public inbox for [email protected]
help / color / mirror / Atom feed[PATCH 01/10] Allow alternate compression methods for wal_compression
3+ messages / 3 participants
[nested] [flat]
* [PATCH 01/10] Allow alternate compression methods for wal_compression
@ 2021-02-27 04:03 Andrey Borodin <[email protected]>
0 siblings, 0 replies; 3+ messages in thread
From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw)
TODO: bump XLOG_PAGE_MAGIC
---
doc/src/sgml/config.sgml | 17 +++++
src/backend/Makefile | 2 +-
src/backend/access/transam/xlog.c | 10 +++
src/backend/access/transam/xloginsert.c | 52 +++++++++++++--
src/backend/access/transam/xlogreader.c | 63 ++++++++++++++++++-
src/backend/utils/misc/guc.c | 11 ++++
src/backend/utils/misc/postgresql.conf.sample | 1 +
src/include/access/xlog.h | 1 +
src/include/access/xlog_internal.h | 8 +++
src/include/access/xlogreader.h | 1 +
src/include/access/xlogrecord.h | 9 +--
11 files changed, 163 insertions(+), 12 deletions(-)
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index a218d78bef..7fb2a84626 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -3072,6 +3072,23 @@ include_dir 'conf.d'
</listitem>
</varlistentry>
+ <varlistentry id="guc-wal-compression-method" xreflabel="wal_compression_method">
+ <term><varname>wal_compressionion_method</varname> (<type>enum</type>)
+ <indexterm>
+ <primary><varname>wal_compression_method</varname> configuration parameter</primary>
+ </indexterm>
+ </term>
+ <listitem>
+ <para>
+ This parameter selects the compression method used to compress WAL when
+ <varname>wal_compression</varname> is enabled.
+ The supported methods are pglz and zlib.
+ The default value is <literal>pglz</literal>.
+ Only superusers can change this setting.
+ </para>
+ </listitem>
+ </varlistentry>
+
<varlistentry id="guc-wal-init-zero" xreflabel="wal_init_zero">
<term><varname>wal_init_zero</varname> (<type>boolean</type>)
<indexterm>
diff --git a/src/backend/Makefile b/src/backend/Makefile
index 0da848b1fd..3af216ddfc 100644
--- a/src/backend/Makefile
+++ b/src/backend/Makefile
@@ -48,7 +48,7 @@ OBJS = \
LIBS := $(filter-out -lpgport -lpgcommon, $(LIBS)) $(LDAP_LIBS_BE) $(ICU_LIBS)
# The backend doesn't need everything that's in LIBS, however
-LIBS := $(filter-out -lz -lreadline -ledit -ltermcap -lncurses -lcurses, $(LIBS))
+LIBS := $(filter-out -lreadline -ledit -ltermcap -lncurses -lcurses, $(LIBS))
ifeq ($(with_systemd),yes)
LIBS += -lsystemd
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..04192b7add 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -99,6 +99,7 @@ bool EnableHotStandby = false;
bool fullPageWrites = true;
bool wal_log_hints = false;
bool wal_compression = false;
+int wal_compression_method = WAL_COMPRESSION_PGLZ;
char *wal_consistency_checking_string = NULL;
bool *wal_consistency_checking = NULL;
bool wal_init_zero = true;
@@ -180,6 +181,15 @@ const struct config_enum_entry recovery_target_action_options[] = {
{NULL, 0, false}
};
+/* Note that due to conditional compilation, offsets within the array are not static */
+const struct config_enum_entry wal_compression_options[] = {
+ {"pglz", WAL_COMPRESSION_PGLZ, false},
+#ifdef HAVE_LIBZ
+ {"zlib", WAL_COMPRESSION_ZLIB, false},
+#endif
+ {NULL, 0, false}
+};
+
/*
* Statistics for current checkpoint are collected in this global struct.
