public inbox for [email protected]help / color / mirror / Atom feed
[PATCH 01/12] Allow alternate compression methods for wal_compression 105+ messages / 2 participants [nested] [flat]
* [PATCH 01/12] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 deletions(-) diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index ee4925d6d9..bed89d055d 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 6f8810e149..5b657ec724 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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 3b36a31a47..47d30d6cda 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -517,6 +517,7 @@ 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 @@ -4760,6 +4761,16 @@ 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 86425965d0..530007f55a 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 77187c12be..2f3886431a 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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --45Z9DzgjV8m4Oswq 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH v9 1/9] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- doc/src/sgml/config.sgml | 9 +- doc/src/sgml/installation.sgml | 4 +- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 14 ++- src/backend/access/transam/xloginsert.c | 65 ++++++++++-- src/backend/access/transam/xlogreader.c | 99 ++++++++++++++++--- src/backend/utils/misc/guc.c | 21 ++-- src/backend/utils/misc/postgresql.conf.sample | 2 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlog.h | 2 +- src/include/access/xlog_internal.h | 10 ++ src/include/access/xlogrecord.h | 15 ++- 12 files changed, 207 insertions(+), 49 deletions(-) diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index aa3e178240..1df56d8034 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -3127,23 +3127,26 @@ include_dir 'conf.d' </varlistentry> <varlistentry id="guc-wal-compression" xreflabel="wal_compression"> - <term><varname>wal_compression</varname> (<type>boolean</type>) + <term><varname>wal_compression</varname> (<type>enum</type>) <indexterm> <primary><varname>wal_compression</varname> configuration parameter</primary> </indexterm> </term> <listitem> <para> - When this parameter is <literal>on</literal>, the <productname>PostgreSQL</productname> + This parameter enables compression of WAL using the specified + compression method. + When enabled, the <productname>PostgreSQL</productname> server compresses full page images written to WAL when <xref linkend="guc-full-page-writes"/> is on or during a base backup. A compressed page image will be decompressed during WAL replay. + The supported methods are pglz and zlib. The default value is <literal>off</literal>. Only superusers can change this setting. </para> <para> - Turning this parameter on can reduce the WAL volume without + Enabling compression can reduce the WAL volume without increasing the risk of unrecoverable data corruption, but at the cost of some extra CPU spent on the compression during WAL logging and on the decompression during WAL replay. diff --git a/doc/src/sgml/installation.sgml b/doc/src/sgml/installation.sgml index 3c0aa118c7..073d5089f7 100644 --- a/doc/src/sgml/installation.sgml +++ b/doc/src/sgml/installation.sgml @@ -147,7 +147,7 @@ su - postgres specify the <option>--without-zlib</option> option to <filename>configure</filename>. Using this option disables support for compressed archives in <application>pg_dump</application> and - <application>pg_restore</application>. + <application>pg_restore</application>, and compressed WAL. </para> </listitem> </itemizedlist> @@ -1236,7 +1236,7 @@ build-postgresql: Prevents use of the <application>Zlib</application> library. This disables support for compressed archives in <application>pg_dump</application> - and <application>pg_restore</application>. + and <application>pg_restore</application> and compressed WAL. </para> </listitem> </varlistentry> 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 17eeff0720..1ccc51575a 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -98,7 +98,7 @@ char *XLogArchiveCommand = NULL; bool EnableHotStandby = false; bool fullPageWrites = true; bool wal_log_hints = false; -bool wal_compression = false; +int wal_compression = WAL_COMPRESSION_NONE; char *wal_consistency_checking_string = NULL; bool *wal_consistency_checking = NULL; bool wal_init_zero = true; @@ -10470,7 +10470,17 @@ xlog_block_info(StringInfo buf, XLogReaderState *record) rnode.spcNode, rnode.dbNode, rnode.relNode, blk); if (XLogRecHasBlockImage(record, block_id)) - appendStringInfoString(buf, " FPW"); + { + int compression = + BKPIMAGE_IS_COMPRESSED(record->blocks[block_id].bimg_info) ? + BKPIMAGE_COMPRESSION(record->blocks[block_id].bimg_info) : -1; + if (compression == -1) + appendStringInfoString(buf, " FPW"); + else + appendStringInfo(buf, " FPW, compression method %d/%s", + compression, wal_compression_name(compression)); + } + } } diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c index 32b4cc84e7..4f81f19c49 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -628,13 +639,14 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, /* * Try to compress a block image if wal_compression is enabled */ - if (wal_compression) + if (wal_compression != WAL_COMPRESSION_NONE) { is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression); } /* @@ -665,8 +677,13 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, if (is_compressed) { + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression); /* Range check */ + Assert(wal_compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.length = compressed_len; - bimg.bimg_info |= BKPIMAGE_IS_COMPRESSED; + bimg.bimg_info |= + wal_compression << BKPIMAGE_COMPRESS_OFFSET_BITS; rdt_datas_last->data = regbuf->compressed_page; rdt_datas_last->len = compressed_len; @@ -827,7 +844,7 @@ 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 +870,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..3a922caaeb 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -26,6 +26,7 @@ #include "catalog/pg_control.h" #include "common/pg_lzcompress.h" #include "replication/origin.h" +#include "utils/guc.h" #ifndef FRONTEND #include "miscadmin.h" @@ -33,6 +34,10 @@ #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); @@ -50,6 +55,29 @@ static void WALOpenSegmentInit(WALOpenSegment *seg, WALSegmentContext *segcxt, /* size of the buffer allocated for error message. */ #define MAX_ERRORMSG_LEN 1000 +/* + * Accept the likely variants for none and pglz, for compatibility with old + * server versions where wal_compression was a boolean. + */ +const struct config_enum_entry wal_compression_options[] = { + {"off", WAL_COMPRESSION_NONE, false}, + {"none", WAL_COMPRESSION_NONE, false}, + {"false", WAL_COMPRESSION_NONE, true}, + {"no", WAL_COMPRESSION_NONE, true}, + {"0", WAL_COMPRESSION_NONE, true}, + {"pglz", WAL_COMPRESSION_PGLZ, false}, + {"true", WAL_COMPRESSION_PGLZ, true}, + {"yes", WAL_COMPRESSION_PGLZ, true}, + {"on", WAL_COMPRESSION_PGLZ, true}, + {"1", WAL_COMPRESSION_PGLZ, true}, + +#ifdef HAVE_LIBZ + {"zlib", WAL_COMPRESSION_ZLIB, false}, +#endif + + {NULL, 0, false} +}; + /* * Construct a string in state->errormsg_buf explaining what's wrong with * the current record being read. @@ -1290,7 +1318,7 @@ DecodeXLogRecord(XLogReaderState *state, XLogRecord *record, char **errormsg) blk->apply_image = ((blk->bimg_info & BKPIMAGE_APPLY) != 0); - if (blk->bimg_info & BKPIMAGE_IS_COMPRESSED) + if (BKPIMAGE_IS_COMPRESSED(blk->bimg_info)) { if (blk->bimg_info & BKPIMAGE_HAS_HOLE) COPY_HEADER_FIELD(&blk->hole_length, sizeof(uint16)); @@ -1335,29 +1363,28 @@ DecodeXLogRecord(XLogReaderState *state, XLogRecord *record, char **errormsg) } /* - * cross-check that bimg_len < BLCKSZ if the IS_COMPRESSED - * flag is set. + * cross-check that bimg_len < BLCKSZ if it's compressed */ - if ((blk->bimg_info & BKPIMAGE_IS_COMPRESSED) && + if (BKPIMAGE_IS_COMPRESSED(blk->bimg_info) && blk->bimg_len == BLCKSZ) { report_invalid_record(state, - "BKPIMAGE_IS_COMPRESSED set, but block image length %u at %X/%X", + "BKPIMAGE_IS_COMPRESSED, but block image length %u at %X/%X", (unsigned int) blk->bimg_len, LSN_FORMAT_ARGS(state->ReadRecPtr)); goto err; } /* - * cross-check that bimg_len = BLCKSZ if neither HAS_HOLE nor - * IS_COMPRESSED flag is set. + * cross-check that bimg_len = BLCKSZ if neither HAS_HOLE is + * set nor IS_COMPRESSED(). */ if (!(blk->bimg_info & BKPIMAGE_HAS_HOLE) && - !