes needs further initialization. ambuildempty should + * do WAL-log and file sync by itself but otherwise we do that by + * ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion is canceled. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init fork is created in the current transaction. We remove + * both the init fork and mark file immediately in that case. Otherwise + * register an at-commit pending-unlink for the existing init fork. See + * RelationCreateInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + + /* + * Some AMs initializes INIT fork via buffer manager. Drop all buffers + * for the INIT fork then unlink the INIT fork along with the mark + * file. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + log_smgrunlinkmark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -197,6 +442,88 @@ log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileLocator *rlocator, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator *rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = *rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -711,6 +1038,95 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * part of the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->rlocator, pending->backend); + + Assert((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + + if (pending->op & PCOP_UNLINK_FORK) + { + /* + * Unlink the fork file. Currently we use this only for + * init forks and we're sure that the init fork is not + * loaded on shared buffers. For RelationDropInitFork + * case, the function dropped that buffers. For + * RelationCreateInitFork case, PCOP_SET_PERSISTENCE(true) + * is set and the buffers have been dropped just before. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->rlocator, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + SMgrRelation srel; + + if (!InRecovery) + log_smgrunlinkmark(&pending->rlocator, + pending->unlink_forknum, + pending->unlink_mark); + srel = smgropen(pending->rlocator, pending->backend); + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -971,6 +1387,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rlocator, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1059,6 +1484,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index a2f577024a..d8de3f496b 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -54,6 +54,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5376,6 +5377,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take the hard way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. We don't + * emit this if wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5506,48 +5688,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c index 95440013c0..0ef4b2bf01 100644 --- a/src/backend/replication/basebackup.c +++ b/src/backend/replication/basebackup.c @@ -1167,6 +1167,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, bool excludeFound; ForkNumber relForkNum; /* Type of fork if file is a relation */ int relOidChars; /* Chars in filename that are the rel oid */ + StorageMarks mark; /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1217,7 +1218,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relOidChars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) @@ -1442,6 +1443,7 @@ is_checksummed_file(const char *fullpath, const char *filename) strncmp(fullpath, "/", 1) == 0) { int excludeIdx; + char *p; /* Compare file against noChecksumFiles skip list */ for (excludeIdx = 0; noChecksumFiles[excludeIdx].name != NULL; excludeIdx++) @@ -1455,6 +1457,11 @@ is_checksummed_file(const char *fullpath, const char *filename) return false; } + /* exclude mark files */ + p = strchr(filename, '.'); + if (p && isalpha(p[1]) && p[2] == 0) + return false; + return true; } else diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index c7d7abcd73..c5d96b2fb7 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3161,6 +3161,91 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileLocatorBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(&srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (bufHdr->tag.forkNum == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr, true); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(bufHdr->tag.forkNum != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index f904f60c08..bcdbbad0f1 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -349,8 +349,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3696,7 +3694,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index f053fe0495..684c08fc1e 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlogrecovery.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { Oid reloid; /* hash key */ -} unlogged_relation_entry; + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,227 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(Oid); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("relation files cleanup hash", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + int oidchars; + ForkNumber forkNum; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + Oid key; + relfile_entry *ent; + bool found; + + /* + * Record the relation file information. If it has + * SMGR_MARK_UNCOMMITTED mark files, the storage file is in dirty + * state, where clean up is needed. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then + * the smgr object before initializing the unlogged relation. This is + * safe as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while ((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileLocatorBackend rel; + + /* + * The relation is persistent and stays persistent. Don't drop the + * buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.locator.spcOid = tspid; + rel.locator.dbOid = dbid; + rel.locator.relNumber = ent->reloid; + + srels[nrels++] = smgropen(rel.locator, InvalidBackendId); + } + + DropRelationsAllBuffers(srels, nrels); + + for (i = 0; i < nrels; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; - unlogged_relation_entry ent; + Oid key; + relfile_entry *ent; + RelFileLocatorBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int oidchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed transaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relation files. + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path)); + + rel.backend = InvalidBackendId; + rel.locator.spcOid = tspid; + rel.locator.dbOid = dbid; + rel.locator.relNumber = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } + hash_destroy(hash); + hash = NULL; + /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +419,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +427,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +465,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int oidchars; char oidbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &oidchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +519,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +550,19 @@ parse_filename_for_nontemp_relation(const char *name, int *oidchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 3998296a62..4249df657a 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -141,7 +141,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -173,6 +174,82 @@ mdexists(SMgrRelation reln, ForkNumber forkNum) return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rlocator.locator.spcOid, + reln->smgr_rlocator.locator.dbOid, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not create mark file \"%s\": %m", path)); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path)); + pfree(path); + + if (fd < 0) + return false; + + close(fd); + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1032,6 +1109,16 @@ register_forget_request(RelFileLocatorBackend rlocator, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileLocatorBackend rlocator, + ForkNumber forknum) +{ + register_forget_request(rlocator, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1390,12 +1477,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rlocator, MAIN_FORKNUM); + p = GetRelationPath(ftag->rlocator.dbOid, ftag->rlocator.spcOid, + ftag->rlocator.relNumber,InvalidBackendId, + MAIN_FORKNUM, mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index c1a5febcbf..5044dc21e4 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -371,6 +377,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -693,6 +719,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rlocator, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index 9d6a9e9109..74357efb1c 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -91,7 +91,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -235,7 +236,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -244,6 +246,21 @@ SyncPostCheckpoint(void) * here. rmtree() also has to ignore ENOENT errors, to deal with * the possibility that we delete the file first. */ + if (errno != ENOENT) + ereport(WARNING, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path)); + } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, + path, + SMGR_MARK_UNCOMMITTED) + < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork file has been successfully removed. It's ok if the file + * does not exist. + */ if (errno != ENOENT) ereport(WARNING, (errcode_for_file_access(), diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 53f011a2fe..4da62f71ce 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -418,6 +418,28 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore there. We'll see that the file don't exist in + * the target data dir, and copy them in from the source system. No + * need to do anything special here. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these, The file will be removed from the + * target, if it doesn't exist in the source system. The files are + * empty so we don't need to bother the content. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/bin/pg_rewind/pg_rewind.c b/src/bin/pg_rewind/pg_rewind.c index 1ff8da1676..c48a60cfe1 100644 --- a/src/bin/pg_rewind/pg_rewind.c +++ b/src/bin/pg_rewind/pg_rewind.c @@ -419,7 +419,6 @@ main(int argc, char **argv) if (showprogress) pg_log_info("reading source file list"); source->traverse_files(source, &process_source_file); - if (showprogress) pg_log_info("reading target file list"); traverse_datadir(datadir_target, &process_target_file); diff --git a/src/common/relpath.c b/src/common/relpath.c index 1b6b620ce8..46cfe38fd5 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbOid, Oid spcOid) */ char * GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcOid == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, Assert(dbOid == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNumber, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNumber, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNumber); + path = psprintf("global/%u%s", relNumber, markstr); } else if (spcOid == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbOid, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbOid, relNumber); + path = psprintf("base/%u/%u%s", + dbOid, relNumber, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbOid, backendId, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbOid, backendId, relNumber); + path = psprintf("base/%u/t%d_%u%s", + dbOid, backendId, relNumber, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcOid, TABLESPACE_VERSION_DIRECTORY, dbOid, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcOid, TABLESPACE_VERSION_DIRECTORY, - dbOid, relNumber); + dbOid, relNumber, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcOid, TABLESPACE_VERSION_DIRECTORY, dbOid, backendId, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcOid, TABLESPACE_VERSION_DIRECTORY, - dbOid, backendId, relNumber); + dbOid, backendId, relNumber, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 9964c312aa..ee4179699a 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -25,6 +25,8 @@ extern PGDLLIMPORT int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileLocator rlocator, char relpersistence, bool register_delete); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileLocator rlocator, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -43,6 +45,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileLocator **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 44a5e2043b..9f48fb5e6f 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilelocator.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileLocator rlocator; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileLocator rlocator; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,13 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileLocator *rlocator, + ForkNumber forkNum, StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileLocator *rlocator, + ForkNumber forkNum, StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileLocator *rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index 3ab713247f..eba7a05f4e 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -67,7 +67,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbOid, Oid spcOid); extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -77,7 +77,7 @@ extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, /* First argument is a RelFileLocator */ #define relpathbackend(rlocator, backend, forknum) \ GetRelationPath((rlocator).dbOid, (rlocator).spcOid, (rlocator).relNumber, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileLocator */ #define relpathperm(rlocator, forknum) \ @@ -87,4 +87,9 @@ extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, #define relpath(rlocator, forknum) \ relpathbackend((rlocator).locator, (rlocator).backend, forknum) +/* First argument is a RelFileLocatorBackend */ +#define markpath(rlocator, forknum, mark) \ + GetRelationPath((rlocator).locator.dbOid, (rlocator).locator.spcOid, \ + (rlocator).locator.relNumber, \ + (rlocator).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index bf8cce7ccf..be5d63319e 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -143,6 +143,8 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 2b4a8e0ffe..eb33a9ba4c 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -185,6 +185,7 @@ extern ssize_t pg_pwritev_with_retry(int fd, extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int loglevel); extern int durable_unlink(const char *fname, int loglevel); extern void SyncDataDirectory(void); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 10aa1b0109..294a09444c 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileLocatorBackend rlocator, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index bf2c10d443..e399aec0c7 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,13 +16,15 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); -extern bool parse_filename_for_nontemp_relation(const char *name, - int *oidchars, ForkNumber *fork); +extern bool parse_filename_for_nontemp_relation(const char *name, int *oidchars, + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a07715356b..75b9d41e4b 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilelocator.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -87,7 +99,12 @@ extern void smgrcloseall(void); extern void smgrcloserellocator(RelFileLocatorBackend rlocator); extern void smgrrelease(SMgrRelation reln); extern void smgrreleaseall(void); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 34a76ceb60..5445826998 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -1981,6 +1981,7 @@ PatternInfoArray Pattern_Prefix_Status Pattern_Type PendingFsyncEntry +PendingMarkCleanup PendingRelDelete PendingRelSync PendingUnlinkEntry @@ -2611,6 +2612,7 @@ StdRdOptIndexCleanup StdRdOptions Step StopList +StorageMarks StrategyNumber StreamCtl String @@ -3612,6 +3614,7 @@ registered_buffer regmatch_t regoff_t regproc +relfile_entry relopt_bool relopt_enum relopt_enum_elt_def @@ -3665,6 +3668,7 @@ slist_iter slist_mutable_iter slist_node slock_t +smgr_mark_action socket_set socklen_t spgBulkDeleteState @@ -3863,8 +3867,11 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create +xl_smgr_mark xl_smgr_truncate +xl_smgr_unlink xl_standby_lock xl_standby_locks xl_tblspc_create_rec -- 2.31.1 ----Next_Part(Tue_Jul_19_13_33_33_2022_460)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="v23-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v25 1/2] In-place table persistence change @ 2022-07-19 04:23 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2022-07-19 04:23 UTC (permalink / raw) Even though ALTER TABLE SET LOGGED/UNLOGGED does not require data rewriting, currently it runs heap rewrite which causes large amount of file I/O. This patch makes the command run without heap rewrite. Addition to that, SET LOGGED while wal_level > minimal emits WAL using XLOG_FPI instead of massive number of HEAP_INSERT's, which should be smaller. Also this allows for the cleanup of files left behind in the crash of the transaction that created it. --- src/backend/access/rmgrdesc/smgrdesc.c | 49 ++ src/backend/access/transam/README | 10 + src/backend/access/transam/xact.c | 7 + src/backend/access/transam/xlogrecovery.c | 18 + src/backend/backup/basebackup.c | 9 +- src/backend/catalog/storage.c | 559 +++++++++++++++++++++- src/backend/commands/tablecmds.c | 267 +++++++++-- src/backend/storage/buffer/bufmgr.c | 87 ++++ src/backend/storage/file/fd.c | 4 +- src/backend/storage/file/reinit.c | 340 +++++++++---- src/backend/storage/smgr/md.c | 95 +++- src/backend/storage/smgr/smgr.c | 32 ++ src/backend/storage/sync/sync.c | 21 +- src/bin/pg_rewind/parsexlog.c | 22 + src/bin/pg_rewind/pg_rewind.c | 1 - src/common/relpath.c | 47 +- src/include/catalog/storage.h | 3 + src/include/catalog/storage_xlog.h | 43 +- src/include/common/relpath.h | 9 +- src/include/storage/bufmgr.h | 2 + src/include/storage/fd.h | 1 + src/include/storage/md.h | 8 +- src/include/storage/reinit.h | 8 +- src/include/storage/smgr.h | 17 + src/tools/pgindent/typedefs.list | 7 + 25 files changed, 1483 insertions(+), 183 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index e0ee8a078a..2f92c06f70 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,46 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) rec; + char *path = relpathperm(xlrec->rlocator, xlrec->forkNum); + + appendStringInfoString(buf, path); + pfree(path); + } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) rec; + char *path = GetRelationPath(xlrec->rlocator.dbOid, + xlrec->rlocator.spcOid, + xlrec->rlocator.relNumber, + InvalidBackendId, + xlrec->forkNum, xlrec->mark); + char *action = "<none>"; + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + action = "CREATE"; + break; + case XLOG_SMGR_MARK_UNLINK: + action = "DELETE"; + break; + } + + appendStringInfo(buf, "%s %s", action, path); + pfree(path); + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +95,15 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_UNLINK: + id = "UNLINK"; + break; + case XLOG_SMGR_MARK: + id = "MARK"; + break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/access/transam/README b/src/backend/access/transam/README index 22c8ae9755..617a63e2c5 100644 --- a/src/backend/access/transam/README +++ b/src/backend/access/transam/README @@ -741,6 +741,16 @@ we must panic and abort recovery. The DBA will have to manually clean up then restart recovery. This is part of the reason for not writing a WAL entry until we've successfully done the original action. +================================ +The Smgr MARK files +-------------------------------- + +An smgr mark file is an empty file that is created when a new relation +storage file is created to signal that the storage file needs to be +cleaned up at recovery time. In contrast to the four actions above, +failure to remove smgr mark files will lead to data loss, in which +case the server will shut down. + Skipping WAL for New RelFileLocator -------------------------------- diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index 8086b857b9..17d631a5db 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2213,6 +2213,9 @@ CommitTransaction(void) */ smgrDoPendingSyncs(true, is_parallel_worker); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2463,6 +2466,9 @@ PrepareTransaction(void) */ smgrDoPendingSyncs(true, false); + /* Likewise delete mark files for files created during this transaction. */ + smgrDoPendingCleanups(true); + /* close large objects before lower-level cleanup */ AtEOXact_LargeObject(true); @@ -2788,6 +2794,7 @@ AbortTransaction(void) AfterTriggerEndXact(false); /* 'false' means it's abort */ AtAbort_Portals(); smgrDoPendingSyncs(false, is_parallel_worker); + smgrDoPendingCleanups(false); AtEOXact_LargeObject(false); AtAbort_Notify(); AtEOXact_RelationMap(false, is_parallel_worker); diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index cb07694aea..af49fb5bf5 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -42,6 +42,7 @@ #include "access/xlogutils.h" #include "backup/basebackup.h" #include "catalog/pg_control.h" +#include "catalog/storage.h" #include "commands/tablespace.h" #include "common/file_utils.h" #include "miscadmin.h" @@ -56,6 +57,7 @@ #include "storage/pmsignal.h" #include "storage/proc.h" #include "storage/procarray.h" +#include "storage/reinit.h" #include "storage/spin.h" #include "utils/builtins.h" #include "utils/datetime.h" @@ -1777,6 +1779,14 @@ PerformWalRecovery(void) RmgrCleanup(); + /* cleanup garbage files left during crash recovery */ + if (!InArchiveRecovery) + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + ereport(LOG, (errmsg("redo done at %X/%X system usage: %s", LSN_FORMAT_ARGS(xlogreader->ReadRecPtr), @@ -3113,6 +3123,14 @@ ReadRecord(XLogPrefetcher *xlogprefetcher, int emode, { ereport(DEBUG1, (errmsg_internal("reached end of WAL in pg_wal, entering archive recovery"))); + + /* cleanup garbage files left during crash recovery */ + ResetUnloggedRelations(UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_CLEANUP); + + /* run rollback cleanup if any */ + smgrDoPendingDeletes(false); + InArchiveRecovery = true; if (StandbyModeRequested) StandbyMode = true; diff --git a/src/backend/backup/basebackup.c b/src/backend/backup/basebackup.c index 74fb529380..1736df3d24 100644 --- a/src/backend/backup/basebackup.c +++ b/src/backend/backup/basebackup.c @@ -1190,6 +1190,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, ForkNumber relForkNum; /* Type of fork if file is a relation */ int relnumchars; /* Chars in filename that are the * relnumber */ + StorageMarks mark; /* marker file sign */ /* Skip special stuff */ if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0) @@ -1240,7 +1241,7 @@ sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly, /* Exclude all forks for unlogged tables except the init fork */ if (isDbDir && parse_filename_for_nontemp_relation(de->d_name, &relnumchars, - &relForkNum)) + &relForkNum, &mark)) { /* Never exclude init forks */ if (relForkNum != INIT_FORKNUM) @@ -1447,6 +1448,7 @@ is_checksummed_file(const char *fullpath, const char *filename) strncmp(fullpath, "/", 1) == 0) { int excludeIdx; + char *p; /* Compare file against noChecksumFiles skip list */ for (excludeIdx = 0; noChecksumFiles[excludeIdx].name != NULL; excludeIdx++) @@ -1460,6 +1462,11 @@ is_checksummed_file(const char *fullpath, const char *filename) return false; } + /* exclude mark files */ + p = strchr(filename, '.'); + if (p && isalpha(p[1]) && p[2] == 0) + return false; + return true; } else diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index d708af19ed..595d6c2bb3 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -19,6 +19,7 @@ #include "postgres.h" +#include "access/amapi.h" #include "access/parallel.h" #include "access/visibilitymap.h" #include "access/xact.h" @@ -66,6 +67,23 @@ typedef struct PendingRelDelete struct PendingRelDelete *next; /* linked-list link */ } PendingRelDelete; +#define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) + +typedef struct PendingCleanup +{ + RelFileLocator rlocator; /* relation that need a cleanup */ + int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ + ForkNumber unlink_forknum; /* forknum to unlink */ + StorageMarks unlink_mark; /* mark to unlink */ + BackendId backend; /* InvalidBackendId if not a temp rel */ + bool atCommit; /* T=delete at commit; F=delete at abort */ + int nestLevel; /* xact nesting level of request */ + struct PendingCleanup *next; /* linked-list link */ +} PendingCleanup; + typedef struct PendingRelSync { RelFileLocator rlocator; @@ -73,6 +91,7 @@ typedef struct PendingRelSync } PendingRelSync; static PendingRelDelete *pendingDeletes = NULL; /* head of linked list */ +static PendingCleanup * pendingCleanups = NULL; /* head of linked list */ static HTAB *pendingSyncHash = NULL; @@ -123,6 +142,7 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, SMgrRelation srel; BackendId backend; bool needs_wal; + PendingCleanup *pendingclean; Assert(!IsInParallelMode()); /* couldn't update pendingSyncHash */ @@ -145,9 +165,23 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return NULL; /* placate compiler */ } + /* + * We are going to create a new storage file. If server crashes before the + * current transaction ends the file needs to be cleaned up. The + * SMGR_MARK_UNCOMMITED mark file prompts that work at the next startup. + * We don't need this during WAL-loggged CREATE DATABASE. See + * CreateAndCopyRelationData for detail. + */ srel = smgropen(rlocator, backend); + + if (register_delete) + { + log_smgrcreatemark(&rlocator, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, MAIN_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + } + smgrcreate(srel, MAIN_FORKNUM, false); - + if (needs_wal) log_smgrcreate(&srel->smgr_rlocator.locator, MAIN_FORKNUM); @@ -157,16 +191,29 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, */ if (register_delete) { - PendingRelDelete *pending; + PendingRelDelete *pendingdel; - pending = (PendingRelDelete *) + pendingdel = (PendingRelDelete *) MemoryContextAlloc(TopMemoryContext, sizeof(PendingRelDelete)); - pending->rlocator = rlocator; - pending->backend = backend; - pending->atCommit = false; /* delete if abort */ - pending->nestLevel = GetCurrentTransactionNestLevel(); - pending->next = pendingDeletes; - pendingDeletes = pending; + pendingdel->rlocator = rlocator; + pendingdel->backend = backend; + pendingdel->atCommit = false; /* delete if abort */ + pendingdel->nestLevel = GetCurrentTransactionNestLevel(); + pendingdel->next = pendingDeletes; + pendingDeletes = pendingdel; + + /* drop mark files at commit */ + pendingclean = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pendingclean->rlocator = rlocator; + pendingclean->op = PCOP_UNLINK_MARK; + pendingclean->unlink_forknum = MAIN_FORKNUM; + pendingclean->unlink_mark = SMGR_MARK_UNCOMMITTED; + pendingclean->backend = backend; + pendingclean->atCommit = true; + pendingclean->nestLevel = GetCurrentTransactionNestLevel(); + pendingclean->next = pendingCleanups; + pendingCleanups = pendingclean; } if (relpersistence == RELPERSISTENCE_PERMANENT && !XLogIsNeeded()) @@ -178,6 +225,204 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init-fork exists since before the current transaction + * started. This function reverts that change just by removing the entry. + * See RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the commit-sentinel file */ + srel = smgropen(rlocator, InvalidBackendId); + log_smgrcreatemark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * init fork for indexes needs further initialization. ambuildempty should + * do WAL-log and file sync by itself but otherwise we do that by + * ourselves. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file and revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + * + * Register pending-delete of the init fork. The real deletion is performed by + * smgrDoPendingDeletes at commit. + * + * This function is transactional. If the transaction aborts later on, the + * deletion is canceled. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init fork is created in the current transaction. We remove + * both the init fork and mark file immediately in that case. Otherwise + * register an at-commit pending-unlink for the existing init fork. See + * RelationCreateInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + + /* + * Some AMs initializes INIT fork via buffer manager. Drop all buffers + * for the INIT fork then unlink the INIT fork along with the mark + * file. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + log_smgrunlinkmark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -197,6 +442,88 @@ log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_CREATE | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINK record to WAL. + */ +void +log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum) +{ + xl_smgr_unlink xlrec; + + /* + * Make an XLOG entry reporting the file unlink. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_CREATEMARK record to WAL. + */ +void +log_smgrcreatemark(const RelFileLocator *rlocator, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_CREATE; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_UNLINKMARK record to WAL. + */ +void +log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, + StorageMarks mark) +{ + xl_smgr_mark xlrec; + + /* + * Make an XLOG entry reporting the file creation. + */ + xlrec.rlocator = *rlocator; + xlrec.forkNum = forkNum; + xlrec.mark = mark; + xlrec.action = XLOG_SMGR_MARK_UNLINK; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); +} + +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator *rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = *rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -711,6 +1038,93 @@ smgrDoPendingDeletes(bool isCommit) } } +/* + * smgrDoPendingUnmark() -- Clean up work that emits WAL records + * + * The operations handled in the function emits WAL records, which must be + * part of the current transaction. + */ +void +smgrDoPendingCleanups(bool isCommit) +{ + int nestLevel = GetCurrentTransactionNestLevel(); + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + if (pending->nestLevel < nestLevel) + { + /* outer-level entries should not be processed yet */ + prev = pending; + } + else + { + /* unlink list entry first, so we don't retry on failure */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + /* do cleanup if called for */ + if (pending->atCommit == isCommit) + { + SMgrRelation srel; + + srel = smgropen(pending->rlocator, pending->backend); + + Assert((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } + + if (pending->op & PCOP_UNLINK_FORK) + { + /* + * Unlink the fork file. Currently we use this only for + * init forks and we're sure that the init fork is not + * loaded on shared buffers. For RelationDropInitFork + * case, the function dropped that buffers. For + * RelationCreateInitFork case, PCOP_SET_PERSISTENCE(true) + * is set and the buffers have been dropped just before. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + + /* Don't emit wal while recovery. */ + if (!InRecovery) + log_smgrunlink(&pending->rlocator, + pending->unlink_forknum); + smgrunlink(srel, pending->unlink_forknum, false); + } + + if (pending->op & PCOP_UNLINK_MARK) + { + if (!InRecovery) + log_smgrunlinkmark(&pending->rlocator, + pending->unlink_forknum, + pending->unlink_mark); + + smgrunlinkmark(srel, pending->unlink_forknum, + pending->unlink_mark, InRecovery); + smgrclose(srel); + } + } + + /* must explicitly free the list entry */ + pfree(pending); + /* prev does not change */ + } + } +} + /* * smgrDoPendingSyncs() -- Take care of relation syncs at end of xact. */ @@ -971,6 +1385,15 @@ smgr_redo(XLogReaderState *record) reln = smgropen(xlrec->rlocator, InvalidBackendId); smgrcreate(reln, xlrec->forkNum, true); } + else if (info == XLOG_SMGR_UNLINK) + { + xl_smgr_unlink *xlrec = (xl_smgr_unlink *) XLogRecGetData(record); + SMgrRelation reln; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + smgrunlink(reln, xlrec->forkNum, true); + smgrclose(reln); + } else if (info == XLOG_SMGR_TRUNCATE) { xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record); @@ -1059,6 +1482,124 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_MARK) + { + xl_smgr_mark *xlrec = (xl_smgr_mark *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + bool created = false; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + + switch (xlrec->action) + { + case XLOG_SMGR_MARK_CREATE: + smgrcreatemark(reln, xlrec->forkNum, xlrec->mark, true); + created = true; + break; + case XLOG_SMGR_MARK_UNLINK: + smgrunlinkmark(reln, xlrec->forkNum, xlrec->mark, true); + break; + default: + elog(ERROR, "unknown smgr_mark action \"%c\"", xlrec->mark); + } + + if (created) + { + /* revert mark file operation at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = xlrec->forkNum; + pending->unlink_mark = xlrec->mark; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + else + { + /* + * Delete pending action for this mark file if any. We should have + * at most one entry for this action. + */ + PendingCleanup *prev = NULL; + + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + pending->unlink_forknum == xlrec->forkNum && + (pending->op & PCOP_UNLINK_MARK) != 0) + { + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * Revert buffer-persistence changes at abort if the relation is going + * to different persistence from before this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index f006807852..b01bdaa51f 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5407,6 +5408,187 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: do in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Under the following condition, we need to call ATRewriteTable, which + * cannot be false in the AT_REWRITE_ALTER_PERSISTENCE case. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * First we collect all relations that we need to change persistence. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not bear up an in-place persistence + * change. Specifically, GiST uses page LSNs to figure out whether a + * block has changed, where UNLOGGED GiST indexes use fake LSNs that + * are incompatible with real LSNs used for LOGGED ones. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could skip index rebuild in exchange of some + * extra WAL records emitted while it is unlogged. + * + * Check relam against a positive list so that we take the hard way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * When this relation gets WAL-logged, immediately sync all files but + * initfork to establish the initial state on storage. Buffers have + * already flushed out by RelationCreate(Drop)InitFork called just + * above. Initfork should have been synced as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * While wal_level >= replica, switching to LOGGED requires the + * relation content to be WAL-logged to recover the table. We don't + * emit this if wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5537,48 +5719,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 73d30bf619..6c6590005e 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3169,6 +3169,93 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more accurately, out to kernel disk buffers), + * ensuring that the kernel has an up-to-date view of the relation. + * + * Generally, the caller should be holding AccessExclusiveLock on the + * target relation to ensure that no other backend is busy dirtying + * more blocks of the relation; the effects can't be expected to last + * after the lock is released. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not + * used in any performance-critical code paths, so it's not worth + * adding additional overhead to normal paths to make it go faster; + * but see also DropRelFileLocatorBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* Init fork is being dropped, drop buffers for it. */ + if (BufTagGetForkNum(&bufHdr->tag) == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* we flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(BufTagGetForkNum(&bufHdr->tag) != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index 4151cafec5..f3e087a006 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -346,8 +346,6 @@ static void pre_sync_fname(const char *fname, bool isdir, int elevel); static void datadir_fsync_fname(const char *fname, bool isdir, int elevel); static void unlink_if_exists_fname(const char *fname, bool isdir, int elevel); -static int fsync_parent_path(const char *fname, int elevel); - /* * pg_fsync --- do fsync with or without writethrough @@ -3669,7 +3667,7 @@ fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel) * This is aimed at making file operations persistent on disk in case of * an OS crash or power failure. */ -static int +int fsync_parent_path(const char *fname, int elevel) { char parentpath[MAXPGPATH]; diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 647c458b52..4d51cdaf34 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -16,29 +16,49 @@ #include <unistd.h> +#include "access/xlogrecovery.h" +#include "catalog/pg_tablespace_d.h" #include "common/relpath.h" #include "postmaster/startup.h" +#include "storage/bufmgr.h" #include "storage/copydir.h" #include "storage/fd.h" +#include "storage/md.h" #include "storage/reinit.h" +#include "storage/smgr.h" #include "utils/hsearch.h" #include "utils/memutils.h" static void ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, - int op); + Oid tspid, int op); static void ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, - int op); + Oid tspid, Oid dbid, int op); typedef struct { - Oid reloid; /* hash key */ -} unlogged_relation_entry; + RelFileNumber relNumber; /* hash key */ + bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ + bool dirty_all; /* needs to remove all forks */ +} relfile_entry; /* - * Reset unlogged relations from before the last restart. + * Clean up and reset relation files from before the last restart. * - * If op includes UNLOGGED_RELATION_CLEANUP, we remove all forks of any - * relation with an "init" fork, except for the "init" fork itself. + * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. + * + * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the + * whole relation along with the mark file. + * + * Otherwise, if the "init" fork is found. we remove all forks of any relation + * with the "init" fork, except for the "init" fork itself. + * + * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -72,7 +92,7 @@ ResetUnloggedRelations(int op) /* * First process unlogged files in pg_default ($PGDATA/base) */ - ResetUnloggedRelationsInTablespaceDir("base", op); + ResetUnloggedRelationsInTablespaceDir("base", DEFAULTTABLESPACE_OID, op); /* * Cycle through directories for all non-default tablespaces. @@ -81,13 +101,19 @@ ResetUnloggedRelations(int op) while ((spc_de = ReadDir(spc_dir, "pg_tblspc")) != NULL) { + Oid tspid; + if (strcmp(spc_de->d_name, ".") == 0 || strcmp(spc_de->d_name, "..") == 0) continue; snprintf(temp_path, sizeof(temp_path), "pg_tblspc/%s/%s", spc_de->d_name, TABLESPACE_VERSION_DIRECTORY); - ResetUnloggedRelationsInTablespaceDir(temp_path, op); + + tspid = atooid(spc_de->d_name); + + Assert(tspid != 0); + ResetUnloggedRelationsInTablespaceDir(temp_path, tspid, op); } FreeDir(spc_dir); @@ -103,7 +129,8 @@ ResetUnloggedRelations(int op) * Process one tablespace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) +ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, + Oid tspid, int op) { DIR *ts_dir; struct dirent *de; @@ -130,6 +157,8 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) while ((de = ReadDir(ts_dir, tsdirname)) != NULL) { + Oid dbid; + /* * We're only interested in the per-database directories, which have * numeric names. Note that this code will also (properly) ignore "." @@ -148,7 +177,10 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) ereport_startup_progress("resetting unlogged relations (cleanup), elapsed time: %ld.%02d s, current path: %s", dbspace_path); - ResetUnloggedRelationsInDbspaceDir(dbspace_path, op); + dbid = atooid(de->d_name); + Assert(dbid != 0); + + ResetUnloggedRelationsInDbspaceDir(dbspace_path, tspid, dbid, op); } FreeDir(ts_dir); @@ -158,125 +190,225 @@ ResetUnloggedRelationsInTablespaceDir(const char *tsdirname, int op) * Process one per-dbspace directory for ResetUnloggedRelations */ static void -ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) +ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, + Oid tspid, Oid dbid, int op) { DIR *dbspace_dir; struct dirent *de; char rm_path[MAXPGPATH * 2]; + HTAB *hash; + HASHCTL ctl; /* Caller must specify at least one operation. */ - Assert((op & (UNLOGGED_RELATION_CLEANUP | UNLOGGED_RELATION_INIT)) != 0); + Assert((op & (UNLOGGED_RELATION_CLEANUP | + UNLOGGED_RELATION_DROP_BUFFER | + UNLOGGED_RELATION_INIT)) != 0); /* * Cleanup is a two-pass operation. First, we go through and identify all * the files with init forks. Then, we go through again and nuke * everything with the same OID except the init fork. */ + + /* + * It's possible that someone could create tons of unlogged relations in + * the same database & tablespace, so we'd better use a hash table rather + * than an array or linked list to keep track of which files need to be + * reset. Otherwise, this cleanup operation would be O(n^2). + */ + memset(&ctl, 0, sizeof(ctl)); + ctl.keysize = sizeof(RelFileNumber); + ctl.entrysize = sizeof(relfile_entry); + hash = hash_create("unlogged relation RelFileNumbers", + 32, &ctl, HASH_ELEM | HASH_BLOBS); + + /* Collect INIT fork and mark files in the directory. */ + dbspace_dir = AllocateDir(dbspacedirname); + while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) + { + ForkNumber forkNum; + int relnumchars; + StorageMarks mark; + + /* Skip anything that doesn't look like a relation data file. */ + if (!parse_filename_for_nontemp_relation(de->d_name, &relnumchars, + &forkNum, &mark)) + continue; + + if (forkNum == INIT_FORKNUM || mark == SMGR_MARK_UNCOMMITTED) + { + RelFileNumber key; + relfile_entry *ent; + bool found; + + /* + * Put the OID portion of the name into the hash table, + * if it isn't already. If it has SMGR_MARK_UNCOMMITTED mark + * files, the storage file is in dirty state, where clean up is + * needed. isn't already. + */ + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_ENTER, &found); + + if (!found) + { + ent->has_init = false; + ent->dirty_init = false; + ent->dirty_all = false; + } + + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_all = true; + else + { + Assert(forkNum == INIT_FORKNUM); + ent->has_init = true; + } + } + } + + /* Done with the first pass. */ + FreeDir(dbspace_dir); + + /* nothing to do if we don't have init nor cleanup forks */ + if (hash_get_num_entries(hash) < 1) + { + hash_destroy(hash); + return; + } + + if ((op & UNLOGGED_RELATION_DROP_BUFFER) != 0) + { + /* + * When we come here after recovery, smgr object for this file might + * have been created. In that case we need to drop all buffers then + * the smgr object before initializing the unlogged relation. This is + * safe as far as no other backends have accessed the relation before + * starting archive recovery. + */ + HASH_SEQ_STATUS status; + relfile_entry *ent; + SMgrRelation *srels = palloc(sizeof(SMgrRelation) * 8); + int maxrels = 8; + int nrels = 0; + int i; + + Assert(!HotStandbyActive()); + + hash_seq_init(&status, hash); + while ((ent = (relfile_entry *) hash_seq_search(&status)) != NULL) + { + RelFileLocatorBackend rel; + + /* + * The relation is persistent and stays persistent. Don't drop the + * buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + + if (maxrels <= nrels) + { + maxrels *= 2; + srels = repalloc(srels, sizeof(SMgrRelation) * maxrels); + } + + rel.backend = InvalidBackendId; + rel.locator.spcOid = tspid; + rel.locator.dbOid = dbid; + rel.locator.relNumber = ent->relNumber; + + srels[nrels++] = smgropen(rel.locator, InvalidBackendId); + } + + DropRelationsAllBuffers(srels, nrels); + + for (i = 0; i < nrels; i++) + smgrclose(srels[i]); + } + + /* + * Now, make a second pass and remove anything that matches. + */ if ((op & UNLOGGED_RELATION_CLEANUP) != 0) { - HTAB *hash; - HASHCTL ctl; - - /* - * It's possible that someone could create a ton of unlogged relations - * in the same database & tablespace, so we'd better use a hash table - * rather than an array or linked list to keep track of which files - * need to be reset. Otherwise, this cleanup operation would be - * O(n^2). - */ - ctl.keysize = sizeof(Oid); - ctl.entrysize = sizeof(unlogged_relation_entry); - ctl.hcxt = CurrentMemoryContext; - hash = hash_create("unlogged relation OIDs", 32, &ctl, - HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); - - /* Scan the directory. */ dbspace_dir = AllocateDir(dbspacedirname); while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int relnumchars; - unlogged_relation_entry ent; + RelFileNumber key; + relfile_entry *ent; + RelFileLocatorBackend rel; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &relnumchars, - &forkNum)) - continue; - - /* Also skip it unless this is the init