* Because only the checkpointer or a stand-alone backend can perform
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..34e1227381 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -33,6 +33,10 @@
#include "storage/proc.h"
#include "utils/memutils.h"
+#ifdef HAVE_LIBZ
+#include <zlib.h>
+#endif
+
/* Buffer size required to store a compressed version of backup block image */
#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ)
@@ -113,7 +117,8 @@ static XLogRecData *XLogRecordAssemble(RmgrId rmid, uint8 info,
XLogRecPtr RedoRecPtr, bool doPageWrites,
XLogRecPtr *fpw_lsn, int *num_fpi);
static bool XLogCompressBackupBlock(char *page, uint16 hole_offset,
- uint16 hole_length, char *dest, uint16 *dlen);
+ uint16 hole_length, char *dest,
+ uint16 *dlen, WalCompression compression);
/*
* Begin constructing a WAL record. This must be called before the
@@ -630,11 +635,12 @@ XLogRecordAssemble(RmgrId rmid, uint8 info,
*/
if (wal_compression)
{
+ bimg.compression_method = wal_compression_method;
is_compressed =
XLogCompressBackupBlock(page, bimg.hole_offset,
cbimg.hole_length,
regbuf->compressed_page,
- &compressed_len);
+ &compressed_len, bimg.compression_method);
}
/*
@@ -827,7 +833,7 @@ XLogRecordAssemble(RmgrId rmid, uint8 info,
*/
static bool
XLogCompressBackupBlock(char *page, uint16 hole_offset, uint16 hole_length,
- char *dest, uint16 *dlen)
+ char *dest, uint16 *dlen, WalCompression compression)
{
int32 orig_len = BLCKSZ - hole_length;
int32 len;
@@ -853,12 +859,48 @@ XLogCompressBackupBlock(char *page, uint16 hole_offset, uint16 hole_length,
else
source = page;
+ switch (compression)
+ {
+ case WAL_COMPRESSION_PGLZ:
+ len = pglz_compress(source, orig_len, dest, PGLZ_strategy_default);
+ break;
+
+#ifdef HAVE_LIBZ
+ case WAL_COMPRESSION_ZLIB:
+ {
+ unsigned long len_l = PGLZ_MAX_BLCKSZ;
+ int ret;
+ ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1);
+ if (ret != Z_OK)
+ {
+ // XXX: using an interface other than compress() would allow giving a better error message
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("failed compressing zlib (%d)", ret)));
+ len_l = -1;
+ }
+ len = len_l;
+ break;
+ }
+#endif
+
+ default:
+ /*
+ * It should be impossible to get here for unsupported algorithms,
+ * which cannot be assigned if they're not enabled at compile time.
+ */
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("unknown compression method requested: %d(%s)",
+ compression, wal_compression_name(compression))));
+
+ }
+
/*
- * We recheck the actual size even if pglz_compress() reports success and
+ * We recheck the actual size even if compression reports success and
* see if the number of bytes saved by compression is larger than the
* length of extra data needed for the compressed version of block image.
*/
- len = pglz_compress(source, orig_len, dest, PGLZ_strategy_default);
if (len >= 0 &&
len + extra_bytes < orig_len)
{
diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c
index 42738eb940..afca22a26c 100644
--- a/src/backend/access/transam/xlogreader.c
+++ b/src/backend/access/transam/xlogreader.c
@@ -33,6 +33,10 @@
#include "utils/memutils.h"
#endif
+#ifdef HAVE_LIBZ
+#include <zlib.h>
+#endif
+
static void report_invalid_record(XLogReaderState *state, const char *fmt,...)
pg_attribute_printf(2, 3);
static bool allocate_recordbuf(XLogReaderState *state, uint32 reclength);
@@ -1286,6 +1290,7 @@ DecodeXLogRecord(XLogReaderState *state, XLogRecord *record, char **errormsg)
{
COPY_HEADER_FIELD(&blk->bimg_len, sizeof(uint16));
COPY_HEADER_FIELD(&blk->hole_offset, sizeof(uint16));
+ COPY_HEADER_FIELD(&blk->compression_method, sizeof(uint8));
COPY_HEADER_FIELD(&blk->bimg_info, sizeof(uint8));
blk->apply_image = ((blk->bimg_info & BKPIMAGE_APPLY) != 0);
@@ -1535,6 +1540,29 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len)
}
}
+/*
+ * Return a statically allocated string associated with the given compression
+ * method. This is similar to the guc, but isn't subject to conditional
+ * compilation.
+ */
+const char *
+wal_compression_name(WalCompression compression)
+{
+ /*
+ * This could index into the guc array, except that it's compiled
+ * conditionally and unsupported methods are elided.
+ */
+ switch (compression)
+ {
+ case WAL_COMPRESSION_PGLZ:
+ return "pglz";
+ case WAL_COMPRESSION_ZLIB:
+ return "zlib";
+ default:
+ return "???";
+ }
+}
+
/*
* Restore a full-page image from a backup block attached to an XLOG record.