(blk->bimg_info & BKPIMAGE_IS_COMPRESSED) && + !BKPIMAGE_IS_COMPRESSED(blk->bimg_info) && blk->bimg_len != BLCKSZ) { report_invalid_record(state, - "neither BKPIMAGE_HAS_HOLE nor BKPIMAGE_IS_COMPRESSED set, but block image length is %u at %X/%X", + "neither BKPIMAGE_HAS_HOLE nor BKPIMAGE_IS_COMPRESSED, but block image length is %u at %X/%X", (unsigned int) blk->data_len, LSN_FORMAT_ARGS(state->ReadRecPtr)); goto err; @@ -1535,6 +1562,22 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* + * Return a statically allocated string associated with the given compression + * method. + */ +const char * +wal_compression_name(WalCompression compression) +{ + for (int i=0; wal_compression_options[i].name != NULL; ++i) + { + if (wal_compression_options[i].val == compression) + return wal_compression_options[i].name; + } + + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1555,11 +1598,43 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) bkpb = &record->blocks[block_id]; ptr = bkpb->bkp_image; - if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) + if (BKPIMAGE_IS_COMPRESSED(bkpb->bimg_info)) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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 68b62d523d..ce1149bed5 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -548,6 +548,7 @@ 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 @@ -1304,16 +1305,6 @@ static struct config_bool ConfigureNamesBool[] = NULL, NULL, NULL }, - { - {"wal_compression", PGC_SUSET, WAL_SETTINGS, - gettext_noop("Compresses full-page writes written in WAL file."), - NULL - }, - &wal_compression, - false, - NULL, NULL, NULL - }, - { {"wal_init_zero", PGC_SUSET, WAL_SETTINGS, gettext_noop("Writes zeroes to new WAL files before first use."), @@ -4825,6 +4816,16 @@ static struct config_enum ConfigureNamesEnum[] = NULL, NULL, NULL }, + { + {"wal_compression", PGC_SUSET, WAL_SETTINGS, + gettext_noop("Set the method used to compress full page images in the WAL."), + NULL + }, + &wal_compression, + WAL_COMPRESSION_NONE, 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 ddbb6dc2be..61786c6e07 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -218,7 +218,7 @@ #full_page_writes = on # recover from partial page writes #wal_log_hints = off # also do full page writes of non-critical updates # (change requires restart) -#wal_compression = off # enable compression of full-page writes +#wal_compression = off # enable compression of full-page writes: off, pglz, zlib #wal_init_zero = on # zero-fill new WAL files #wal_recycle = on # recycle WAL files #wal_buffers = -1 # min 32kB, -1 sets based on shared_buffers diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index f8b8afe4a7..1cd71ac2f7 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -537,18 +537,21 @@ XLogDumpDisplayRecord(XLogDumpConfig *config, XLogReaderState *record) blk); if (XLogRecHasBlockImage(record, block_id)) { - if (record->blocks[block_id].bimg_info & - BKPIMAGE_IS_COMPRESSED) + if (BKPIMAGE_IS_COMPRESSED(record->blocks[block_id].bimg_info)) { + int compression = BKPIMAGE_COMPRESSION( + record->blocks[block_id].bimg_info); + printf(" (FPW%s); hole: offset: %u, length: %u, " - "compression saved: %u", + "compression method %d/%s, saved: %u", XLogRecBlockImageApply(record, block_id) ? "" : " for WAL verification", record->blocks[block_id].hole_offset, record->blocks[block_id].hole_length, + compression, wal_compression_name(compression), BLCKSZ - - record->blocks[block_id].hole_length - - record->blocks[block_id].bimg_len); + record->blocks[block_id].hole_length - + record->blocks[block_id].bimg_len); } else { diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h index 77187c12be..e8b2c53784 100644 --- a/src/include/access/xlog.h +++ b/src/include/access/xlog.h @@ -116,7 +116,7 @@ extern char *XLogArchiveCommand; extern bool EnableHotStandby; extern bool fullPageWrites; extern bool wal_log_hints; -extern bool wal_compression; +extern int wal_compression; 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 26a743b6b6..8b740af66d 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,14 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +/* These are the compression IDs written into bimg_info */ +typedef enum WalCompression +{ + WAL_COMPRESSION_NONE, + 