fork. */ - if (forkNum != INIT_FORKNUM) - continue; - - /* - * Put the OID portion of the name into the hash table, if it - * isn't already. - */ - ent.reloid = atooid(de->d_name); - (void) hash_search(hash, &ent, HASH_ENTER, NULL); - } - - /* Done with the first pass. */ - FreeDir(dbspace_dir); - - /* - * If we didn't find any init forks, there's no point in continuing; - * we can bail out now. - */ - if (hash_get_num_entries(hash) == 0) - { - hash_destroy(hash); - return; - } - - /* - * Now, make a second pass and remove anything that matches. - */ - dbspace_dir = AllocateDir(dbspacedirname); - while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) - { - ForkNumber forkNum; - int relnumchars; - unlogged_relation_entry ent; - - /* Skip anything that doesn't look like a relation data file. */ - if (!parse_filename_for_nontemp_relation(de->d_name, &relnumchars, - &forkNum)) - continue; - - /* We never remove the init fork. */ - if (forkNum == INIT_FORKNUM) + &forkNum, &mark)) continue; /* * See whether the OID portion of the name shows up in the hash * table. If so, nuke it! */ - ent.reloid = atooid(de->d_name); - if (hash_search(hash, &ent, HASH_FIND, NULL)) + key = atooid(de->d_name); + ent = hash_search(hash, &key, HASH_FIND, NULL); + + if (!ent) + continue; + + if (!ent->dirty_all) { - snprintf(rm_path, sizeof(rm_path), "%s/%s", - dbspacedirname, de->d_name); - if (unlink(rm_path) < 0) - ereport(ERROR, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", - rm_path))); + /* clean permanent relations don't need cleanup */ + if (!ent->has_init) + continue; + + if (ent->dirty_init) + { + /* + * The crashed transaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } else - elog(DEBUG2, "unlinked file \"%s\"", rm_path); + { + /* + * we don't remove the INIT fork of a non-dirty + * relation files. + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } + + /* so, nuke it! */ + snprintf(rm_path, sizeof(rm_path), "%s/%s", + dbspacedirname, de->d_name); + if (unlink(rm_path) < 0) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", + rm_path)); + + rel.backend = InvalidBackendId; + rel.locator.spcOid = tspid; + rel.locator.dbOid = dbid; + rel.locator.relNumber = atooid(de->d_name); + + ForgetRelationForkSyncRequests(rel, forkNum); } /* Cleanup is complete. */ FreeDir(dbspace_dir); - hash_destroy(hash); } /* * Initialization happens after cleanup is complete: we copy each init - * fork file to the corresponding main fork file. Note that if we are - * asked to do both cleanup and init, we may never get here: if the - * cleanup code determines that there are no init forks in this dbspace, - * it will return before we get to this point. + * fork file to the corresponding main fork file. */ if ((op & UNLOGGED_RELATION_INIT) != 0) { @@ -285,6 +417,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int relnumchars; char relnumbuf[OIDCHARS + 1]; char srcpath[MAXPGPATH * 2]; @@ -292,9 +425,11 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &relnumchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -328,15 +463,18 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) while ((de = ReadDir(dbspace_dir, dbspacedirname)) != NULL) { ForkNumber forkNum; + StorageMarks mark; int relnumchars; char relnumbuf[OIDCHARS + 1]; char mainpath[MAXPGPATH]; /* Skip anything that doesn't look like a relation data file. */ if (!parse_filename_for_nontemp_relation(de->d_name, &relnumchars, - &forkNum)) + &forkNum, &mark)) continue; + Assert(mark == SMGR_MARK_NONE); + /* Also skip it unless this is the init fork. */ if (forkNum != INIT_FORKNUM) continue; @@ -379,7 +517,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, int op) */ bool parse_filename_for_nontemp_relation(const char *name, int *relnumchars, - ForkNumber *fork) + ForkNumber *fork, StorageMarks *mark) { int pos; @@ -410,11 +548,19 @@ parse_filename_for_nontemp_relation(const char *name, int *relnumchars, for (segchar = 1; isdigit((unsigned char) name[pos + segchar]); ++segchar) ; - if (segchar <= 1) - return false; - pos += segchar; + if (segchar > 1) + pos += segchar; } + /* mark file? */ + if (name[pos] == '.' && name[pos + 1] != 0) + { + *mark = name[pos + 1]; + pos += 2; + } + else + *mark = SMGR_MARK_NONE; + /* Now we should be at the end. */ if (name[pos] != '\0') return false; diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index 14b6fa0fd9..094b4191c4 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -141,7 +141,8 @@ static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno, bool skipFsync, int behavior); static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg); - +static bool mdmarkexists(SMgrRelation reln, ForkNumber forkNum, + StorageMarks mark); /* * mdinit() -- Initialize private state for magnetic disk storage manager. @@ -173,6 +174,82 @@ mdexists(SMgrRelation reln, ForkNumber forknum) return (mdopenfork(reln, forknum, EXTENSION_RETURN_NULL) != NULL); } +/* + * mdcreatemark() -- Create a mark file. + * + * If isRedo is true, it's okay for the file to exist already. + */ +void +mdcreatemark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + int fd; + + /* See mdcreate for details.. */ + TablespaceCreateDbspace(reln->smgr_rlocator.locator.spcOid, + reln->smgr_rlocator.locator.dbOid, + isRedo); + + fd = BasicOpenFile(path, O_WRONLY | O_CREAT | O_EXCL); + if (fd < 0 && (!isRedo || errno != EEXIST)) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not create mark file \"%s\": %m", path)); + + pg_fsync(fd); + close(fd); + + /* + * To guarantee that the creation of the file is persistent, fsync its + * parent directory. + */ + fsync_parent_path(path, ERROR); + + pfree(path); +} + + +/* + * mdunlinkmark() -- Delete the mark file + * + * If isRedo is true, it's okay for the file being not found. + */ +void +mdunlinkmark(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark, + bool isRedo) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + + if (!isRedo || mdmarkexists(reln, forkNum, mark)) + durable_unlink(path, ERROR); + + pfree(path); +} + +/* + * mdmarkexists() -- Check if the file exists. + */ +static bool +mdmarkexists(SMgrRelation reln, ForkNumber forkNum, StorageMarks mark) +{ + char *path = markpath(reln->smgr_rlocator, forkNum, mark); + int fd; + + fd = BasicOpenFile(path, O_RDONLY); + if (fd < 0 && errno != ENOENT) + ereport(ERROR, + errcode_for_file_access(), + errmsg("could not access mark file \"%s\": %m", path)); + pfree(path); + + if (fd < 0) + return false; + + close(fd); + return true; +} + /* * mdcreate() -- Create a new relation on magnetic disk. * @@ -1063,6 +1140,16 @@ register_forget_request(RelFileLocatorBackend rlocator, ForkNumber forknum, RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ ); } +/* + * ForgetRelationForkSyncRequests -- forget any fsyncs and unlinks for a fork + */ +void +ForgetRelationForkSyncRequests(RelFileLocatorBackend rlocator, + ForkNumber forknum) +{ + register_forget_request(rlocator, forknum, 0); +} + /* * ForgetDatabaseSyncRequests -- forget any fsyncs and unlinks for a DB */ @@ -1421,12 +1508,14 @@ mdsyncfiletag(const FileTag *ftag, char *path) * Return 0 on success, -1 on failure, with errno set. */ int -mdunlinkfiletag(const FileTag *ftag, char *path) +mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark) { char *p; /* Compute the path. */ - p = relpathperm(ftag->rlocator, MAIN_FORKNUM); + p = GetRelationPath(ftag->rlocator.dbOid, ftag->rlocator.spcOid, + ftag->rlocator.relNumber,InvalidBackendId, + MAIN_FORKNUM, mark); strlcpy(path, p, MAXPGPATH); pfree(p); diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c index c1a5febcbf..5044dc21e4 100644 --- a/src/backend/storage/smgr/smgr.c +++ b/src/backend/storage/smgr/smgr.c @@ -62,6 +62,10 @@ typedef struct f_smgr void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum); + void (*smgr_createmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); + void (*smgr_unlinkmark) (SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); } f_smgr; static const f_smgr smgrsw[] = { @@ -82,6 +86,8 @@ static const f_smgr smgrsw[] = { .smgr_nblocks = mdnblocks, .smgr_truncate = mdtruncate, .smgr_immedsync = mdimmedsync, + .smgr_createmark = mdcreatemark, + .smgr_unlinkmark = mdunlinkmark, } }; @@ -371,6 +377,26 @@ smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo) smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo); } +/* + * smgrcreatemark() -- Create a mark file + */ +void +smgrcreatemark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_createmark(reln, forknum, mark, isRedo); +} + +/* + * smgrunlinkmark() -- Delete a mark file + */ +void +smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, StorageMarks mark, + bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlinkmark(reln, forknum, mark, isRedo); +} + /* * smgrdosyncall() -- Immediately sync all forks of all given relations * @@ -693,6 +719,12 @@ smgrimmedsync(SMgrRelation reln, ForkNumber forknum) smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum); } +void +smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo) +{ + smgrsw[reln->smgr_which].smgr_unlink(reln->smgr_rlocator, forknum, isRedo); +} + /* * AtEOXact_SMgr * diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c index 9d6a9e9109..74357efb1c 100644 --- a/src/backend/storage/sync/sync.c +++ b/src/backend/storage/sync/sync.c @@ -91,7 +91,8 @@ static CycleCtr checkpoint_cycle_ctr = 0; typedef struct SyncOps { int (*sync_syncfiletag) (const FileTag *ftag, char *path); - int (*sync_unlinkfiletag) (const FileTag *ftag, char *path); + int (*sync_unlinkfiletag) (const FileTag *ftag, char *path, + StorageMarks mark); bool (*sync_filetagmatches) (const FileTag *ftag, const FileTag *candidate); } SyncOps; @@ -235,7 +236,8 @@ SyncPostCheckpoint(void) /* Unlink the file */ if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, - path) < 0) + path, + SMGR_MARK_NONE) < 0) { /* * There's a race condition, when the database is dropped at the @@ -244,6 +246,21 @@ SyncPostCheckpoint(void) * here. rmtree() also has to ignore ENOENT errors, to deal with * the possibility that we delete the file first. */ + if (errno != ENOENT) + ereport(WARNING, + errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path)); + } + else if (syncsw[entry->tag.handler].sync_unlinkfiletag(&entry->tag, + path, + SMGR_MARK_UNCOMMITTED) + < 0) + { + /* + * And we may have SMGR_MARK_UNCOMMITTED file. Remove it if the + * fork file has been successfully removed. It's ok if the file + * does not exist. + */ if (errno != ENOENT) ereport(WARNING, (errcode_for_file_access(), diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 53f011a2fe..4da62f71ce 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -418,6 +418,28 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_UNLINK) + { + /* + * We can safely ignore there. We'll see that the file don't exist in + * the target data dir, and copy them in from the source system. No + * need to do anything special here. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_MARK) + { + /* + * We can safely ignore these, The file will be removed from the + * target, if it doesn't exist in the source system. The files are + * empty so we don't need to bother the content. + */ + } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/bin/pg_rewind/pg_rewind.c b/src/bin/pg_rewind/pg_rewind.c index 3cd77c09b1..f87dfda8fe 100644 --- a/src/bin/pg_rewind/pg_rewind.c +++ b/src/bin/pg_rewind/pg_rewind.c @@ -419,7 +419,6 @@ main(int argc, char **argv) if (showprogress) pg_log_info("reading source file list"); source->traverse_files(source, &process_source_file); - if (showprogress) pg_log_info("reading target file list"); traverse_datadir(datadir_target, &process_target_file); diff --git a/src/common/relpath.c b/src/common/relpath.c index 1b6b620ce8..46cfe38fd5 100644 --- a/src/common/relpath.c +++ b/src/common/relpath.c @@ -139,9 +139,15 @@ GetDatabasePath(Oid dbOid, Oid spcOid) */ char * GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, - int backendId, ForkNumber forkNumber) + int backendId, ForkNumber forkNumber, char mark) { char *path; + char markstr[4]; + + if (mark == 0) + markstr[0] = 0; + else + snprintf(markstr, sizeof(markstr), ".%c", mark); if (spcOid == GLOBALTABLESPACE_OID) { @@ -149,10 +155,10 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, Assert(dbOid == 0); Assert(backendId == InvalidBackendId); if (forkNumber != MAIN_FORKNUM) - path = psprintf("global/%u_%s", - relNumber, forkNames[forkNumber]); + path = psprintf("global/%u_%s%s", + relNumber, forkNames[forkNumber], markstr); else - path = psprintf("global/%u", relNumber); + path = psprintf("global/%u%s", relNumber, markstr); } else if (spcOid == DEFAULTTABLESPACE_OID) { @@ -160,22 +166,22 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/%u_%s", + path = psprintf("base/%u/%u_%s%s", dbOid, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/%u", - dbOid, relNumber); + path = psprintf("base/%u/%u%s", + dbOid, relNumber, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("base/%u/t%d_%u_%s", + path = psprintf("base/%u/t%d_%u_%s%s", dbOid, backendId, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("base/%u/t%d_%u", - dbOid, backendId, relNumber); + path = psprintf("base/%u/t%d_%u%s", + dbOid, backendId, relNumber, markstr); } } else @@ -184,27 +190,28 @@ GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, if (backendId == InvalidBackendId) { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/%u_%s%s", spcOid, TABLESPACE_VERSION_DIRECTORY, dbOid, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/%u", + path = psprintf("pg_tblspc/%u/%s/%u/%u%s", spcOid, TABLESPACE_VERSION_DIRECTORY, - dbOid, relNumber); + dbOid, relNumber, markstr); } else { if (forkNumber != MAIN_FORKNUM) - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u_%s%s", spcOid, TABLESPACE_VERSION_DIRECTORY, dbOid, backendId, relNumber, - forkNames[forkNumber]); + forkNames[forkNumber], markstr); else - path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u", + path = psprintf("pg_tblspc/%u/%s/%u/t%d_%u%s", spcOid, TABLESPACE_VERSION_DIRECTORY, - dbOid, backendId, relNumber); + dbOid, backendId, relNumber, markstr); } } + return path; } diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h index 9964c312aa..ee4179699a 100644 --- a/src/include/catalog/storage.h +++ b/src/include/catalog/storage.h @@ -25,6 +25,8 @@ extern PGDLLIMPORT int wal_skip_threshold; extern SMgrRelation RelationCreateStorage(RelFileLocator rlocator, char relpersistence, bool register_delete); +extern void RelationCreateInitFork(Relation rel); +extern void RelationDropInitFork(Relation rel); extern void RelationDropStorage(Relation rel); extern void RelationPreserveStorage(RelFileLocator rlocator, bool atCommit); extern void RelationPreTruncate(Relation rel); @@ -43,6 +45,7 @@ extern void RestorePendingSyncs(char *startAddress); extern void smgrDoPendingDeletes(bool isCommit); extern void smgrDoPendingSyncs(bool isCommit, bool isParallelWorker); extern int smgrGetPendingDeletes(bool forCommit, RelFileLocator **ptr); +extern void smgrDoPendingCleanups(bool isCommit); extern void AtSubCommit_smgr(void); extern void AtSubAbort_smgr(void); extern void PostPrepare_smgr(void); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 44a5e2043b..9f48fb5e6f 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -18,17 +18,23 @@ #include "lib/stringinfo.h" #include "storage/block.h" #include "storage/relfilelocator.h" +#include "storage/smgr.h" /* * Declarations for smgr-related XLOG records * - * Note: we log file creation and truncation here, but logging of deletion - * actions is handled by xact.c, because it is part of transaction commit. + * Note: we log file creation, truncation and buffer persistence change here, + * but logging of deletion actions is handled mainly by xact.c, because it is + * part of transaction commit in most cases. However, there's a case where + * init forks are deleted outside control of transaction. */ /* XLOG gives us high 4 bits */ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 +#define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_MARK 0x40 +#define XLOG_SMGR_BUFPERSISTENCE 0x50 typedef struct xl_smgr_create { @@ -36,6 +42,32 @@ typedef struct xl_smgr_create ForkNumber forkNum; } xl_smgr_create; +typedef struct xl_smgr_unlink +{ + RelFileLocator rlocator; + ForkNumber forkNum; +} xl_smgr_unlink; + +typedef enum smgr_mark_action +{ + XLOG_SMGR_MARK_CREATE = 'c', + XLOG_SMGR_MARK_UNLINK = 'u' +} smgr_mark_action; + +typedef struct xl_smgr_mark +{ + RelFileLocator rlocator; + ForkNumber forkNum; + StorageMarks mark; + smgr_mark_action action; +} xl_smgr_mark; + +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -51,6 +83,13 @@ typedef struct xl_smgr_truncate } xl_smgr_truncate; extern void log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrcreatemark(const RelFileLocator *rlocator, + ForkNumber forkNum, StorageMarks mark); +extern void log_smgrunlinkmark(const RelFileLocator *rlocator, + ForkNumber forkNum, StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileLocator *rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h index 4bbd94393c..4281d35d1a 100644 --- a/src/include/common/relpath.h +++ b/src/include/common/relpath.h @@ -74,7 +74,7 @@ extern int forkname_chars(const char *str, ForkNumber *fork); extern char *GetDatabasePath(Oid dbOid, Oid spcOid); extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, - int backendId, ForkNumber forkNumber); + int backendId, ForkNumber forkNumber, char mark); /* * Wrapper macros for GetRelationPath. Beware of multiple @@ -84,7 +84,7 @@ extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, /* First argument is a RelFileLocator */ #define relpathbackend(rlocator, backend, forknum) \ GetRelationPath((rlocator).dbOid, (rlocator).spcOid, (rlocator).relNumber, \ - backend, forknum) + backend, forknum, 0) /* First argument is a RelFileLocator */ #define relpathperm(rlocator, forknum) \ @@ -94,4 +94,9 @@ extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber, #define relpath(rlocator, forknum) \ relpathbackend((rlocator).locator, (rlocator).backend, forknum) +/* First argument is a RelFileLocatorBackend */ +#define markpath(rlocator, forknum, mark) \ + GetRelationPath((rlocator).locator.dbOid, (rlocator).locator.spcOid, \ + (rlocator).locator.relNumber, \ + (rlocator).backend, forknum, mark) #endif /* RELPATH_H */ diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index e1bd22441b..036fdc35b3 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -151,6 +151,8 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index c0a212487d..7c0fa86b15 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -179,6 +179,7 @@ extern void pg_flush_data(int fd, off_t offset, off_t nbytes); extern int pg_truncate(const char *path, off_t length); extern void fsync_fname(const char *fname, bool isdir); extern int fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel); +extern int fsync_parent_path(const char *fname, int elevel); extern int durable_rename(const char *oldfile, const char *newfile, int elevel); extern int durable_unlink(const char *fname, int elevel); extern void SyncDataDirectory(void); diff --git a/src/include/storage/md.h b/src/include/storage/md.h index 10aa1b0109..294a09444c 100644 --- a/src/include/storage/md.h +++ b/src/include/storage/md.h @@ -23,6 +23,10 @@ extern void mdinit(void); extern void mdopen(SMgrRelation reln); extern void mdclose(SMgrRelation reln, ForkNumber forknum); +extern void mdcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void mdunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern bool mdexists(SMgrRelation reln, ForkNumber forknum); extern void mdunlink(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo); @@ -41,12 +45,14 @@ extern void mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks); extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum); +extern void ForgetRelationForkSyncRequests(RelFileLocatorBackend rlocator, + ForkNumber forknum); extern void ForgetDatabaseSyncRequests(Oid dbid); extern void DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo); /* md sync callbacks */ extern int mdsyncfiletag(const FileTag *ftag, char *path); -extern int mdunlinkfiletag(const FileTag *ftag, char *path); +extern int mdunlinkfiletag(const FileTag *ftag, char *path, StorageMarks mark); extern bool mdfiletagmatches(const FileTag *ftag, const FileTag *candidate); #endif /* MD_H */ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index b990d28d38..cacd5e7d2c 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -16,14 +16,16 @@ #define REINIT_H #include "common/relpath.h" - +#include "storage/smgr.h" extern void ResetUnloggedRelations(int op); extern bool parse_filename_for_nontemp_relation(const char *name, int *relnumchars, - ForkNumber *fork); + ForkNumber *fork, + StorageMarks *mark); #define UNLOGGED_RELATION_CLEANUP 0x0001 -#define UNLOGGED_RELATION_INIT 0x0002 +#define UNLOGGED_RELATION_DROP_BUFFER 0x0002 +#define UNLOGGED_RELATION_INIT 0x0004 #endif /* REINIT_H */ diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h index a07715356b..75b9d41e4b 100644 --- a/src/include/storage/smgr.h +++ b/src/include/storage/smgr.h @@ -18,6 +18,18 @@ #include "storage/block.h" #include "storage/relfilelocator.h" +/* + * Storage marks is a file of which existence suggests something about a + * file. The name of such files is "<filename>.<mark>", where the mark is one + * of the values of StorageMarks. Since ".