*
@@ -1558,8 +1586,39 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page)
if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED)
{
/* If a backup block image is compressed, decompress it */
- if (pglz_decompress(ptr, bkpb->bimg_len, tmp.data,
- BLCKSZ - bkpb->hole_length, true) < 0)
+ int32 decomp_result = -1;
+ switch (bkpb->compression_method)
+ {
+ case WAL_COMPRESSION_PGLZ:
+ decomp_result = pglz_decompress(ptr, bkpb->bimg_len, tmp.data,
+ BLCKSZ - bkpb->hole_length, true);
+ break;
+
+#ifdef HAVE_LIBZ
+ case WAL_COMPRESSION_ZLIB:
+ {
+ unsigned long decomp_result_l;
+ decomp_result_l = BLCKSZ - bkpb->hole_length;
+ if (uncompress((Bytef*)tmp.data, &decomp_result_l,
+ (Bytef*)ptr, bkpb->bimg_len) == Z_OK)
+ decomp_result = decomp_result_l;
+ else
+ decomp_result = -1;
+ break;
+ }
+#endif
+
+ default:
+ report_invalid_record(record, "image at %X/%X is compressed with unsupported codec, block %d (%d/%s)",
+ (uint32) (record->ReadRecPtr >> 32),
+ (uint32) record->ReadRecPtr,
+ block_id,
+ bkpb->compression_method,
+ wal_compression_name(bkpb->compression_method));
+ return false;
+ }
+
+ if (decomp_result < 0)
{
report_invalid_record(record, "invalid compressed image at %X/%X, block %d",
LSN_FORMAT_ARGS(record->ReadRecPtr),
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index 855076b1fd..8084027465 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -508,6 +508,7 @@ extern const struct config_enum_entry archive_mode_options[];
extern const struct config_enum_entry recovery_target_action_options[];
extern const struct config_enum_entry sync_method_options[];
extern const struct config_enum_entry dynamic_shared_memory_options[];
+extern const struct config_enum_entry wal_compression_options[];
/*
* GUC option variables that are exported from this module
@@ -4721,6 +4722,16 @@ static struct config_enum ConfigureNamesEnum[] =
NULL, NULL, NULL
},
+ {
+ {"wal_compression_method", PGC_SIGHUP, WAL_SETTINGS,
+ gettext_noop("Set the method used to compress full page images in the WAL."),
+ NULL
+ },
+ &wal_compression_method,
+ WAL_COMPRESSION_PGLZ, wal_compression_options,
+ NULL, NULL, NULL
+ },
+
{
{"dynamic_shared_memory_type", PGC_POSTMASTER, RESOURCES_MEM,
gettext_noop("Selects the dynamic shared memory implementation used."),
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index f46c2dd7a8..ef69a94492 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -213,6 +213,7 @@
# open_sync
#full_page_writes = on # recover from partial page writes
#wal_compression = off # enable compression of full-page writes
+#wal_compression_method = pglz # pglz, zlib
#wal_log_hints = off # also do full page writes of non-critical updates
# (change requires restart)
#wal_init_zero = on # zero-fill new WAL files
diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 6d384d3ce6..fa2e5c611f 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -117,6 +117,7 @@ extern bool EnableHotStandby;
extern bool fullPageWrites;
extern bool wal_log_hints;
extern bool wal_compression;
+extern int wal_compression_method;
extern bool wal_init_zero;
extern bool wal_recycle;
extern bool *wal_consistency_checking;
diff --git a/src/include/access/xlog_internal.h b/src/include/access/xlog_internal.h
index b23e286406..d653839b97 100644
--- a/src/include/access/xlog_internal.h
+++ b/src/include/access/xlog_internal.h
@@ -324,4 +324,12 @@ extern bool InArchiveRecovery;
extern bool StandbyMode;
extern char *recoveryRestoreCommand;
+typedef enum WalCompression
+{
+ WAL_COMPRESSION_PGLZ,
+ WAL_COMPRESSION_ZLIB,
+} WalCompression;
+
+extern const char *wal_compression_name(WalCompression compression);
+
#endif /* XLOG_INTERNAL_H */
diff --git a/src/include/access/xlogreader.h b/src/include/access/xlogreader.h
index 21d200d3df..3d19c315d7 100644
--- a/src/include/access/xlogreader.h
+++ b/src/include/access/xlogreader.h
@@ -133,6 +133,7 @@ typedef struct
bool apply_image; /* has image that should be restored */
char *bkp_image;
uint16 hole_offset;
+ uint8 compression_method;
uint16 hole_length;
uint16 bimg_len;
uint8 bimg_info;
diff --git a/src/include/access/xlogrecord.h b/src/include/access/xlogrecord.h
index 80c92a2498..0d4c212f15 100644
--- a/src/include/access/xlogrecord.h
+++ b/src/include/access/xlogrecord.h
@@ -114,7 +114,7 @@ typedef struct XLogRecordBlockHeader
* present is (BLCKSZ - <length of "hole" bytes>).
*
* Additionally, when wal_compression is enabled, we will try to compress full
- * page images using the PGLZ compression algorithm, after removing the "hole".
+ * page images, after removing the "hole".