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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..2a60c0fb92 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 @@ -144,9 +144,18 @@ typedef struct XLogRecordBlockImageHeader /* Information stored in bimg_info */ #define BKPIMAGE_HAS_HOLE 0x01 /* page image has "hole" */ -#define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ -#define BKPIMAGE_APPLY 0x04 /* page image should be restored during +#define BKPIMAGE_APPLY 0x02 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x04 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x08 /* 0=none, 1=pglz, 2=zlib */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 2 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) +#define BKPIMAGE_IS_COMPRESSED(info) (BKPIMAGE_COMPRESSION(info) != 0) /* * Extra header information used when page image has "hole" and -- 2.17.0 --neYutvxvOLaeuPCA Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="v9-0002-Run-011_crash_recovery.pl-with-wal_level-minimal.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH v7 1/9] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 deletions(-) diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 7e32b0686c..70effa6345 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -3137,6 +3137,23 @@ 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 8d163f190f..b7f8e12aea 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 32b4cc84e7..5ab07621d6 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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 ee731044b6..99932582ba 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -548,6 +548,7 @@ 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 @@ -4825,6 +4826,16 @@ static struct config_enum ConfigureNamesEnum[] = NULL, NULL, NULL }, + { + {"wal_compression_method", PGC_SUSET, 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 6e36e4c2ef..baed4d9228 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -219,6 +219,7 @@ #wal_log_hints = off # also do full page writes of non-critical updates # (change requires restart) #wal_compression = off # enable compression of full-page writes +#wal_compression_method = pglz # pglz, zlib #wal_init_zero = on # zero-fill new WAL files #wal_recycle = on # recycle WAL files #wal_buffers = -1 # min 32kB, -1 sets based on shared_buffers diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h index 77187c12be..2f3886431a 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 26a743b6b6..ceca0f7189 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --JYK4vJDZwFMowpUq Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="v7-0002-Run-011_crash_recovery.pl-with-wal_level-minimal.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH v8 1/9] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- doc/src/sgml/config.sgml | 9 +- doc/src/sgml/installation.sgml | 4 +- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 2 +- src/backend/access/transam/xloginsert.c | 65 +++++++++++-- src/backend/access/transam/xlogreader.c | 97 ++++++++++++++++--- src/backend/utils/misc/guc.c | 21 ++-- src/backend/utils/misc/postgresql.conf.sample | 2 +- src/bin/pg_waldump/pg_waldump.c | 13 ++- src/include/access/xlog.h | 2 +- src/include/access/xlog_internal.h | 10 ++ src/include/access/xlogrecord.h | 15 ++- 12 files changed, 194 insertions(+), 48 deletions(-) diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 7e32b0686c..218a98cfc6 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -3113,23 +3113,26 @@ include_dir 'conf.d' </varlistentry> <varlistentry id="guc-wal-compression" xreflabel="wal_compression"> - <term><varname>wal_compression</varname> (<type>boolean</type>) + <term><varname>wal_compression</varname> (<type>enum</type>) <indexterm> <primary><varname>wal_compression</varname> configuration parameter</primary> </indexterm> </term> <listitem> <para> - When this parameter is <literal>on</literal>, the <productname>PostgreSQL</productname> + This parameter enables compression of WAL using the specified + compression method. + When enabled, the <productname>PostgreSQL</productname> server compresses full page images written to WAL when <xref linkend="guc-full-page-writes"/> is on or during a