<digit>" means segment files so don't + * use digits for the mark character. + */ +typedef enum StorageMarks +{ + SMGR_MARK_NONE = 0, + SMGR_MARK_UNCOMMITTED = 'u' /* the file is not committed yet */ +} StorageMarks; + /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) @@ -87,7 +99,12 @@ extern void smgrcloseall(void); extern void smgrcloserellocator(RelFileLocatorBackend rlocator); extern void smgrrelease(SMgrRelation reln); extern void smgrreleaseall(void); +extern void smgrcreatemark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); +extern void smgrunlinkmark(SMgrRelation reln, ForkNumber forknum, + StorageMarks mark, bool isRedo); extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo); +extern void smgrunlink(SMgrRelation reln, ForkNumber forknum, bool isRedo); extern void smgrdosyncall(SMgrRelation *rels, int nrels); extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo); extern void smgrextend(SMgrRelation reln, ForkNumber forknum, diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index f8302f1ed1..83b455752e 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -1970,6 +1970,7 @@ PatternInfoArray Pattern_Prefix_Status Pattern_Type PendingFsyncEntry +PendingMarkCleanup PendingRelDelete PendingRelSync PendingUnlinkEntry @@ -2600,6 +2601,7 @@ StdRdOptIndexCleanup StdRdOptions Step StopList +StorageMarks StrategyNumber StreamCtl String @@ -3609,6 +3611,7 @@ registered_buffer regmatch_t regoff_t regproc +relfile_entry relopt_bool relopt_enum relopt_enum_elt_def @@ -3662,6 +3665,7 @@ slist_iter slist_mutable_iter slist_node slock_t +smgr_mark_action socket_set socklen_t spgBulkDeleteState @@ -3861,8 +3865,11 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create +xl_smgr_mark xl_smgr_truncate +xl_smgr_unlink xl_standby_lock xl_standby_locks xl_tblspc_create_rec -- 2.31.1 ----Next_Part(Fri_Nov_18_17_25_11_2022_330)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="v25-0002-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v27 2/3] In-place table persistence change @ 2023-03-15 06:42 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2023-03-15 06:42 UTC (permalink / raw) Currently, the command cuases a large amount of file I/O due to heap rewrite, even though ALTER TABLE SET UNLOGGED does not require any data rewrites. In addition, this patch changes ALTER TABLE SET LOGGED to emit XLOG_FPI records instead of a large number of HEAP_INSERT's when wal_level > minimal, as this option is likely to be less resource intensive. --- src/backend/access/rmgrdesc/smgrdesc.c | 12 + src/backend/catalog/storage.c | 290 ++++++++++++++++++++++- src/backend/commands/tablecmds.c | 269 ++++++++++++++++++--- src/backend/storage/buffer/bufmgr.c | 85 +++++++ src/backend/storage/file/reinit.c | 51 +++- src/bin/pg_rewind/parsexlog.c | 6 + src/bin/pg_rewind/pg_rewind.c | 1 - src/include/catalog/storage_xlog.h | 8 + src/include/storage/bufmgr.h | 2 + src/test/recovery/t/013_crash_restart.pl | 21 -- src/tools/pgindent/typedefs.list | 1 + 11 files changed, 673 insertions(+), 73 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index f8187385c4..e2998a3ee4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -71,6 +71,15 @@ smgr_desc(StringInfo buf, XLogReaderState *record) appendStringInfo(buf, "%s %s", action, path); pfree(path); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -92,6 +101,9 @@ smgr_identify(uint8 info) case XLOG_SMGR_MARK: id = "MARK"; break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 03e06246be..97d1230ee8 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -69,11 +69,13 @@ typedef struct PendingRelDelete #define PCOP_UNLINK_FORK (1 << 0) #define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) typedef struct PendingCleanup { RelFileLocator rlocator; /* relation that need a cleanup */ int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ ForkNumber unlink_forknum; /* forknum to unlink */ StorageMarks unlink_mark; /* mark to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ @@ -223,6 +225,202 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If we have a pending-unlink for the init-fork of this relation, that + * means the init-fork exists since before the current transaction + * started. This function reverts that change just by removing the entry. + * See RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the mark file */ + srel = smgropen(rlocator, InvalidBackendId); + log_smgrcreatemark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * For index relations, WAL-logging and file sync are performed by + * ambuildempty. On the other hand, we manually perform these tasks here + * for heap relations. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file then revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * Search for pending-unlink associated with the init-fork of the + * relation. The presence of one indicates that the init fork was created + * within the current transaction. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + /* + * If the init-fork was created in this transaction, we immediately remove + * both the init fork and mark file. Otherwise, we register an at-commit + * pending-unlink for the existing init fork. See + * RelationCreateInitFork. + */ + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + + /* + * Some AMs initialize INIT fork via buffer manager. To properly drop + * the init fork, we need to drop all buffers for the INIT fork first, + * then unlink the INIT fork along with the mark file. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + log_smgrunlinkmark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -305,6 +503,25 @@ log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator *rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = *rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -858,10 +1075,28 @@ smgrDoPendingCleanups(bool isCommit) srel = smgropen(pending->rlocator, pending->backend); Assert((pending->op & - ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK)) == 0); + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } if (pending->op & PCOP_UNLINK_FORK) { + /* + * Unlink the fork file. Currently we only apply this + * operation for init forks and it is ceratin that the init + * fork is not loaded on shared buffers at this point. In + * the case of RelationDropInitFork, the function should + * have dropped buffers. In the case of + * RelationCreateInitFork, PCOP_SET_PERSISTENCE is set and + * the buffers were dropped just before. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + /* Don't emit wal while recovery. */ if (!InRecovery) log_smgrunlink(&pending->rlocator, @@ -1311,6 +1546,59 @@ smgr_redo(XLogReaderState *record) } } } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete pending action for persistence change if any. We should have + * at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * At abort time, revert any changes to buffer-persistence that were + * made in this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 3e2c5f797c..becef96927 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5439,6 +5440,189 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: perform in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * ATRewriteTable should be used instead of this function under the + * following condition. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * Initially we gather all relations that require persistence change. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods do not support in-place persistence + * changes. GiST uses page LSNs to figure out whether a block has been + * modified. However UNLOGGED GiST indexes use fake LSNs that are + * incompatible with the real LSNs used for LOGGED indexes. + * + * Maybe if gistGetFakeLSN behaved the same way for permanent and + * unlogged indexes, we could potentially avoid index rebuilds in + * exchange for emitting some extra WAL records while the index is + * unlogged. + * + * Check relam against a positive list so that we take the hard way for + * unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * If this relation is changed to WAL-logged, immediately sync all + * files except for init fork to establish the initial state on + * storage. The buffers should have already been flushed out by + * RelationCreate(Drop)InitFork called immediately above. The init fork + * should have already been synchronized as needed. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * If wal_level >= replica, switching to LOGGED requires the relation + * content to be WAL-logged for later recovery. We don't emit this if + * wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5569,48 +5753,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 0a05577b68..2b00ec3eed 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3240,6 +3240,91 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages of the relation out to disk when switching + * to PERMANENT. (or more precisely, to kernel disk buffers), ensuring + * that the kernel has an up-to-date view of the relation. + * + * The caller must be holding AccessExclusiveLock on the target relation + * to ensure that no other backend is busy dirtying more blocks of the + * relation. + * + * XXX currently it sequentially searches the buffer pool, should be + * changed to more clever ways of searching. This routine is not used in + * any performance-critical code paths, so it's not worth additional + * overhead to make it go faster; but see also DropRelationBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* The init fork is being dropped, drop buffers for it. */ + if (BufTagGetForkNum(&bufHdr->tag) == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(BufTagGetForkNum(&bufHdr->tag) != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index 250cfe9e44..bdd1200132 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -38,6 +38,7 @@ typedef struct { RelFileNumber relNumber; /* hash key */ bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ bool dirty_all; /* needs to remove all forks */ } relfile_entry; @@ -45,7 +46,10 @@ typedef struct * Clean up and reset relation files from before the last restart. * * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations - * depending on the existence of mark files. + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. * * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the * whole relation along with the mark file. @@ -54,7 +58,7 @@ typedef struct * with the "init" fork, except for the "init" fork itself. * * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all - * relations that are to be cleaned up. + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -241,7 +245,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, * Put the OID portion of the name into the hash table, * if it isn't already. If it has SMGR_MARK_UNCOMMITTED mark * files, the storage file is in dirty state, where clean up is - * needed. + * needed. isn't already. */ key = atooid(de->d_name); ent = hash_search(hash, &key, HASH_ENTER, &found); @@ -249,10 +253,13 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, if (!found) { ent->has_init = false; + ent->dirty_init = false; ent->dirty_all = false; } - if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) ent->dirty_all = true; else { @@ -276,11 +283,10 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, { /* * When we come here after recovery, smgr object for this file might - * have been created. In that case we need to drop all buffers then the - * smgr object. Otherwise checkpointer wrongly tries to flush buffers - * for nonexistent relation storage. This is safe as far as no other - * backends have accessed the relation before starting archive - * recovery. + * have been created. In that case we need to drop all buffers then + * the smgr object before initializing the unlogged relation. This is + * safe as far as no other backends have accessed the relation before + * starting archive recovery. */ HASH_SEQ_STATUS status; relfile_entry *ent; @@ -296,6 +302,13 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, { RelFileLocatorBackend rel; + /* + * The relation is persistent and stays persistent. Don't drop the + * buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + if (maxrels <= nrels) { maxrels *= 2; @@ -352,8 +365,24 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, if (!ent->has_init) continue; - if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) - continue; + if (ent->dirty_init) + { + /* + * The crashed transaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } + else + { + /* + * we don't remove the INIT fork of a non-dirty + * relation files. + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } /* so, nuke it! */ diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index e9e4bafb01..ddc8014e55 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -434,6 +434,12 @@ extractPageInfo(XLogReaderState *record) * empty so we don't need to bother the content. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/bin/pg_rewind/pg_rewind.c b/src/bin/pg_rewind/pg_rewind.c index f7f3b8227f..b3a1f255d7 100644 --- a/src/bin/pg_rewind/pg_rewind.c +++ b/src/bin/pg_rewind/pg_rewind.c @@ -460,7 +460,6 @@ main(int argc, char **argv) if (showprogress) pg_log_info("reading source file list"); source->traverse_files(source, &process_source_file); - if (showprogress) pg_log_info("reading target file list"); traverse_datadir(datadir_target, &process_target_file); diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index a36646c6ee..6e79c68f5b 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -62,6 +62,12 @@ typedef struct xl_smgr_mark smgr_mark_action action; } xl_smgr_mark; +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -82,6 +88,8 @@ extern void log_smgrcreatemark(const RelFileLocator *rlocator, ForkNumber forkNum, StorageMarks mark); extern void log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileLocator *rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index b8a18b8081..fd34810dc2 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -156,6 +156,8 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/test/recovery/t/013_crash_restart.pl b/src/test/recovery/t/013_crash_restart.pl index 9def8d2062..92e7b367df 100644 --- a/src/test/recovery/t/013_crash_restart.pl +++ b/src/test/recovery/t/013_crash_restart.pl @@ -86,24 +86,6 @@ ok( pump_until( $killme_stdout = ''; $killme_stderr = ''; -#create a table that should *not* survive, but has rows. -#the table's contents is requried to cause access to the storage file -#after a restart. -$killme_stdin .= q[ -CREATE TABLE not_alive AS SELECT 1 as a; -SELECT pg_relation_filepath('not_alive'); -]; -ok( pump_until( - $killme, $psql_timeout, - \$killme_stdout, qr/[[:alnum:]\/]+[\r\n]$/m), - 'added in-creation table'); -my $not_alive_relfile = $node->data_dir . "/" . $killme_stdout; -chomp($not_alive_relfile); -$killme_stdout = ''; -$killme_stderr = ''; - -# The relfile must be exists now -ok ( -e $not_alive_relfile, 'relfile for in-creation table'); # Start longrunning query in second session; its failure will signal that # crash-restart has occurred. The initial wait for the trivial select is to @@ -162,9 +144,6 @@ $killme->run(); ($monitor_stdin, $monitor_stdout, $monitor_stderr) = ('', '', ''); $monitor->run(); -# The relfile must have been removed due to the recent restart. -ok ( ! -e $not_alive_relfile, - 'relfile for the in-creation table should be removed after restart'); # Acquire pid of new backend $killme_stdin .= q[ diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 747b7557dc..8dbbb09e8c 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -3886,6 +3886,7 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create xl_smgr_mark xl_smgr_truncate -- 2.31.1 ----Next_Part(Fri_Mar_17_15_16_34_2023_294)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="v27-0003-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v28 2/4] In-place table persistence change @ 2023-04-25 06:49 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2023-04-25 06:49 UTC (permalink / raw) Previously, the command caused a large amount of file I/O due to heap rewrites, even though ALTER TABLE SET UNLOGGED does not require any data rewrites. This patch eliminates the need for rewrites. Additionally, ALTER TABLE SET LOGGED is updated to emit XLOG_FPI records instead of numerous HEAP_INSERTs when wal_level > minimal, reducing resource consumption. --- src/backend/access/rmgrdesc/smgrdesc.c | 12 + src/backend/catalog/storage.c | 295 ++++++++++++++++++++++++- src/backend/commands/tablecmds.c | 268 ++++++++++++++++++---- src/backend/storage/buffer/bufmgr.c | 84 +++++++ src/backend/storage/file/reinit.c | 51 ++++- src/bin/pg_rewind/parsexlog.c | 6 + src/include/catalog/storage_xlog.h | 8 + src/include/storage/bufmgr.h | 2 + src/tools/pgindent/typedefs.list | 1 + 9 files changed, 674 insertions(+), 53 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index f8187385c4..e2998a3ee4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -71,6 +71,15 @@ smgr_desc(StringInfo buf, XLogReaderState *record) appendStringInfo(buf, "%s %s", action, path); pfree(path); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -92,6 +101,9 @@ smgr_identify(uint8 info) case XLOG_SMGR_MARK: id = "MARK"; break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index fe06c3c31d..6106376525 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -69,11 +69,13 @@ typedef struct PendingRelDelete #define PCOP_UNLINK_FORK (1 << 0) #define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) typedef struct PendingCleanup { RelFileLocator rlocator; /* relation that need a cleanup */ int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ ForkNumber unlink_forknum; /* forknum to unlink */ StorageMarks unlink_mark; /* mark to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ @@ -223,6 +225,202 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If a pending-unlink exists for this relation's init-fork, it indicates + * the init-fork's existed before the current transaction; this function + * reverts the pending-unlink by removing the entry. See + * RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the mark file */ + srel = smgropen(rlocator, InvalidBackendId); + log_smgrcreatemark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * For index relations, WAL-logging and file sync are handled by + * ambuildempty. In contrast, for heap relations, these tasks are performed + * directly. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file then revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * Search for a pending-unlink associated with the init-fork of the + * relation. Its presence indicates that the init-fork was created within + * the current transaction. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + /* + * If the init-fork was created in this transaction, remove both the + * init-fork and mark file. Otherwise, register an at-commit pending-unlink + * for the existing init-fork. See RelationCreateInitFork. + */ + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + + /* + * Some AMs initialize init-fork via the buffer manager. To properly + * drop the init-fork, first drop all buffers for the init-fork, then + * unlink the init-fork and the mark file. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + log_smgrunlinkmark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -305,6 +503,25 @@ log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -858,10 +1075,29 @@ smgrDoPendingCleanups(bool isCommit) srel = smgropen(pending->rlocator, pending->backend); Assert((pending->op & - ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK)) == 0); + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } if (pending->op & PCOP_UNLINK_FORK) { + BlockNumber firstblock = 0; + + /* + * Unlink the fork file. Currently this operation is + * applied only to init-forks. As it is not ceratin that + * the init-fork is not loaded on shared buffers, drop all + * buffers for it. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + DropRelationBuffers(srel, &pending->unlink_forknum, 1, + &firstblock); + /* Don't emit wal while recovery. */ if (!InRecovery) log_smgrunlink(&pending->rlocator, @@ -1286,8 +1522,8 @@ smgr_redo(XLogReaderState *record) else { /* - * Delete pending action for this mark file if any. We should have - * at most one entry for this action. + * Delete any pending action for this mark file, if present. There + * should be at most one entry for this action. */ PendingCleanup *prev = NULL; @@ -1311,6 +1547,59 @@ smgr_redo(XLogReaderState *record) } } } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 343fe61115..26446db085 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5464,6 +5465,188 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: perform in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Use ATRewriteTable instead of this function under the following + * condition. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * Initially, gather all relations that require a persistence change. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods don't support in-place persistence + * changes. GiST uses page LSNs to figure out whether a block has been + * modified. However, UNLOGGED GiST indexes use fake LSNs, which are + * incompatible with the real LSNs used for LOGGED indexes. + * + * Potentially, if gistGetFakeLSN behaved similarly for both permanent + * and unlogged indexes, we could avoid index rebuilds by emitting + * extra WAL records while the index is unlogged. + * + * Compare relam against a positive list to ensure the hard way is + * taken for unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * If this relation becomes WAL-logged, immediately sync all files + * except the init-fork to establish the initial state on storage. The + * buffers should have already been flushed out by + * RelationCreate(Drop)InitFork called just above. The init-fork should + * already be synchronized as required. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * If wal_level >= replica, switching to LOGGED necessitates WAL-logging + * the relation content for later recovery. This is not emitted when + * wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5594,48 +5777,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 1fa689052e..14f42c283f 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3702,6 +3702,90 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages to disk (or kernel disk buffers) when + * switching to PERMANENT, ensuring the kernel has an up-to-date view of + * the relation. + * + * The caller must be holding AccessExclusiveLock on the target relation + * to ensure no other backend is busy dirtying more blocks. + * + * XXX currently it sequentially searches the buffer pool; consider + * implementing more efficient search methods. This routine isn't used in + * performance-critical code paths, so it's not worth additional overhead + * to make it go faster; see also DropRelationBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* The init fork is being dropped, drop buffers for it. */ + if (BufTagGetForkNum(&bufHdr->tag) == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel, IOOBJECT_RELATION, IOCONTEXT_NORMAL); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(BufTagGetForkNum(&bufHdr->tag) != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index e84fcbf884..a5d8763e15 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -38,6 +38,7 @@ typedef struct { RelFileNumber relNumber; /* hash key */ bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ bool dirty_all; /* needs to remove all forks */ } relfile_entry; @@ -45,7 +46,10 @@ typedef struct * Clean up and reset relation files from before the last restart. * * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations - * depending on the existence of mark files. + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. * * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the * whole relation along with the mark file. @@ -54,7 +58,7 @@ typedef struct * with the "init" fork, except for the "init" fork itself. * * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all - * relations that are to be cleaned up. + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -241,7 +245,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, * Put the OID portion of the name into the hash table, * if it isn't already. If it has SMGR_MARK_UNCOMMITTED mark * files, the storage file is in dirty state, where clean up is - * needed. + * needed. isn't already. */ key = atooid(de->d_name); ent = hash_search(hash, &key, HASH_ENTER, &found); @@ -249,10 +253,13 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, if (!found) { ent->has_init = false; + ent->dirty_init = false; ent->dirty_all = false; } - if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) ent->dirty_all = true; else { @@ -276,11 +283,10 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, { /* * When we come here after recovery, smgr object for this file might - * have been created. In that case we need to drop all buffers then the - * smgr object. Otherwise checkpointer wrongly tries to flush buffers - * for nonexistent relation storage. This is safe as far as no other - * backends have accessed the relation before starting archive - * recovery. + * have been created. In that case we need to drop all buffers then + * the smgr object before initializing the unlogged relation. This is + * safe as far as no other backends have accessed the relation before + * starting archive recovery. */ HASH_SEQ_STATUS status; relfile_entry *ent; @@ -296,6 +302,13 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, { RelFileLocatorBackend rel; + /* + * The relation is persistent and stays persistent. Don't drop the + * buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + if (maxrels <= nrels) { maxrels *= 2; @@ -352,8 +365,24 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, if (!ent->has_init) continue; - if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) - continue; + if (ent->dirty_init) + { + /* + * The crashed transaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } + else + { + /* + * we don't remove the INIT fork of a non-dirty + * relation files. + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } /* so, nuke it! */ diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index e9e4bafb01..ddc8014e55 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -434,6 +434,12 @@ extractPageInfo(XLogReaderState *record) * empty so we don't need to bother the content. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index a36646c6ee..847660b6af 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -62,6 +62,12 @@ typedef struct xl_smgr_mark smgr_mark_action action; } xl_smgr_mark; +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -82,6 +88,8 @@ extern void log_smgrcreatemark(const RelFileLocator *rlocator, ForkNumber forkNum, StorageMarks mark); extern void log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileLocator rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index 6ab00daa2e..2440803a6e 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -222,6 +222,8 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 72e81c084c..3bdbb189a3 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -3896,6 +3896,7 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create xl_smgr_mark xl_smgr_truncate -- 2.31.1 ----Next_Part(Tue_Apr_25_16_54_57_2023_250)-- Content-Type: Text/X-Patch; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="v28-0003-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v29 2/3] In-place table persistence change @ 2023-04-25 06:49 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2023-04-25 06:49 UTC (permalink / raw) Previously, the command caused a large amount of file I/O due to heap rewrites, even though ALTER TABLE SET UNLOGGED does not require any data rewrites. This patch eliminates the need for rewrites. Additionally, ALTER TABLE SET LOGGED is updated to emit XLOG_FPI records instead of numerous HEAP_INSERTs when wal_level > minimal, reducing resource consumption. --- src/backend/access/rmgrdesc/smgrdesc.c | 12 + src/backend/catalog/storage.c | 295 ++++++++++++++++++++++++- src/backend/commands/tablecmds.c | 270 ++++++++++++++++++---- src/backend/storage/buffer/bufmgr.c | 84 +++++++ src/backend/storage/file/reinit.c | 51 ++++- src/bin/pg_rewind/parsexlog.c | 6 + src/include/catalog/storage_xlog.h | 8 + src/include/storage/bufmgr.h | 2 + src/tools/pgindent/typedefs.list | 1 + 9 files changed, 675 insertions(+), 54 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index f8187385c4..e2998a3ee4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -71,6 +71,15 @@ smgr_desc(StringInfo buf, XLogReaderState *record) appendStringInfo(buf, "%s %s", action, path); pfree(path); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -92,6 +101,9 @@ smgr_identify(uint8 info) case XLOG_SMGR_MARK: id = "MARK"; break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index fe06c3c31d..6106376525 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -69,11 +69,13 @@ typedef struct PendingRelDelete #define PCOP_UNLINK_FORK (1 << 0) #define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) typedef struct PendingCleanup { RelFileLocator rlocator; /* relation that need a cleanup */ int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ ForkNumber unlink_forknum; /* forknum to unlink */ StorageMarks unlink_mark; /* mark to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ @@ -223,6 +225,202 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If a pending-unlink exists for this relation's init-fork, it indicates + * the init-fork's existed before the current transaction; this function + * reverts the pending-unlink by removing the entry. See + * RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the mark file */ + srel = smgropen(rlocator, InvalidBackendId); + log_smgrcreatemark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * For index relations, WAL-logging and file sync are handled by + * ambuildempty. In contrast, for heap relations, these tasks are performed + * directly. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file then revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * Search for a pending-unlink associated with the init-fork of the + * relation. Its presence indicates that the init-fork was created within + * the current transaction. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + /* + * If the init-fork was created in this transaction, remove both the + * init-fork and mark file. Otherwise, register an at-commit pending-unlink + * for the existing init-fork. See RelationCreateInitFork. + */ + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + + /* + * Some AMs initialize init-fork via the buffer manager. To properly + * drop the init-fork, first drop all buffers for the init-fork, then + * unlink the init-fork and the mark file. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + log_smgrunlinkmark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -305,6 +503,25 @@ log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -858,10 +1075,29 @@ smgrDoPendingCleanups(bool isCommit) srel = smgropen(pending->rlocator, pending->backend); Assert((pending->op & - ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK)) == 0); + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } if (pending->op & PCOP_UNLINK_FORK) { + BlockNumber firstblock = 0; + + /* + * Unlink the fork file. Currently this operation is + * applied only to init-forks. As it is not ceratin that + * the init-fork is not loaded on shared buffers, drop all + * buffers for it. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + DropRelationBuffers(srel, &pending->unlink_forknum, 1, + &firstblock); + /* Don't emit wal while recovery. */ if (!InRecovery) log_smgrunlink(&pending->rlocator, @@ -1286,8 +1522,8 @@ smgr_redo(XLogReaderState *record) else { /* - * Delete pending action for this mark file if any. We should have - * at most one entry for this action. + * Delete any pending action for this mark file, if present. There + * should be at most one entry for this action. */ PendingCleanup *prev = NULL; @@ -1311,6 +1547,59 @@ smgr_redo(XLogReaderState *record) } } } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 727f151750..3192a97b0e 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5464,6 +5465,188 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: perform in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Use ATRewriteTable instead of this function under the following + * condition. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * Initially, gather all relations that require a persistence change. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods don't support in-place persistence + * changes. GiST uses page LSNs to figure out whether a block has been + * modified. However, UNLOGGED GiST indexes use fake LSNs, which are + * incompatible with the real LSNs used for LOGGED indexes. + * + * Potentially, if gistGetFakeLSN behaved similarly for both permanent + * and unlogged indexes, we could avoid index rebuilds by emitting + * extra WAL records while the index is unlogged. + * + * Compare relam against a positive list to ensure the hard way is + * taken for unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * If this relation becomes WAL-logged, immediately sync all files + * except the init-fork to establish the initial state on storage. The + * buffers should have already been flushed out by + * RelationCreate(Drop)InitFork called just above. The init-fork should + * already be synchronized as required. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * If wal_level >= replica, switching to LOGGED necessitates WAL-logging + * the relation content for later recovery. This is not emitted when + * wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5594,48 +5777,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); - - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index df22aaa1c5..7d21fd5ac5 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3708,6 +3708,90 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages to disk (or kernel disk buffers) when + * switching to PERMANENT, ensuring the kernel has an up-to-date view of + * the relation. + * + * The caller must be holding AccessExclusiveLock on the target relation + * to ensure no other backend is busy dirtying more blocks. + * + * XXX currently it sequentially searches the buffer pool; consider + * implementing more efficient search methods. This routine isn't used in + * performance-critical code paths, so it's not worth additional overhead + * to make it go faster; see also DropRelationBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* The init fork is being dropped, drop buffers for it. */ + if (BufTagGetForkNum(&bufHdr->tag) == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel, IOOBJECT_RELATION, IOCONTEXT_NORMAL); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(BufTagGetForkNum(&bufHdr->tag) != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index e84fcbf884..a5d8763e15 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -38,6 +38,7 @@ typedef struct { RelFileNumber relNumber; /* hash key */ bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ bool dirty_all; /* needs to remove all forks */ } relfile_entry; @@ -45,7 +46,10 @@ typedef struct * Clean up and reset relation files from before the last restart. * * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations - * depending on the existence of mark files. + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. * * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the * whole relation along with the mark file. @@ -54,7 +58,7 @@ typedef struct * with the "init" fork, except for the "init" fork itself. * * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all - * relations that are to be cleaned up. + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -241,7 +245,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, * Put the OID portion of the name into the hash table, * if it isn't already. If it has SMGR_MARK_UNCOMMITTED mark * files, the storage file is in dirty state, where clean up is - * needed. + * needed. isn't already. */ key = atooid(de->d_name); ent = hash_search(hash, &key, HASH_ENTER, &found); @@ -249,10 +253,13 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, if (!found) { ent->has_init = false; + ent->dirty_init = false; ent->dirty_all = false; } - if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) ent->dirty_all = true; else { @@ -276,11 +283,10 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, { /* * When we come here after recovery, smgr object for this file might - * have been created. In that case we need to drop all buffers then the - * smgr object. Otherwise checkpointer wrongly tries to flush buffers - * for nonexistent relation storage. This is safe as far as no other - * backends have accessed the relation before starting archive - * recovery. + * have been created. In that case we need to drop all buffers then + * the smgr object before initializing the unlogged relation. This is + * safe as far as no other backends have accessed the relation before + * starting archive recovery. */ HASH_SEQ_STATUS status; relfile_entry *ent; @@ -296,6 +302,13 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, { RelFileLocatorBackend rel; + /* + * The relation is persistent and stays persistent. Don't drop the + * buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + if (maxrels <= nrels) { maxrels *= 2; @@ -352,8 +365,24 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, if (!ent->has_init) continue; - if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) - continue; + if (ent->dirty_init) + { + /* + * The crashed transaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } + else + { + /* + * we don't remove the INIT fork of a non-dirty + * relation files. + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } /* so, nuke it! */ diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index e9e4bafb01..ddc8014e55 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -434,6 +434,12 @@ extractPageInfo(XLogReaderState *record) * empty so we don't need to bother the content. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index a36646c6ee..847660b6af 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -62,6 +62,12 @@ typedef struct xl_smgr_mark smgr_mark_action action; } xl_smgr_mark; +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -82,6 +88,8 @@ extern void log_smgrcreatemark(const RelFileLocator *rlocator, ForkNumber forkNum, StorageMarks mark); extern void log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileLocator rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index 0f5fb6be00..e3d9273710 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -222,6 +222,8 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 057e3d3104..f72c5c8a9e 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -3939,6 +3939,7 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create xl_smgr_mark xl_smgr_truncate -- 2.25.1 --EeQfGwPcQSOJBaQU Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="v29-0003-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v29 2/3] In-place table persistence change @ 2023-04-25 06:49 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2023-04-25 06:49 UTC (permalink / raw) Previously, the command caused a large amount of file I/O due to heap rewrites, even though ALTER TABLE SET UNLOGGED does not require any data rewrites. This patch eliminates the need for rewrites. Additionally, ALTER TABLE SET LOGGED is updated to emit XLOG_FPI records instead of numerous HEAP_INSERTs when wal_level > minimal, reducing resource consumption. --- src/backend/access/rmgrdesc/smgrdesc.c | 12 + src/backend/catalog/storage.c | 295 ++++++++++++++++++++++++- src/backend/commands/tablecmds.c | 270 ++++++++++++++++++---- src/backend/storage/buffer/bufmgr.c | 84 +++++++ src/backend/storage/file/reinit.c | 51 ++++- src/bin/pg_rewind/parsexlog.c | 6 + src/include/catalog/storage_xlog.h | 8 + src/include/storage/bufmgr.h | 2 + src/tools/pgindent/typedefs.list | 1 + 9 files changed, 675 insertions(+), 54 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index f8187385c4..e2998a3ee4 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -71,6 +71,15 @@ smgr_desc(StringInfo buf, XLogReaderState *record) appendStringInfo(buf, "%s %s", action, path); pfree(path); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -92,6 +101,9 @@ smgr_identify(uint8 info) case XLOG_SMGR_MARK: id = "MARK"; break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index fe06c3c31d..6106376525 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -69,11 +69,13 @@ typedef struct PendingRelDelete #define PCOP_UNLINK_FORK (1 << 0) #define PCOP_UNLINK_MARK (1 << 1) +#define PCOP_SET_PERSISTENCE (1 << 2) typedef struct PendingCleanup { RelFileLocator rlocator; /* relation that need a cleanup */ int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ ForkNumber unlink_forknum; /* forknum to unlink */ StorageMarks unlink_mark; /* mark to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ @@ -223,6 +225,202 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If a pending-unlink exists for this relation's init-fork, it indicates + * the init-fork's existed before the current transaction; this function + * reverts the pending-unlink by removing the entry. See + * RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create the init fork, along with the mark file */ + srel = smgropen(rlocator, InvalidBackendId); + log_smgrcreatemark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrcreatemark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * For index relations, WAL-logging and file sync are handled by + * ambuildempty. In contrast, for heap relations, these tasks are performed + * directly. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file then revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* drop mark file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_MARK; + pending->unlink_forknum = INIT_FORKNUM; + pending->unlink_mark = SMGR_MARK_UNCOMMITTED; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * Search for a pending-unlink associated with the init-fork of the + * relation. Its presence indicates that the init-fork was created within + * the current transaction. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum != INIT_FORKNUM) + { + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + /* + * If the init-fork was created in this transaction, remove both the + * init-fork and mark file. Otherwise, register an at-commit pending-unlink + * for the existing init-fork. See RelationCreateInitFork. + */ + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + + /* + * Some AMs initialize init-fork via the buffer manager. To properly + * drop the init-fork, first drop all buffers for the init-fork, then + * unlink the init-fork and the mark file. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + log_smgrunlinkmark(&rlocator, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED); + smgrunlinkmark(srel, INIT_FORKNUM, SMGR_MARK_UNCOMMITTED, false); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + + /* revert buffer-persistence changes at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = false; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -305,6 +503,25 @@ log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, XLogInsert(RM_SMGR_ID, XLOG_SMGR_MARK | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -858,10 +1075,29 @@ smgrDoPendingCleanups(bool isCommit) srel = smgropen(pending->rlocator, pending->backend); Assert((pending->op & - ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK)) == 0); + ~(PCOP_UNLINK_FORK | PCOP_UNLINK_MARK | + PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } if (pending->op & PCOP_UNLINK_FORK) { + BlockNumber firstblock = 0; + + /* + * Unlink the fork file. Currently this operation is + * applied only to init-forks. As it is not ceratin that + * the init-fork is not loaded on shared buffers, drop all + * buffers for it. + */ + Assert(pending->unlink_forknum == INIT_FORKNUM); + DropRelationBuffers(srel, &pending->unlink_forknum, 1, + &firstblock); + /* Don't emit wal while recovery. */ if (!InRecovery) log_smgrunlink(&pending->rlocator, @@ -1286,8 +1522,8 @@ smgr_redo(XLogReaderState *record) else { /* - * Delete pending action for this mark file if any. We should have - * at most one entry for this action. + * Delete any pending action for this mark file, if present. There + * should be at most one entry for this action. */ PendingCleanup *prev = NULL; @@ -1311,6 +1547,59 @@ smgr_redo(XLogReaderState *record) } } } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 727f151750..3192a97b0e 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5464,6 +5465,188 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: perform in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Use ATRewriteTable instead of this function under the following + * condition. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * Initially, gather all relations that require a persistence change. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods don't support in-place persistence + * changes. GiST uses page LSNs to figure out whether a block has been + * modified. However, UNLOGGED GiST indexes use fake LSNs, which are + * incompatible with the real LSNs used for LOGGED indexes. + * + * Potentially, if gistGetFakeLSN behaved similarly for both permanent + * and unlogged indexes, we could avoid index rebuilds by emitting + * extra WAL records while the index is unlogged. + * + * Compare relam against a positive list to ensure the hard way is + * taken for unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * If this relation becomes WAL-logged, immediately sync all files + * except the init-fork to establish the initial state on storage. The + * buffers should have already been flushed out by + * RelationCreate(Drop)InitFork called just above. The init-fork should + * already be synchronized as required. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * If wal_level >= replica, switching to LOGGED necessitates WAL-logging + * the relation content for later recovery. This is not emitted when + * wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5594,48 +5777,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); - - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); + + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index df22aaa1c5..7d21fd5ac5 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3708,6 +3708,90 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages to disk (or kernel disk buffers) when + * switching to PERMANENT, ensuring the kernel has an up-to-date view of + * the relation. + * + * The caller must be holding AccessExclusiveLock on the target relation + * to ensure no other backend is busy dirtying more blocks. + * + * XXX currently it sequentially searches the buffer pool; consider + * implementing more efficient search methods. This routine isn't used in + * performance-critical code paths, so it's not worth additional overhead + * to make it go faster; see also DropRelationBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* The init fork is being dropped, drop buffers for it. */ + if (BufTagGetForkNum(&bufHdr->tag) == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel, IOOBJECT_RELATION, IOCONTEXT_NORMAL); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(BufTagGetForkNum(&bufHdr->tag) != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/backend/storage/file/reinit.c b/src/backend/storage/file/reinit.c index e84fcbf884..a5d8763e15 100644 --- a/src/backend/storage/file/reinit.c +++ b/src/backend/storage/file/reinit.c @@ -38,6 +38,7 @@ typedef struct { RelFileNumber relNumber; /* hash key */ bool has_init; /* has INIT fork */ + bool dirty_init; /* needs to remove INIT fork */ bool dirty_all; /* needs to remove all forks */ } relfile_entry; @@ -45,7 +46,10 @@ typedef struct * Clean up and reset relation files from before the last restart. * * If op includes UNLOGGED_RELATION_CLEANUP, we perform different operations - * depending on the existence of mark files. + * depending on the existence of the "cleanup" forks. + * + * If SMGR_MARK_UNCOMMITTED mark file for init fork is present, we remove the + * init fork along with the mark file. * * If SMGR_MARK_UNCOMMITTED mark file for main fork is present we remove the * whole relation along with the mark file. @@ -54,7 +58,7 @@ typedef struct * with the "init" fork, except for the "init" fork itself. * * If op includes UNLOGGED_RELATION_DROP_BUFFER, we drop all buffers for all - * relations that are to be cleaned up. + * relations that have the "cleanup" and/or the "init" forks. * * If op includes UNLOGGED_RELATION_INIT, we copy the "init" fork to the main * fork. @@ -241,7 +245,7 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, * Put the OID portion of the name into the hash table, * if it isn't already. If it has SMGR_MARK_UNCOMMITTED mark * files, the storage file is in dirty state, where clean up is - * needed. + * needed. isn't already. */ key = atooid(de->d_name); ent = hash_search(hash, &key, HASH_ENTER, &found); @@ -249,10 +253,13 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, if (!found) { ent->has_init = false; + ent->dirty_init = false; ent->dirty_all = false; } - if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) + ent->dirty_init = true; + else if (forkNum == MAIN_FORKNUM && mark == SMGR_MARK_UNCOMMITTED) ent->dirty_all = true; else { @@ -276,11 +283,10 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, { /* * When we come here after recovery, smgr object for this file might - * have been created. In that case we need to drop all buffers then the - * smgr object. Otherwise checkpointer wrongly tries to flush buffers - * for nonexistent relation storage. This is safe as far as no other - * backends have accessed the relation before starting archive - * recovery. + * have been created. In that case we need to drop all buffers then + * the smgr object before initializing the unlogged relation. This is + * safe as far as no other backends have accessed the relation before + * starting archive recovery. */ HASH_SEQ_STATUS status; relfile_entry *ent; @@ -296,6 +302,13 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, { RelFileLocatorBackend rel; + /* + * The relation is persistent and stays persistent. Don't drop the + * buffers for this relation. + */ + if (ent->has_init && ent->dirty_init) + continue; + if (maxrels <= nrels) { maxrels *= 2; @@ -352,8 +365,24 @@ ResetUnloggedRelationsInDbspaceDir(const char *dbspacedirname, if (!ent->has_init) continue; - if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) - continue; + if (ent->dirty_init) + { + /* + * The crashed transaction did SET UNLOGGED. This relation + * is restored to a LOGGED relation. + */ + if (forkNum != INIT_FORKNUM) + continue; + } + else + { + /* + * we don't remove the INIT fork of a non-dirty + * relation files. + */ + if (forkNum == INIT_FORKNUM && mark == SMGR_MARK_NONE) + continue; + } } /* so, nuke it! */ diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index e9e4bafb01..ddc8014e55 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -434,6 +434,12 @@ extractPageInfo(XLogReaderState *record) * empty so we don't need to bother the content. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index a36646c6ee..847660b6af 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -62,6 +62,12 @@ typedef struct xl_smgr_mark smgr_mark_action action; } xl_smgr_mark; +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -82,6 +88,8 @@ extern void log_smgrcreatemark(const RelFileLocator *rlocator, ForkNumber forkNum, StorageMarks mark); extern void log_smgrunlinkmark(const RelFileLocator *rlocator, ForkNumber forkNum, StorageMarks mark); +extern void log_smgrbufpersistence(const RelFileLocator rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index 0f5fb6be00..e3d9273710 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -222,6 +222,8 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 057e3d3104..f72c5c8a9e 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -3939,6 +3939,7 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create xl_smgr_mark xl_smgr_truncate -- 2.25.1 --EeQfGwPcQSOJBaQU Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="v29-0003-New-command-ALTER-TABLE-ALL-IN-TABLESPACE-SET-LO.patch" ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v29 2/2] In-place table persistence change @ 2023-09-04 08:23 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2023-09-04 08:23 UTC (permalink / raw) Previously, the command caused a large amount of file I/O due to heap rewrites, even though ALTER TABLE SET UNLOGGED does not require any data rewrites. This patch eliminates the need for rewrites. Additionally, ALTER TABLE SET LOGGED is updated to emit XLOG_FPI records instead of numerous HEAP_INSERTs when wal_level > minimal, reducing resource consumption. --- src/backend/access/rmgrdesc/smgrdesc.c | 12 + src/backend/catalog/storage.c | 338 ++++++++++++++++++++++++- src/backend/commands/tablecmds.c | 268 +++++++++++++++++--- src/backend/storage/buffer/bufmgr.c | 84 ++++++ src/bin/pg_rewind/parsexlog.c | 6 + src/include/catalog/storage_xlog.h | 10 + src/include/storage/bufmgr.h | 3 + src/include/storage/reinit.h | 2 +- src/tools/pgindent/typedefs.list | 1 + 9 files changed, 683 insertions(+), 41 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index bd841b96e8..620e02bc26 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,15 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +64,9 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 1778801bbd..e7c917c50f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -71,11 +71,13 @@ typedef struct PendingRelDelete } PendingRelDelete; #define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_SET_PERSISTENCE (1 << 1) typedef struct PendingCleanup { RelFileLocator rlocator; /* relation that need a cleanup */ int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ ForkNumber unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ @@ -209,6 +211,208 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + ul_uncommitted_storage ul_storage; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If a pending-unlink exists for this relation's init-fork, it indicates + * the init-fork's existed before the current transaction; this function + * reverts the pending-unlink by removing the entry. See + * RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + /* write cancel log for preceding undo log entry */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create undo log entry, then the init fork */ + srel = smgropen(rlocator, InvalidBackendId); + + /* write undo log */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = true; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * For index relations, WAL-logging and file sync are handled by + * ambuildempty. In contrast, for heap relations, these tasks are performed + * directly. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file then revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * Search for a pending-unlink associated with the init-fork of the + * relation. Its presence indicates that the init-fork was created within + * the current transaction. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + ul_uncommitted_storage ul_storage; + + /* write cancel log for preceding undo log entry */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + /* + * If the init-fork was created in this transaction, remove the init-fork + * and cancel preceding undo log. Otherwise, register an at-commit + * pending-unlink for the existing init-fork. See RelationCreateInitFork. + */ + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + ul_uncommitted_storage ul_storage; + + /* + * Some AMs initialize init-fork via the buffer manager. To properly + * drop the init-fork, first drop all buffers for the init-fork, then + * unlink the init-fork and cancel preceding undo log. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + + /* cancel existing undo log */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -248,6 +452,25 @@ log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -800,7 +1023,14 @@ smgrDoPendingCleanups(bool isCommit) srel = smgropen(pending->rlocator, pending->backend); - Assert((pending->op & ~(PCOP_UNLINK_FORK)) == 0); + Assert((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } if (pending->op & PCOP_UNLINK_FORK) { @@ -1200,6 +1430,112 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 47c556669f..6c4cfbfa78 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5571,6 +5572,188 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: perform in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Use ATRewriteTable instead of this function under the following + * condition. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * Initially, gather all relations that require a persistence change. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods don't support in-place persistence + * changes. GiST uses page LSNs to figure out whether a block has been + * modified. However, UNLOGGED GiST indexes use fake LSNs, which are + * incompatible with the real LSNs used for LOGGED indexes. + * + * Potentially, if gistGetFakeLSN behaved similarly for both permanent + * and unlogged indexes, we could avoid index rebuilds by emitting + * extra WAL records while the index is unlogged. + * + * Compare relam against a positive list to ensure the hard way is + * taken for unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * If this relation becomes WAL-logged, immediately sync all files + * except the init-fork to establish the initial state on storage. The + * buffers should have already been flushed out by + * RelationCreate(Drop)InitFork called just above. The init-fork should + * already be synchronized as required. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * If wal_level >= replica, switching to LOGGED necessitates WAL-logging + * the relation content for later recovery. This is not emitted when + * wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5701,48 +5884,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 3bd82dbfca..04ab6ec8a7 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3708,6 +3708,90 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages to disk (or kernel disk buffers) when + * switching to PERMANENT, ensuring the kernel has an up-to-date view of + * the relation. + * + * The caller must be holding AccessExclusiveLock on the target relation + * to ensure no other backend is busy dirtying more blocks. + * + * XXX currently it sequentially searches the buffer pool; consider + * implementing more efficient search methods. This routine isn't used in + * performance-critical code paths, so it's not worth additional overhead + * to make it go faster; see also DropRelationBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* The init fork is being dropped, drop buffers for it. */ + if (BufTagGetForkNum(&bufHdr->tag) == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel, IOOBJECT_RELATION, IOCONTEXT_NORMAL); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(BufTagGetForkNum(&bufHdr->tag) != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 87b4659e27..db12f4f397 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -418,6 +418,12 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 5122f5b61d..eaa162f0c7 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -14,6 +14,7 @@ #ifndef STORAGE_XLOG_H #define STORAGE_XLOG_H +#include "access/simpleundolog.h" #include "access/xlogreader.h" #include "lib/stringinfo.h" #include "storage/block.h" @@ -30,6 +31,7 @@ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 #define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -44,6 +46,12 @@ typedef struct xl_smgr_unlink ForkNumber forkNum; } xl_smgr_unlink; +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -60,6 +68,8 @@ typedef struct xl_smgr_truncate extern void log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum); extern void log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileLocator rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index b379c76e27..0e4e290392 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -222,6 +222,9 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); + extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index ccd182531d..e59fb7892e 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -20,10 +20,10 @@ extern void ResetUnloggedRelations(int op); -extern void ResetUnloggedRelationIgnore(RelFileLocator rloc); extern bool parse_filename_for_nontemp_relation(const char *name, int *relnumchars, ForkNumber *fork); +extern void ResetUnloggedRelationIgnore(RelFileLocator rloc); #define UNLOGGED_RELATION_CLEANUP 0x0001 #define UNLOGGED_RELATION_INIT 0x0002 diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index b9255e5e25..2c34434555 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -3941,6 +3941,7 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create xl_smgr_truncate xl_smgr_unlink -- 2.39.3 ----Next_Part(Mon_Sep__4_17_37_48_2023_852)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v29 2/2] In-place table persistence change @ 2023-09-04 08:23 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2023-09-04 08:23 UTC (permalink / raw) Previously, the command caused a large amount of file I/O due to heap rewrites, even though ALTER TABLE SET UNLOGGED does not require any data rewrites. This patch eliminates the need for rewrites. Additionally, ALTER TABLE SET LOGGED is updated to emit XLOG_FPI records instead of numerous HEAP_INSERTs when wal_level > minimal, reducing resource consumption. --- src/backend/access/rmgrdesc/smgrdesc.c | 12 + src/backend/catalog/storage.c | 338 ++++++++++++++++++++++++- src/backend/commands/tablecmds.c | 268 +++++++++++++++++--- src/backend/storage/buffer/bufmgr.c | 84 ++++++ src/bin/pg_rewind/parsexlog.c | 6 + src/include/catalog/storage_xlog.h | 10 + src/include/storage/bufmgr.h | 3 + src/include/storage/reinit.h | 2 +- src/tools/pgindent/typedefs.list | 1 + 9 files changed, 683 insertions(+), 41 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index bd841b96e8..620e02bc26 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,15 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +64,9 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 1778801bbd..e7c917c50f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -71,11 +71,13 @@ typedef struct PendingRelDelete } PendingRelDelete; #define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_SET_PERSISTENCE (1 << 1) typedef struct PendingCleanup { RelFileLocator rlocator; /* relation that need a cleanup */ int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ ForkNumber unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ @@ -209,6 +211,208 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + ul_uncommitted_storage ul_storage; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If a pending-unlink exists for this relation's init-fork, it indicates + * the init-fork's existed before the current transaction; this function + * reverts the pending-unlink by removing the entry. See + * RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + /* write cancel log for preceding undo log entry */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create undo log entry, then the init fork */ + srel = smgropen(rlocator, InvalidBackendId); + + /* write undo log */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = true; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * For index relations, WAL-logging and file sync are handled by + * ambuildempty. In contrast, for heap relations, these tasks are performed + * directly. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file then revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * Search for a pending-unlink associated with the init-fork of the + * relation. Its presence indicates that the init-fork was created within + * the current transaction. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + ul_uncommitted_storage ul_storage; + + /* write cancel log for preceding undo log entry */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + /* + * If the init-fork was created in this transaction, remove the init-fork + * and cancel preceding undo log. Otherwise, register an at-commit + * pending-unlink for the existing init-fork. See RelationCreateInitFork. + */ + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + ul_uncommitted_storage ul_storage; + + /* + * Some AMs initialize init-fork via the buffer manager. To properly + * drop the init-fork, first drop all buffers for the init-fork, then + * unlink the init-fork and cancel preceding undo log. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + + /* cancel existing undo log */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -248,6 +452,25 @@ log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -800,7 +1023,14 @@ smgrDoPendingCleanups(bool isCommit) srel = smgropen(pending->rlocator, pending->backend); - Assert((pending->op & ~(PCOP_UNLINK_FORK)) == 0); + Assert((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } if (pending->op & PCOP_UNLINK_FORK) { @@ -1200,6 +1430,112 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 47c556669f..6c4cfbfa78 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5571,6 +5572,188 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: perform in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Use ATRewriteTable instead of this function under the following + * condition. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * Initially, gather all relations that require a persistence change. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods don't support in-place persistence + * changes. GiST uses page LSNs to figure out whether a block has been + * modified. However, UNLOGGED GiST indexes use fake LSNs, which are + * incompatible with the real LSNs used for LOGGED indexes. + * + * Potentially, if gistGetFakeLSN behaved similarly for both permanent + * and unlogged indexes, we could avoid index rebuilds by emitting + * extra WAL records while the index is unlogged. + * + * Compare relam against a positive list to ensure the hard way is + * taken for unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * If this relation becomes WAL-logged, immediately sync all files + * except the init-fork to establish the initial state on storage. The + * buffers should have already been flushed out by + * RelationCreate(Drop)InitFork called just above. The init-fork should + * already be synchronized as required. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * If wal_level >= replica, switching to LOGGED necessitates WAL-logging + * the relation content for later recovery. This is not emitted when + * wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5701,48 +5884,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 3bd82dbfca..04ab6ec8a7 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3708,6 +3708,90 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages to disk (or kernel disk buffers) when + * switching to PERMANENT, ensuring the kernel has an up-to-date view of + * the relation. + * + * The caller must be holding AccessExclusiveLock on the target relation + * to ensure no other backend is busy dirtying more blocks. + * + * XXX currently it sequentially searches the buffer pool; consider + * implementing more efficient search methods. This routine isn't used in + * performance-critical code paths, so it's not worth additional overhead + * to make it go faster; see also DropRelationBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* The init fork is being dropped, drop buffers for it. */ + if (BufTagGetForkNum(&bufHdr->tag) == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel, IOOBJECT_RELATION, IOCONTEXT_NORMAL); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(BufTagGetForkNum(&bufHdr->tag) != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 87b4659e27..db12f4f397 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -418,6 +418,12 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 5122f5b61d..eaa162f0c7 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -14,6 +14,7 @@ #ifndef STORAGE_XLOG_H #define STORAGE_XLOG_H +#include "access/simpleundolog.h" #include "access/xlogreader.h" #include "lib/stringinfo.h" #include "storage/block.h" @@ -30,6 +31,7 @@ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 #define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -44,6 +46,12 @@ typedef struct xl_smgr_unlink ForkNumber forkNum; } xl_smgr_unlink; +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -60,6 +68,8 @@ typedef struct xl_smgr_truncate extern void log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum); extern void log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileLocator rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index b379c76e27..0e4e290392 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -222,6 +222,9 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); + extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index ccd182531d..e59fb7892e 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -20,10 +20,10 @@ extern void ResetUnloggedRelations(int op); -extern void ResetUnloggedRelationIgnore(RelFileLocator rloc); extern bool parse_filename_for_nontemp_relation(const char *name, int *relnumchars, ForkNumber *fork); +extern void ResetUnloggedRelationIgnore(RelFileLocator rloc); #define UNLOGGED_RELATION_CLEANUP 0x0001 #define UNLOGGED_RELATION_INIT 0x0002 diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index b9255e5e25..2c34434555 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -3941,6 +3941,7 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create xl_smgr_truncate xl_smgr_unlink -- 2.39.3 ----Next_Part(Mon_Sep__4_17_37_48_2023_852)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v29 2/2] In-place table persistence change @ 2023-09-04 08:23 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2023-09-04 08:23 UTC (permalink / raw) Previously, the command caused a large amount of file I/O due to heap rewrites, even though ALTER TABLE SET UNLOGGED does not require any data rewrites. This patch eliminates the need for rewrites. Additionally, ALTER TABLE SET LOGGED is updated to emit XLOG_FPI records instead of numerous HEAP_INSERTs when wal_level > minimal, reducing resource consumption. --- src/backend/access/rmgrdesc/smgrdesc.c | 12 + src/backend/catalog/storage.c | 338 ++++++++++++++++++++++++- src/backend/commands/tablecmds.c | 268 +++++++++++++++++--- src/backend/storage/buffer/bufmgr.c | 84 ++++++ src/bin/pg_rewind/parsexlog.c | 6 + src/include/catalog/storage_xlog.h | 10 + src/include/storage/bufmgr.h | 3 + src/include/storage/reinit.h | 2 +- src/tools/pgindent/typedefs.list | 1 + 9 files changed, 683 insertions(+), 41 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index bd841b96e8..620e02bc26 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,15 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +64,9 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 1778801bbd..e7c917c50f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -71,11 +71,13 @@ typedef struct PendingRelDelete } PendingRelDelete; #define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_SET_PERSISTENCE (1 << 1) typedef struct PendingCleanup { RelFileLocator rlocator; /* relation that need a cleanup */ int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ ForkNumber unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ @@ -209,6 +211,208 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + ul_uncommitted_storage ul_storage; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If a pending-unlink exists for this relation's init-fork, it indicates + * the init-fork's existed before the current transaction; this function + * reverts the pending-unlink by removing the entry. See + * RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + /* write cancel log for preceding undo log entry */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create undo log entry, then the init fork */ + srel = smgropen(rlocator, InvalidBackendId); + + /* write undo log */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = true; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * For index relations, WAL-logging and file sync are handled by + * ambuildempty. In contrast, for heap relations, these tasks are performed + * directly. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file then revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * Search for a pending-unlink associated with the init-fork of the + * relation. Its presence indicates that the init-fork was created within + * the current transaction. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + ul_uncommitted_storage ul_storage; + + /* write cancel log for preceding undo log entry */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + /* + * If the init-fork was created in this transaction, remove the init-fork + * and cancel preceding undo log. Otherwise, register an at-commit + * pending-unlink for the existing init-fork. See RelationCreateInitFork. + */ + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + ul_uncommitted_storage ul_storage; + + /* + * Some AMs initialize init-fork via the buffer manager. To properly + * drop the init-fork, first drop all buffers for the init-fork, then + * unlink the init-fork and cancel preceding undo log. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + + /* cancel existing undo log */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -248,6 +452,25 @@ log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -800,7 +1023,14 @@ smgrDoPendingCleanups(bool isCommit) srel = smgropen(pending->rlocator, pending->backend); - Assert((pending->op & ~(PCOP_UNLINK_FORK)) == 0); + Assert((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } if (pending->op & PCOP_UNLINK_FORK) { @@ -1200,6 +1430,112 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 47c556669f..6c4cfbfa78 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5571,6 +5572,188 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: perform in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Use ATRewriteTable instead of this function under the following + * condition. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * Initially, gather all relations that require a persistence change. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods don't support in-place persistence + * changes. GiST uses page LSNs to figure out whether a block has been + * modified. However, UNLOGGED GiST indexes use fake LSNs, which are + * incompatible with the real LSNs used for LOGGED indexes. + * + * Potentially, if gistGetFakeLSN behaved similarly for both permanent + * and unlogged indexes, we could avoid index rebuilds by emitting + * extra WAL records while the index is unlogged. + * + * Compare relam against a positive list to ensure the hard way is + * taken for unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * If this relation becomes WAL-logged, immediately sync all files + * except the init-fork to establish the initial state on storage. The + * buffers should have already been flushed out by + * RelationCreate(Drop)InitFork called just above. The init-fork should + * already be synchronized as required. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * If wal_level >= replica, switching to LOGGED necessitates WAL-logging + * the relation content for later recovery. This is not emitted when + * wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5701,48 +5884,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 3bd82dbfca..04ab6ec8a7 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3708,6 +3708,90 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages to disk (or kernel disk buffers) when + * switching to PERMANENT, ensuring the kernel has an up-to-date view of + * the relation. + * + * The caller must be holding AccessExclusiveLock on the target relation + * to ensure no other backend is busy dirtying more blocks. + * + * XXX currently it sequentially searches the buffer pool; consider + * implementing more efficient search methods. This routine isn't used in + * performance-critical code paths, so it's not worth additional overhead + * to make it go faster; see also DropRelationBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* The init fork is being dropped, drop buffers for it. */ + if (BufTagGetForkNum(&bufHdr->tag) == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel, IOOBJECT_RELATION, IOCONTEXT_NORMAL); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(BufTagGetForkNum(&bufHdr->tag) != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 87b4659e27..db12f4f397 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -418,6 +418,12 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 5122f5b61d..eaa162f0c7 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -14,6 +14,7 @@ #ifndef STORAGE_XLOG_H #define STORAGE_XLOG_H +#include "access/simpleundolog.h" #include "access/xlogreader.h" #include "lib/stringinfo.h" #include "storage/block.h" @@ -30,6 +31,7 @@ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 #define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -44,6 +46,12 @@ typedef struct xl_smgr_unlink ForkNumber forkNum; } xl_smgr_unlink; +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -60,6 +68,8 @@ typedef struct xl_smgr_truncate extern void log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum); extern void log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileLocator rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index b379c76e27..0e4e290392 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -222,6 +222,9 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); + extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index ccd182531d..e59fb7892e 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -20,10 +20,10 @@ extern void ResetUnloggedRelations(int op); -extern void ResetUnloggedRelationIgnore(RelFileLocator rloc); extern bool parse_filename_for_nontemp_relation(const char *name, int *relnumchars, ForkNumber *fork); +extern void ResetUnloggedRelationIgnore(RelFileLocator rloc); #define UNLOGGED_RELATION_CLEANUP 0x0001 #define UNLOGGED_RELATION_INIT 0x0002 diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index b9255e5e25..2c34434555 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -3941,6 +3941,7 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create xl_smgr_truncate xl_smgr_unlink -- 2.39.3 ----Next_Part(Mon_Sep__4_17_37_48_2023_852)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v29 2/2] In-place table persistence change @ 2023-09-04 08:23 Kyotaro Horiguchi <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Kyotaro Horiguchi @ 2023-09-04 08:23 UTC (permalink / raw) Previously, the command caused a large amount of file I/O due to heap rewrites, even though ALTER TABLE SET UNLOGGED does not require any data rewrites. This patch eliminates the need for rewrites. Additionally, ALTER TABLE SET LOGGED is updated to emit XLOG_FPI records instead of numerous HEAP_INSERTs when wal_level > minimal, reducing resource consumption. --- src/backend/access/rmgrdesc/smgrdesc.c | 12 + src/backend/catalog/storage.c | 338 ++++++++++++++++++++++++- src/backend/commands/tablecmds.c | 268 +++++++++++++++++--- src/backend/storage/buffer/bufmgr.c | 84 ++++++ src/bin/pg_rewind/parsexlog.c | 6 + src/include/catalog/storage_xlog.h | 10 + src/include/storage/bufmgr.h | 3 + src/include/storage/reinit.h | 2 +- src/tools/pgindent/typedefs.list | 1 + 9 files changed, 683 insertions(+), 41 deletions(-) diff --git a/src/backend/access/rmgrdesc/smgrdesc.c b/src/backend/access/rmgrdesc/smgrdesc.c index bd841b96e8..620e02bc26 100644 --- a/src/backend/access/rmgrdesc/smgrdesc.c +++ b/src/backend/access/rmgrdesc/smgrdesc.c @@ -40,6 +40,15 @@ smgr_desc(StringInfo buf, XLogReaderState *record) xlrec->blkno, xlrec->flags); pfree(path); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = (xl_smgr_bufpersistence *) rec; + char *path = relpathperm(xlrec->rlocator, MAIN_FORKNUM); + + appendStringInfoString(buf, path); + appendStringInfo(buf, " persistence %d", xlrec->persistence); + pfree(path); + } } const char * @@ -55,6 +64,9 @@ smgr_identify(uint8 info) case XLOG_SMGR_TRUNCATE: id = "TRUNCATE"; break; + case XLOG_SMGR_BUFPERSISTENCE: + id = "BUFPERSISTENCE"; + break; } return id; diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index 1778801bbd..e7c917c50f 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -71,11 +71,13 @@ typedef struct PendingRelDelete } PendingRelDelete; #define PCOP_UNLINK_FORK (1 << 0) +#define PCOP_SET_PERSISTENCE (1 << 1) typedef struct PendingCleanup { RelFileLocator rlocator; /* relation that need a cleanup */ int op; /* operation mask */ + bool bufpersistence; /* buffer persistence to set */ ForkNumber unlink_forknum; /* forknum to unlink */ BackendId backend; /* InvalidBackendId if not a temp rel */ bool atCommit; /* T=delete at commit; F=delete at abort */ @@ -209,6 +211,208 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence, return srel; } +/* + * RelationCreateInitFork + * Create physical storage for the init fork of a relation. + * + * Create the init fork for the relation. + * + * This function is transactional. The creation is WAL-logged, and if the + * transaction aborts later on, the init fork will be removed. + */ +void +RelationCreateInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + SMgrRelation srel; + ul_uncommitted_storage ul_storage; + bool create = true; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), false, false); + + /* + * If a pending-unlink exists for this relation's init-fork, it indicates + * the init-fork's existed before the current transaction; this function + * reverts the pending-unlink by removing the entry. See + * RelationDropInitFork. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + /* write cancel log for preceding undo log entry */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + /* prev does not change */ + + create = false; + } + else + prev = pending; + } + + if (!create) + return; + + /* create undo log entry, then the init fork */ + srel = smgropen(rlocator, InvalidBackendId); + + /* write undo log */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = true; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + /* We don't have existing init fork, create it. */ + smgrcreate(srel, INIT_FORKNUM, false); + + /* + * For index relations, WAL-logging and file sync are handled by + * ambuildempty. In contrast, for heap relations, these tasks are performed + * directly. + */ + if (rel->rd_rel->relkind == RELKIND_INDEX) + rel->rd_indam->ambuildempty(rel); + else + { + log_smgrcreate(&rlocator, INIT_FORKNUM); + smgrimmedsync(srel, INIT_FORKNUM); + } + + /* drop the init fork, mark file then revert persistence at abort */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK | PCOP_SET_PERSISTENCE; + pending->unlink_forknum = INIT_FORKNUM; + pending->bufpersistence = true; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + +/* + * RelationDropInitFork + * Delete physical storage for the init fork of a relation. + */ +void +RelationDropInitFork(Relation rel) +{ + RelFileLocator rlocator = rel->rd_locator; + PendingCleanup *pending; + PendingCleanup *prev; + PendingCleanup *next; + bool inxact_created = false; + + /* switch buffer persistence */ + SetRelationBuffersPersistence(RelationGetSmgr(rel), true, false); + + /* + * Search for a pending-unlink associated with the init-fork of the + * relation. Its presence indicates that the init-fork was created within + * the current transaction. + */ + prev = NULL; + for (pending = pendingCleanups; pending != NULL; pending = next) + { + next = pending->next; + + if (RelFileLocatorEquals(rlocator, pending->rlocator) && + pending->unlink_forknum == INIT_FORKNUM) + { + ul_uncommitted_storage ul_storage; + + /* write cancel log for preceding undo log entry */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + + /* unlink list entry */ + if (prev) + prev->next = next; + else + pendingCleanups = next; + + pfree(pending); + + /* prev does not change */ + + inxact_created = true; + } + else + prev = pending; + } + + /* + * If the init-fork was created in this transaction, remove the init-fork + * and cancel preceding undo log. Otherwise, register an at-commit + * pending-unlink for the existing init-fork. See RelationCreateInitFork. + */ + if (inxact_created) + { + SMgrRelation srel = smgropen(rlocator, InvalidBackendId); + ForkNumber forknum = INIT_FORKNUM; + BlockNumber firstblock = 0; + ul_uncommitted_storage ul_storage; + + /* + * Some AMs initialize init-fork via the buffer manager. To properly + * drop the init-fork, first drop all buffers for the init-fork, then + * unlink the init-fork and cancel preceding undo log. + */ + DropRelationBuffers(srel, &forknum, 1, &firstblock); + + /* cancel existing undo log */ + ul_storage.rlocator = rlocator; + ul_storage.forknum = INIT_FORKNUM; + ul_storage.remove = false; + SimpleUndoLogWrite(RM_SMGR_ID, ULOG_SMGR_UNCOMMITED_STORAGE, + GetCurrentTransactionId(), + &ul_storage, sizeof(ul_storage)); + log_smgrunlink(&rlocator, INIT_FORKNUM); + smgrunlink(srel, INIT_FORKNUM, false); + return; + } + + /* register drop of this init fork file at commit */ + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = rlocator; + pending->op = PCOP_UNLINK_FORK; + pending->unlink_forknum = INIT_FORKNUM; + pending->backend = InvalidBackendId; + pending->atCommit = true; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; +} + /* * Perform XLogInsert of an XLOG_SMGR_CREATE record to WAL. */ @@ -248,6 +452,25 @@ log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum) XLogInsert(RM_SMGR_ID, XLOG_SMGR_UNLINK | XLR_SPECIAL_REL_UPDATE); } +/* + * Perform XLogInsert of an XLOG_SMGR_BUFPERSISTENCE record to WAL. + */ +void +log_smgrbufpersistence(const RelFileLocator rlocator, bool persistence) +{ + xl_smgr_bufpersistence xlrec; + + /* + * Make an XLOG entry reporting the change of buffer persistence. + */ + xlrec.rlocator = rlocator; + xlrec.persistence = persistence; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + XLogInsert(RM_SMGR_ID, XLOG_SMGR_BUFPERSISTENCE | XLR_SPECIAL_REL_UPDATE); +} + /* * RelationDropStorage * Schedule unlinking of physical storage at transaction commit. @@ -800,7 +1023,14 @@ smgrDoPendingCleanups(bool isCommit) srel = smgropen(pending->rlocator, pending->backend); - Assert((pending->op & ~(PCOP_UNLINK_FORK)) == 0); + Assert((pending->op & + ~(PCOP_UNLINK_FORK | PCOP_SET_PERSISTENCE)) == 0); + + if (pending->op & PCOP_SET_PERSISTENCE) + { + SetRelationBuffersPersistence(srel, pending->bufpersistence, + InRecovery); + } if (pending->op & PCOP_UNLINK_FORK) { @@ -1200,6 +1430,112 @@ smgr_redo(XLogReaderState *record) FreeFakeRelcacheEntry(rel); } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } + else if (info == XLOG_SMGR_BUFPERSISTENCE) + { + xl_smgr_bufpersistence *xlrec = + (xl_smgr_bufpersistence *) XLogRecGetData(record); + SMgrRelation reln; + PendingCleanup *pending; + PendingCleanup *prev = NULL; + + reln = smgropen(xlrec->rlocator, InvalidBackendId); + SetRelationBuffersPersistence(reln, xlrec->persistence, true); + + /* + * Delete any pending action for persistence change, if present. There + * should be at most one entry for this action. + */ + for (pending = pendingCleanups; pending != NULL; + pending = pending->next) + { + if (RelFileLocatorEquals(xlrec->rlocator, pending->rlocator) && + (pending->op & PCOP_SET_PERSISTENCE) != 0) + { + Assert(pending->bufpersistence == xlrec->persistence); + + if (prev) + prev->next = pending->next; + else + pendingCleanups = pending->next; + + pfree(pending); + break; + } + + prev = pending; + } + + /* + * During abort, revert any changes to buffer persistence made made in + * this transaction. + */ + if (!pending) + { + pending = (PendingCleanup *) + MemoryContextAlloc(TopMemoryContext, sizeof(PendingCleanup)); + pending->rlocator = xlrec->rlocator; + pending->op = PCOP_SET_PERSISTENCE; + pending->bufpersistence = !xlrec->persistence; + pending->backend = InvalidBackendId; + pending->atCommit = false; + pending->nestLevel = GetCurrentTransactionNestLevel(); + pending->next = pendingCleanups; + pendingCleanups = pending; + } + } else elog(PANIC, "smgr_redo: unknown op code %u", info); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 47c556669f..6c4cfbfa78 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -55,6 +55,7 @@ #include "commands/defrem.h" #include "commands/event_trigger.h" #include "commands/policy.h" +#include "commands/progress.h" #include "commands/sequence.h" #include "commands/tablecmds.h" #include "commands/tablespace.h" @@ -5571,6 +5572,188 @@ ATParseTransformCmd(List **wqueue, AlteredTableInfo *tab, Relation rel, return newcmd; } +/* + * RelationChangePersistence: perform in-place persistence change of a relation + */ +static void +RelationChangePersistence(AlteredTableInfo *tab, char persistence, + LOCKMODE lockmode) +{ + Relation rel; + Relation classRel; + HeapTuple tuple, + newtuple; + Datum new_val[Natts_pg_class]; + bool new_null[Natts_pg_class], + new_repl[Natts_pg_class]; + int i; + List *relids; + ListCell *lc_oid; + + Assert(tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE); + Assert(lockmode == AccessExclusiveLock); + + /* + * Use ATRewriteTable instead of this function under the following + * condition. + */ + Assert(tab->constraints == NULL && tab->partition_constraint == NULL && + tab->newvals == NULL && !tab->verify_new_notnull); + + rel = table_open(tab->relid, lockmode); + + Assert(rel->rd_rel->relpersistence != persistence); + + elog(DEBUG1, "perform in-place persistence change"); + + /* + * Initially, gather all relations that require a persistence change. + */ + + /* Collect OIDs of indexes and toast relations */ + relids = RelationGetIndexList(rel); + relids = lcons_oid(rel->rd_id, relids); + + /* Add toast relation if any */ + if (OidIsValid(rel->rd_rel->reltoastrelid)) + { + List *toastidx; + Relation toastrel = table_open(rel->rd_rel->reltoastrelid, lockmode); + + relids = lappend_oid(relids, rel->rd_rel->reltoastrelid); + toastidx = RelationGetIndexList(toastrel); + relids = list_concat(relids, toastidx); + pfree(toastidx); + table_close(toastrel, NoLock); + } + + table_close(rel, NoLock); + + /* Make changes in storage */ + classRel = table_open(RelationRelationId, RowExclusiveLock); + + foreach(lc_oid, relids) + { + Oid reloid = lfirst_oid(lc_oid); + Relation r = relation_open(reloid, lockmode); + + /* + * XXXX: Some access methods don't support in-place persistence + * changes. GiST uses page LSNs to figure out whether a block has been + * modified. However, UNLOGGED GiST indexes use fake LSNs, which are + * incompatible with the real LSNs used for LOGGED indexes. + * + * Potentially, if gistGetFakeLSN behaved similarly for both permanent + * and unlogged indexes, we could avoid index rebuilds by emitting + * extra WAL records while the index is unlogged. + * + * Compare relam against a positive list to ensure the hard way is + * taken for unknown AMs. + */ + if (r->rd_rel->relkind == RELKIND_INDEX && + /* GiST is excluded */ + r->rd_rel->relam != BTREE_AM_OID && + r->rd_rel->relam != HASH_AM_OID && + r->rd_rel->relam != GIN_AM_OID && + r->rd_rel->relam != SPGIST_AM_OID && + r->rd_rel->relam != BRIN_AM_OID) + { + int reindex_flags; + ReindexParams params = {0}; + + /* reindex doesn't allow concurrent use of the index */ + table_close(r, NoLock); + + reindex_flags = + REINDEX_REL_SUPPRESS_INDEX_USE | + REINDEX_REL_CHECK_CONSTRAINTS; + + /* Set the same persistence with the parent relation. */ + if (persistence == RELPERSISTENCE_UNLOGGED) + reindex_flags |= REINDEX_REL_FORCE_INDEXES_UNLOGGED; + else + reindex_flags |= REINDEX_REL_FORCE_INDEXES_PERMANENT; + + reindex_index(reloid, reindex_flags, persistence, ¶ms); + + continue; + } + + /* Create or drop init fork */ + if (persistence == RELPERSISTENCE_UNLOGGED) + RelationCreateInitFork(r); + else + RelationDropInitFork(r); + + /* + * If this relation becomes WAL-logged, immediately sync all files + * except the init-fork to establish the initial state on storage. The + * buffers should have already been flushed out by + * RelationCreate(Drop)InitFork called just above. The init-fork should + * already be synchronized as required. + */ + if (persistence == RELPERSISTENCE_PERMANENT) + { + for (i = 0; i < INIT_FORKNUM; i++) + { + if (smgrexists(RelationGetSmgr(r), i)) + smgrimmedsync(RelationGetSmgr(r), i); + } + } + + /* Update catalog */ + tuple = SearchSysCacheCopy1(RELOID, ObjectIdGetDatum(reloid)); + if (!HeapTupleIsValid(tuple)) + elog(ERROR, "cache lookup failed for relation %u", reloid); + + memset(new_val, 0, sizeof(new_val)); + memset(new_null, false, sizeof(new_null)); + memset(new_repl, false, sizeof(new_repl)); + + new_val[Anum_pg_class_relpersistence - 1] = CharGetDatum(persistence); + new_null[Anum_pg_class_relpersistence - 1] = false; + new_repl[Anum_pg_class_relpersistence - 1] = true; + + newtuple = heap_modify_tuple(tuple, RelationGetDescr(classRel), + new_val, new_null, new_repl); + + CatalogTupleUpdate(classRel, &newtuple->t_self, newtuple); + heap_freetuple(newtuple); + + /* + * If wal_level >= replica, switching to LOGGED necessitates WAL-logging + * the relation content for later recovery. This is not emitted when + * wal_level = minimal. + */ + if (persistence == RELPERSISTENCE_PERMANENT && XLogIsNeeded()) + { + ForkNumber fork; + xl_smgr_truncate xlrec; + + xlrec.blkno = 0; + xlrec.rlocator = r->rd_locator; + xlrec.flags = SMGR_TRUNCATE_ALL; + + XLogBeginInsert(); + XLogRegisterData((char *) &xlrec, sizeof(xlrec)); + + XLogInsert(RM_SMGR_ID, XLOG_SMGR_TRUNCATE | XLR_SPECIAL_REL_UPDATE); + + for (fork = 0; fork < INIT_FORKNUM; fork++) + { + if (smgrexists(RelationGetSmgr(r), fork)) + log_newpage_range(r, fork, 0, + smgrnblocks(RelationGetSmgr(r), fork), + false); + } + } + + table_close(r, NoLock); + } + + table_close(classRel, NoLock); +} + /* * ATRewriteTables: ALTER TABLE phase 3 */ @@ -5701,48 +5884,55 @@ ATRewriteTables(AlterTableStmt *parsetree, List **wqueue, LOCKMODE lockmode, tab->relid, tab->rewrite); - /* - * Create transient table that will receive the modified data. - * - * Ensure it is marked correctly as logged or unlogged. We have - * to do this here so that buffers for the new relfilenumber will - * have the right persistence set, and at the same time ensure - * that the original filenumbers's buffers will get read in with - * the correct setting (i.e. the original one). Otherwise a - * rollback after the rewrite would possibly result with buffers - * for the original filenumbers having the wrong persistence - * setting. - * - * NB: This relies on swap_relation_files() also swapping the - * persistence. That wouldn't work for pg_class, but that can't be - * unlogged anyway. - */ - OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, NewAccessMethod, - persistence, lockmode); + if (tab->rewrite == AT_REWRITE_ALTER_PERSISTENCE) + RelationChangePersistence(tab, persistence, lockmode); + else + { + /* + * Create transient table that will receive the modified data. + * + * Ensure it is marked correctly as logged or unlogged. We + * have to do this here so that buffers for the new + * relfilenumber will have the right persistence set, and at + * the same time ensure that the original filenumbers's buffers + * will get read in with the correct setting (i.e. the original + * one). Otherwise a rollback after the rewrite would possibly + * result with buffers for the original filenumbers having the + * wrong persistence setting. + * + * NB: This relies on swap_relation_files() also swapping the + * persistence. That wouldn't work for pg_class, but that + * can't be unlogged anyway. + */ + OIDNewHeap = make_new_heap(tab->relid, NewTableSpace, + NewAccessMethod, + persistence, lockmode); - /* - * Copy the heap data into the new table with the desired - * modifications, and test the current data within the table - * against new constraints generated by ALTER TABLE commands. - */ - ATRewriteTable(tab, OIDNewHeap, lockmode); + /* + * Copy the heap data into the new table with the desired + * modifications, and test the current data within the table + * against new constraints generated by ALTER TABLE commands. + */ + ATRewriteTable(tab, OIDNewHeap, lockmode); - /* - * Swap the physical files of the old and new heaps, then rebuild - * indexes and discard the old heap. We can use RecentXmin for - * the table's new relfrozenxid because we rewrote all the tuples - * in ATRewriteTable, so no older Xid remains in the table. Also, - * we never try to swap toast tables by content, since we have no - * interest in letting this code work on system catalogs. - */ - finish_heap_swap(tab->relid, OIDNewHeap, - false, false, true, - !OidIsValid(tab->newTableSpace), - RecentXmin, - ReadNextMultiXactId(), - persistence); + /* + * Swap the physical files of the old and new heaps, then + * rebuild indexes and discard the old heap. We can use + * RecentXmin for the table's new relfrozenxid because we + * rewrote all the tuples in ATRewriteTable, so no older Xid + * remains in the table. Also, we never try to swap toast + * tables by content, since we have no interest in letting + * this code work on system catalogs. + */ + finish_heap_swap(tab->relid, OIDNewHeap, + false, false, true, + !OidIsValid(tab->newTableSpace), + RecentXmin, + ReadNextMultiXactId(), + persistence); - InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + InvokeObjectPostAlterHook(RelationRelationId, tab->relid, 0); + } } else if (tab->rewrite > 0 && tab->relkind == RELKIND_SEQUENCE) { diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 3bd82dbfca..04ab6ec8a7 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -3708,6 +3708,90 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum, } } +/* --------------------------------------------------------------------- + * SetRelationBuffersPersistence + * + * This function changes the persistence of all buffer pages of a relation + * then writes all dirty pages to disk (or kernel disk buffers) when + * switching to PERMANENT, ensuring the kernel has an up-to-date view of + * the relation. + * + * The caller must be holding AccessExclusiveLock on the target relation + * to ensure no other backend is busy dirtying more blocks. + * + * XXX currently it sequentially searches the buffer pool; consider + * implementing more efficient search methods. This routine isn't used in + * performance-critical code paths, so it's not worth additional overhead + * to make it go faster; see also DropRelationBuffers. + * -------------------------------------------------------------------- + */ +void +SetRelationBuffersPersistence(SMgrRelation srel, bool permanent, bool isRedo) +{ + int i; + RelFileLocatorBackend rlocator = srel->smgr_rlocator; + + Assert(!RelFileLocatorBackendIsTemp(rlocator)); + + if (!isRedo) + log_smgrbufpersistence(srel->smgr_rlocator.locator, permanent); + + ResourceOwnerEnlargeBuffers(CurrentResourceOwner); + + for (i = 0; i < NBuffers; i++) + { + BufferDesc *bufHdr = GetBufferDescriptor(i); + uint32 buf_state; + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + continue; + + ReservePrivateRefCountEntry(); + + buf_state = LockBufHdr(bufHdr); + + if (!RelFileLocatorEquals(BufTagGetRelFileLocator(&bufHdr->tag), + rlocator.locator)) + { + UnlockBufHdr(bufHdr, buf_state); + continue; + } + + if (permanent) + { + /* The init fork is being dropped, drop buffers for it. */ + if (BufTagGetForkNum(&bufHdr->tag) == INIT_FORKNUM) + { + InvalidateBuffer(bufHdr); + continue; + } + + buf_state |= BM_PERMANENT; + pg_atomic_write_u32(&bufHdr->state, buf_state); + + /* flush this buffer when switching to PERMANENT */ + if ((buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY)) + { + PinBuffer_Locked(bufHdr); + LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), + LW_SHARED); + FlushBuffer(bufHdr, srel, IOOBJECT_RELATION, IOCONTEXT_NORMAL); + LWLockRelease(BufferDescriptorGetContentLock(bufHdr)); + UnpinBuffer(bufHdr); + } + else + UnlockBufHdr(bufHdr, buf_state); + } + else + { + /* There shouldn't be an init fork */ + Assert(BufTagGetForkNum(&bufHdr->tag) != INIT_FORKNUM); + UnlockBufHdr(bufHdr, buf_state); + } + } +} + /* --------------------------------------------------------------------- * DropRelationsAllBuffers * diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c index 87b4659e27..db12f4f397 100644 --- a/src/bin/pg_rewind/parsexlog.c +++ b/src/bin/pg_rewind/parsexlog.c @@ -418,6 +418,12 @@ extractPageInfo(XLogReaderState *record) * source system. */ } + else if (rmid == RM_SMGR_ID && rminfo == XLOG_SMGR_BUFPERSISTENCE) + { + /* + * We can safely ignore these. These don't make any on-disk changes. + */ + } else if (rmid == RM_XACT_ID && ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || diff --git a/src/include/catalog/storage_xlog.h b/src/include/catalog/storage_xlog.h index 5122f5b61d..eaa162f0c7 100644 --- a/src/include/catalog/storage_xlog.h +++ b/src/include/catalog/storage_xlog.h @@ -14,6 +14,7 @@ #ifndef STORAGE_XLOG_H #define STORAGE_XLOG_H +#include "access/simpleundolog.h" #include "access/xlogreader.h" #include "lib/stringinfo.h" #include "storage/block.h" @@ -30,6 +31,7 @@ #define XLOG_SMGR_CREATE 0x10 #define XLOG_SMGR_TRUNCATE 0x20 #define XLOG_SMGR_UNLINK 0x30 +#define XLOG_SMGR_BUFPERSISTENCE 0x40 typedef struct xl_smgr_create { @@ -44,6 +46,12 @@ typedef struct xl_smgr_unlink ForkNumber forkNum; } xl_smgr_unlink; +typedef struct xl_smgr_bufpersistence +{ + RelFileLocator rlocator; + bool persistence; +} xl_smgr_bufpersistence; + /* flags for xl_smgr_truncate */ #define SMGR_TRUNCATE_HEAP 0x0001 #define SMGR_TRUNCATE_VM 0x0002 @@ -60,6 +68,8 @@ typedef struct xl_smgr_truncate extern void log_smgrcreate(const RelFileLocator *rlocator, ForkNumber forkNum); extern void log_smgrunlink(const RelFileLocator *rlocator, ForkNumber forkNum); +extern void log_smgrbufpersistence(const RelFileLocator rlocator, + bool persistence); extern void smgr_redo(XLogReaderState *record); extern void smgr_desc(StringInfo buf, XLogReaderState *record); diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h index b379c76e27..0e4e290392 100644 --- a/src/include/storage/bufmgr.h +++ b/src/include/storage/bufmgr.h @@ -222,6 +222,9 @@ extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln, int nforks, BlockNumber *firstDelBlock); extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln, int nlocators); +extern void SetRelationBuffersPersistence(struct SMgrRelationData *srel, + bool permanent, bool isRedo); + extern void DropDatabaseBuffers(Oid dbid); #define RelationGetNumberOfBlocks(reln) \ diff --git a/src/include/storage/reinit.h b/src/include/storage/reinit.h index ccd182531d..e59fb7892e 100644 --- a/src/include/storage/reinit.h +++ b/src/include/storage/reinit.h @@ -20,10 +20,10 @@ extern void ResetUnloggedRelations(int op); -extern void ResetUnloggedRelationIgnore(RelFileLocator rloc); extern bool parse_filename_for_nontemp_relation(const char *name, int *relnumchars, ForkNumber *fork); +extern void ResetUnloggedRelationIgnore(RelFileLocator rloc); #define UNLOGGED_RELATION_CLEANUP 0x0001 #define UNLOGGED_RELATION_INIT 0x0002 diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index b9255e5e25..2c34434555 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -3941,6 +3941,7 @@ xl_replorigin_set xl_restore_point xl_running_xacts xl_seq_rec +xl_smgr_bufpersistence xl_smgr_create xl_smgr_truncate xl_smgr_unlink -- 2.39.3 ----Next_Part(Mon_Sep__4_17_37_48_2023_852)---- ^ permalink raw reply [nested|flat] 110+ messages in thread
* Re: ON ERROR in json_query and the like @ 2024-06-12 13:53 Markus Winand <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Markus Winand @ 2024-06-12 13:53 UTC (permalink / raw) To: David G. Johnston <[email protected]>; +Cc: pgsql-hackers > On 12.06.2024, at 15:31, David G. Johnston <[email protected]> wrote: > > On Tuesday, May 28, 2024, Markus Winand <[email protected]> wrote: > > 2. EMPTY [ARRAY|OBJECT] ON ERROR implies ERROR ON EMPTY > > 17beta1=# SELECT JSON_QUERY('[]', '$[*]' EMPTY ARRAY ON ERROR) a; > a > ---- > [] > (1 row) > > As NULL ON EMPTY is implied, it should give the same result as > explicitly adding NULL ON EMPTY: > > 17beta1=# SELECT JSON_QUERY('[]', '$[*]' NULL ON EMPTY EMPTY ARRAY ON ERROR) a; > a > --- > > (1 row) > > Interestingly, Oracle DB gives the same (wrong) results. Db2 (LUW) > on the other hand returns NULL for both queries. > > I don’t think that PostgreSQL should follow Oracle DB's suit here > but again, in case this is intentional it should be made explicit > in the docs. > > The docs here don’t seem to cover the on empty clause at all nor fully cover all options. > > Where do you find the claim that the one implies the other? Is it a typo that your examples says “implies null on empty” but the subject line says “implies error on empty”? I see the confusion caused — sorry. The headline was meant to describe the observed behaviour in 17beta1, while the content refers to how the standard defines it. > Without those clauses a result is either empty or an error - they are mutually exclusive (ignoring matches). I would not expect one clause to imply or affect the behavior of the other. There is no chaining. The original result is transformed to the new result specified by the clause. Agreed, that’s why I found the 17beta1 behaviour surprising. > I’d need to figure out whether the example you show is actually producing empty or error; but it seems correct if the result is empty. As I understand the standard, an empty result is not an error. > The first query ignores the error clause - the empty array row seems to be the representation of empty here; the second one matches the empty clause and outputs null instead of the empty array. But the first should behave the same, as the standard implies NULL ON EMPTY if there is no explicit ON EMPTY clause. Oracle DB behaving differently here makes me wonder if there is something in the standard that I haven’t noticed yet... -markus ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v1] aio: io_uring: Fix danger of completion getting reused before being read @ 2026-01-15 15:17 Andres Freund <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Andres Freund @ 2026-01-15 15:17 UTC (permalink / raw) We called io_uring_cqe_seen(..., cqe) before reading cqe->res. That allows the completion to be reused, which in turn could lead to cqe->res being overwritten. The window for that is very narrow and the likelihood of it happening is very low, as we should never actually utilize all CQEs, but the consequences would be bad. This bug was reported to me privately. Backpatch-through: 18 --- src/backend/storage/aio/method_io_uring.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/src/backend/storage/aio/method_io_uring.c b/src/backend/storage/aio/method_io_uring.c index af58c6118ac..f2527ae61dd 100644 --- a/src/backend/storage/aio/method_io_uring.c +++ b/src/backend/storage/aio/method_io_uring.c @@ -559,13 +559,14 @@ pgaio_uring_drain_locked(PgAioUringContext *context) for (int i = 0; i < ncqes; i++) { struct io_uring_cqe *cqe = cqes[i]; - PgAioHandle *ioh; + PgAioHandle *ioh = io_uring_cqe_get_data(cqe); + int result = cqe->res; - ioh = io_uring_cqe_get_data(cqe); errcallback.arg = ioh; + io_uring_cqe_seen(&context->io_uring_ring, cqe); - pgaio_io_process_completion(ioh, cqe->res); + pgaio_io_process_completion(ioh, result); errcallback.arg = NULL; } -- 2.48.1.76.g4e746b1a31.dirty --ppazo27qa47camgo-- ^ permalink raw reply [nested|flat] 110+ messages in thread
* [PATCH v1] aio: io_uring: Fix danger of completion getting reused before being read @ 2026-01-15 15:17 Andres Freund <[email protected]> 0 siblings, 0 replies; 110+ messages in thread From: Andres Freund @ 2026-01-15 15:17 UTC (permalink / raw) We called io_uring_cqe_seen(..., cqe) before reading cqe->res. That allows the completion to be reused, which in turn could lead to cqe->res being overwritten. The window for that is very narrow and the likelihood of it happening is very low, as we should never actually utilize all CQEs, but the consequences would be bad. This bug was reported to me privately. Backpatch-through: 18 --- src/backend/storage/aio/method_io_uring.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/src/backend/storage/aio/method_io_uring.c b/src/backend/storage/aio/method_io_uring.c index af58c6118ac..f2527ae61dd 100644 --- a/src/backend/storage/aio/method_io_uring.c +++ b/src/backend/storage/aio/method_io_uring.c @@ -559,13 +559,14 @@ pgaio_uring_drain_locked(PgAioUringContext *context) for (int i = 0; i < ncqes; i++) { struct io_uring_cqe *cqe = cqes[i]; - PgAioHandle *ioh; + PgAioHandle *ioh = io_uring_cqe_get_data(cqe); + int result = cqe->res; - ioh = io_uring_cqe_get_data(cqe); errcallback.arg = ioh; + io_uring_cqe_seen(&context->io_uring_ring, cqe); - pgaio_io_process_completion(ioh, cqe->res); + pgaio_io_process_completion(ioh, result); errcallback.arg = NULL; } -- 2.48.1.76.g4e746b1a31.dirty --ppazo27qa47camgo-- ^ permalink raw reply [nested|flat] 110+ messages in thread
end of thread, other threads:[~2026-01-15 15:17 UTC | newest] Thread overview: 110+ messages (download: mbox mbox.gz follow: Atom feed) -- links below jump to the message on this page -- 2018-11-27 05:42 [PATCH 5/5] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH 5/5] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH 5/5] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH v22 5/5] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH 6/6] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH 5/5] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH 5/5] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH 5/5] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH 5/5] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH 5/5] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH v23 5/5] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2018-11-27 05:42 [PATCH 6/6] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2020-03-13 08:00 [PATCH v27 7/7] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2020-03-13 08:00 [PATCH v28 7/7] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2020-03-13 08:00 [PATCH v25 8/8] Remove the GUC stats_temp_directory Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v9 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v19 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v14 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v22 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v10 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v12 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v15 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v16 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v2 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v6 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v13 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v21 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v17 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v18 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v20 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v11 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v3 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v8 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v4 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2020-11-11 12:51 [PATCH v5 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2022-07-19 04:23 [PATCH v24 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2022-07-19 04:23 [PATCH v23 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2022-07-19 04:23 [PATCH v25 1/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2023-03-15 06:42 [PATCH v27 2/3] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2023-04-25 06:49 [PATCH v28 2/4] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2023-04-25 06:49 [PATCH v29 2/3] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2023-04-25 06:49 [PATCH v29 2/3] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2023-09-04 08:23 [PATCH v29 2/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2023-09-04 08:23 [PATCH v29 2/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2023-09-04 08:23 [PATCH v29 2/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2023-09-04 08:23 [PATCH v29 2/2] In-place table persistence change Kyotaro Horiguchi <[email protected]> 2024-06-12 13:53 Re: ON ERROR in json_query and the like Markus Winand <[email protected]> 2026-01-15 15:17 [PATCH v1] aio: io_uring: Fix danger of completion getting reused before being read Andres Freund <[email protected]> 2026-01-15 15:17 [PATCH v1] aio: io_uring: Fix danger of completion getting reused before being read Andres Freund <[email protected]>
This inbox is served by agora; see mirroring instructions for how to clone and mirror all data and code used for this inbox