* This can reduce the WAL volume, but at some extra cost of CPU spent
* on the compression during WAL logging. In this case, since the "hole"
* length cannot be calculated by subtracting the number of page image bytes
@@ -129,9 +129,10 @@ typedef struct XLogRecordBlockHeader
*/
typedef struct XLogRecordBlockImageHeader
{
- uint16 length; /* number of page image bytes */
- uint16 hole_offset; /* number of bytes before "hole" */
- uint8 bimg_info; /* flag bits, see below */
+ uint16 length; /* number of page image bytes */
+ uint16 hole_offset; /* number of bytes before "hole" */
+ uint8 compression_method; /* compression method used for image */
+ uint8 bimg_info; /* flag bits, see below */
/*
* If BKPIMAGE_HAS_HOLE and BKPIMAGE_IS_COMPRESSED, an
--
2.17.0
--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
filename="0002-Run-011_crash_recovery.pl-with-wal_level-minimal.patch"
^ permalink raw reply [nested|flat] 3+ messages in thread
* Re: bug: copy progress reporting of backends which run multiple COPYs
@ 2024-05-08 14:07 Robert Haas <[email protected]>
2024-05-08 23:57 ` Re: bug: copy progress reporting of backends which run multiple COPYs Michael Paquier <[email protected]>
0 siblings, 1 reply; 3+ messages in thread
From: Robert Haas @ 2024-05-08 14:07 UTC (permalink / raw)
To: Michael Paquier <[email protected]>; +Cc: Justin Pryzby <[email protected]>; Matthias van de Meent <[email protected]>; Tomas Vondra <[email protected]>; Josef Šimánek <[email protected]>; [email protected]
On Tue, May 7, 2024 at 9:12 PM Michael Paquier <[email protected]> wrote:
> FWIW, I'm rather annoyed by the fact that we rely on the ParseState to
> decide if reporting should happen or not. file_fdw tells, even if
> that's accidental, that status reporting can be useful if working on a
> single table. So, shutting down the whole reporting if a caller if
> BeginCopyFrom() does not give a ParseState is too heavy-handed, IMO.
I think you're hoping for too much. The progress reporting
infrastructure is fundamentally designed around the idea that there
can only be one progress-reporting operation in progress at a time.
For COPY, that is, I believe, true, but for file_fdw, it's false. If
we want to do any kind of progress reporting from within plannable
queries, we need some totally different and much more complex
infrastructure that can report progress for, probably, each plan node
individually. I think it's likely a mistake to try to shoehorn cases
like this into the infrastructure
that we have today. It will just encourage people to try to use the
current infrastructure in ways that are less and less like what it was
actually designed to do; whereas what we should be doing if we want
this kind of functionality, at least IMHO, is building infrastructure
that's actually fit for purpose.
--
Robert Haas
EDB: http://www.enterprisedb.com
^ permalink raw reply [nested|flat] 3+ messages in thread
* Re: bug: copy progress reporting of backends which run multiple COPYs
2024-05-08 14:07 Re: bug: copy progress reporting of backends which run multiple COPYs Robert Haas <[email protected]>
@ 2024-05-08 23:57 ` Michael Paquier <[email protected]>
0 siblings, 0 replies; 3+ messages in thread
From: Michael Paquier @ 2024-05-08 23:57 UTC (permalink / raw)
To: Robert Haas <[email protected]>; +Cc: Justin Pryzby <[email protected]>; Matthias van de Meent <[email protected]>; Tomas Vondra <[email protected]>; Josef Šimánek <[email protected]>; [email protected]
On Wed, May 08, 2024 at 10:07:15AM -0400, Robert Haas wrote:
> I think you're hoping for too much. The progress reporting
> infrastructure is fundamentally designed around the idea that there
> can only be one progress-reporting operation in progress at a time.
> For COPY, that is, I believe, true, but for file_fdw, it's false. If
> we want to do any kind of progress reporting from within plannable
> queries, we need some totally different and much more complex
> infrastructure that can report progress for, probably, each plan node
> individually. I think it's likely a mistake to try to shoehorn cases
> like this into the infrastructure
> that we have today. It will just encourage people to try to use the
> current infrastructure in ways that are less and less like what it was
> actually designed to do; whereas what we should be doing if we want
> this kind of functionality, at least IMHO, is building infrastructure
> that's actually fit for purpose.
Hmm. OK. I have been looking around for cases out there where
BeginCopyFrom() could be called with a pstate where the reporting
could matter, and could not find anything worth worrying about. It
still makes me a bit uneasy to not have a separate argument in the
function to control that. Now, if you, Justin and Matthias agree with
this approach, I won't stand in the way either.
--
Michael
Attachments:
[application/pgp-signature] signature.asc (833B, ../../[email protected]/2-signature.asc)
download
^ permalink raw reply [nested|flat] 3+ messages in thread
end of thread, other threads:[~2024-05-08 23:57 UTC | newest]
Thread overview: 3+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]>
2024-05-08 14:07 Re: bug: copy progress reporting of backends which run multiple COPYs Robert Haas <[email protected]>
2024-05-08 23:57 ` Re: bug: copy progress reporting of backends which run multiple COPYs Michael Paquier <[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