base backup. A compressed page image will be decompressed during WAL replay. + The supported methods are pglz and zlib. The default value is <literal>off</literal>. Only superusers can change this setting. </para> <para> - Turning this parameter on can reduce the WAL volume without + Enabling compression can reduce the WAL volume without increasing the risk of unrecoverable data corruption, but at the cost of some extra CPU spent on the compression during WAL logging and on the decompression during WAL replay. diff --git a/doc/src/sgml/installation.sgml b/doc/src/sgml/installation.sgml index 3c0aa118c7..073d5089f7 100644 --- a/doc/src/sgml/installation.sgml +++ b/doc/src/sgml/installation.sgml @@ -147,7 +147,7 @@ su - postgres specify the <option>--without-zlib</option> option to <filename>configure</filename>. Using this option disables support for compressed archives in <application>pg_dump</application> and - <application>pg_restore</application>. + <application>pg_restore</application>, and compressed WAL. </para> </listitem> </itemizedlist> @@ -1236,7 +1236,7 @@ build-postgresql: Prevents use of the <application>Zlib</application> library. This disables support for compressed archives in <application>pg_dump</application> - and <application>pg_restore</application>. + and <application>pg_restore</application> and compressed WAL. </para> </listitem> </varlistentry> 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 441a9124cd..64094e7175 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -98,7 +98,7 @@ char *XLogArchiveCommand = NULL; bool EnableHotStandby = false; bool fullPageWrites = true; bool wal_log_hints = false; -bool wal_compression = false; +int wal_compression = WAL_COMPRESSION_NONE; char *wal_consistency_checking_string = NULL; bool *wal_consistency_checking = NULL; bool wal_init_zero = true; diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c index 32b4cc84e7..4f81f19c49 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -628,13 +639,14 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, /* * Try to compress a block image if wal_compression is enabled */ - if (wal_compression) + if (wal_compression != WAL_COMPRESSION_NONE) { is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression); } /* @@ -665,8 +677,13 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, if (is_compressed) { + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression); /* Range check */ + Assert(wal_compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.length = compressed_len; - bimg.bimg_info |= BKPIMAGE_IS_COMPRESSED; + bimg.bimg_info |= + wal_compression << BKPIMAGE_COMPRESS_OFFSET_BITS; rdt_datas_last->data = regbuf->compressed_page; rdt_datas_last->len = compressed_len; @@ -827,7 +844,7 @@ 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 +870,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..822b2612cd 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -26,6 +26,7 @@ #include "catalog/pg_control.h" #include "common/pg_lzcompress.h" #include "replication/origin.h" +#include "utils/guc.h" #ifndef FRONTEND #include "miscadmin.h" @@ -33,6 +34,10 @@ #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); @@ -50,6 +55,27 @@ static void WALOpenSegmentInit(WALOpenSegment *seg, WALSegmentContext *segcxt, /* size of the buffer allocated for error message. */ #define MAX_ERRORMSG_LEN 1000 +/* + * Accept the likely variants for none and pglz, for compatibility with old + * server versions where wal_compression was a boolean. + */ +const struct config_enum_entry wal_compression_options[] = { + {"off", WAL_COMPRESSION_NONE, false}, + {"none", WAL_COMPRESSION_NONE, false}, + {"false", WAL_COMPRESSION_NONE, true}, + {"no", WAL_COMPRESSION_NONE, true}, + {"0", WAL_COMPRESSION_NONE, true}, + {"pglz", WAL_COMPRESSION_PGLZ, false}, + {"true", WAL_COMPRESSION_PGLZ, true}, + {"yes", WAL_COMPRESSION_PGLZ, true}, + {"on", WAL_COMPRESSION_PGLZ, true}, + {"1", WAL_COMPRESSION_PGLZ, true}, +#ifdef HAVE_LIBZ + {"zlib", WAL_COMPRESSION_ZLIB, false}, +#endif + {NULL, 0, false} +}; + /* * Construct a string in state->errormsg_buf explaining what's wrong with * the current record being read. @@ -1290,7 +1316,7 @@ DecodeXLogRecord(XLogReaderState *state, XLogRecord *record, char **errormsg) blk->apply_image = ((blk->bimg_info & BKPIMAGE_APPLY) != 0); - if (blk->bimg_info & BKPIMAGE_IS_COMPRESSED) + if (BKPIMAGE_IS_COMPRESSED(blk->bimg_info)) { if (blk->bimg_info & BKPIMAGE_HAS_HOLE) COPY_HEADER_FIELD(&blk->hole_length, sizeof(uint16)); @@ -1335,29 +1361,28 @@ DecodeXLogRecord(XLogReaderState *state, XLogRecord *record, char **errormsg) } /* - * cross-check that bimg_len < BLCKSZ if the IS_COMPRESSED - * flag is set. + * cross-check that bimg_len < BLCKSZ if it's compressed */ - if ((blk->bimg_info & BKPIMAGE_IS_COMPRESSED) && + if (BKPIMAGE_IS_COMPRESSED(blk->bimg_info) && blk->bimg_len == BLCKSZ) { report_invalid_record(state, - "BKPIMAGE_IS_COMPRESSED set, but block image length %u at %X/%X", + "BKPIMAGE_IS_COMPRESSED, but block image length %u at %X/%X", (unsigned int) blk->bimg_len, LSN_FORMAT_ARGS(state->ReadRecPtr)); goto err; } /* - * cross-check that bimg_len = BLCKSZ if neither HAS_HOLE nor - * IS_COMPRESSED flag is set. + * cross-check that bimg_len = BLCKSZ if neither HAS_HOLE is + * set nor IS_COMPRESSED(). */ if (!(blk->bimg_info & BKPIMAGE_HAS_HOLE) && - !(blk->bimg_info & BKPIMAGE_IS_COMPRESSED) && + !BKPIMAGE_IS_COMPRESSED(blk->bimg_info) && blk->bimg_len != BLCKSZ) { report_invalid_record(state, - "neither BKPIMAGE_HAS_HOLE nor BKPIMAGE_IS_COMPRESSED set, but block image length is %u at %X/%X", + "neither BKPIMAGE_HAS_HOLE nor BKPIMAGE_IS_COMPRESSED, but block image length is %u at %X/%X", (unsigned int) blk->data_len, LSN_FORMAT_ARGS(state->ReadRecPtr)); goto err; @@ -1535,6 +1560,22 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* + * Return a statically allocated string associated with the given compression + * method. + */ +const char * +wal_compression_name(WalCompression compression) +{ + for (int i=0; wal_compression_options[i].name != NULL; ++i) + { + if (wal_compression_options[i].val == compression) + return wal_compression_options[i].name; + } + + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1555,11 +1596,43 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) bkpb = &record->blocks[block_id]; ptr = bkpb->bkp_image; - if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) + if (BKPIMAGE_IS_COMPRESSED(bkpb->bimg_info)) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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 ee731044b6..8860deda2a 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -548,6 +548,7 @@ 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 @@ -1304,16 +1305,6 @@ static struct config_bool ConfigureNamesBool[] = NULL, NULL, NULL }, - { - {"wal_compression", PGC_SUSET, WAL_SETTINGS, - gettext_noop("Compresses full-page writes written in WAL file."), - NULL - }, - &wal_compression, - false, - NULL, NULL, NULL - }, - { {"wal_init_zero", PGC_SUSET, WAL_SETTINGS, gettext_noop("Writes zeroes to new WAL files before first use."), @@ -4825,6 +4816,16 @@ static struct config_enum ConfigureNamesEnum[] = NULL, NULL, NULL }, + { + {"wal_compression", PGC_SUSET, WAL_SETTINGS, + gettext_noop("Set the method used to compress full page images in the WAL."), + NULL + }, + &wal_compression, + WAL_COMPRESSION_NONE, 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 6e36e4c2ef..3991d35afd 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -218,7 +218,7 @@ #full_page_writes = on # recover from partial page writes #wal_log_hints = off # also do full page writes of non-critical updates # (change requires restart) -#wal_compression = off # enable compression of full-page writes +#wal_compression = off # enable compression of full-page writes: off, pglz, zlib #wal_init_zero = on # zero-fill new WAL files #wal_recycle = on # recycle WAL files #wal_buffers = -1 # min 32kB, -1 sets based on shared_buffers diff --git a/src/bin/pg_waldump/pg_waldump.c b/src/bin/pg_waldump/pg_waldump.c index f8b8afe4a7..1cd71ac2f7 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -537,18 +537,21 @@ XLogDumpDisplayRecord(XLogDumpConfig *config, XLogReaderState *record) blk); if (XLogRecHasBlockImage(record, block_id)) { - if (record->blocks[block_id].bimg_info & - BKPIMAGE_IS_COMPRESSED) + if (BKPIMAGE_IS_COMPRESSED(record->blocks[block_id].bimg_info)) { + int compression = BKPIMAGE_COMPRESSION( + record->blocks[block_id].bimg_info); + printf(" (FPW%s); hole: offset: %u, length: %u, " - "compression saved: %u", + "compression method %d/%s, saved: %u", XLogRecBlockImageApply(record, block_id) ? "" : " for WAL verification", record->blocks[block_id].hole_offset, record->blocks[block_id].hole_length, + compression, wal_compression_name(compression), BLCKSZ - - record->blocks[block_id].hole_length - - record->blocks[block_id].bimg_len); + record->blocks[block_id].hole_length - + record->blocks[block_id].bimg_len); } else { diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h index 77187c12be..e8b2c53784 100644 --- a/src/include/access/xlog.h +++ b/src/include/access/xlog.h @@ -116,7 +116,7 @@ extern char *XLogArchiveCommand; extern bool EnableHotStandby; extern bool fullPageWrites; extern bool wal_log_hints; -extern bool wal_compression; +extern int wal_compression; 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 26a743b6b6..8b740af66d 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,14 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +/* These are the compression IDs written into bimg_info */ +typedef enum WalCompression +{ + WAL_COMPRESSION_NONE, + 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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..2a60c0fb92 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 @@ -144,9 +144,18 @@ typedef struct XLogRecordBlockImageHeader /* Information stored in bimg_info */ #define BKPIMAGE_HAS_HOLE 0x01 /* page image has "hole" */ -#define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ -#define BKPIMAGE_APPLY 0x04 /* page image should be restored during +#define BKPIMAGE_APPLY 0x02 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x04 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x08 /* 0=none, 1=pglz, 2=zlib */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 2 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) +#define BKPIMAGE_IS_COMPRESSED(info) (BKPIMAGE_COMPRESSION(info) != 0) /* * Extra header information used when page image has "hole" and -- 2.17.0 --P+33d92oIH25kiaB Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="v8-0002-Run-011_crash_recovery.pl-with-wal_level-minimal.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 3/8] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: Andrey Borodin @ 2021-02-27 04:03 UTC (permalink / raw) TODO: bump XLOG_PAGE_MAGIC --- src/backend/Makefile | 2 +- src/backend/access/transam/xlog.c | 10 ++++ src/backend/access/transam/xloginsert.c | 52 +++++++++++++++++++-- src/backend/access/transam/xlogreader.c | 62 ++++++++++++++++++++++++- src/backend/utils/misc/guc.c | 11 +++++ 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 ++-- 9 files changed, 144 insertions(+), 12 deletions(-) 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 e3128564e1..0183589b4d 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..ee73bc3afd 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 = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, + Z_DEFAULT_COMPRESSION); + 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..143df55fcb 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 +1586,38 @@ 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 = 0; + 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/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..b80759ed45 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 = 0, + WAL_COMPRESSION_ZLIB = 1, +} 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 --f0KYrhQ4vYSV2aJu Content-Type: text/x-diff; charset=us-ascii Content-Disposition: attachment; filename="0004-wal_compression_method-default-to-zlib.patch" ^ permalink raw reply [nested|flat] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 f4d1ce5dea..15da91a8dd 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 --XsQoSWH+UP9D9v3l 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] 105+ messages in thread
* [PATCH 01/10] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 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 f4d1ce5dea..15da91a8dd 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --jozmn01XJZjDjM3N 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] 105+ messages in thread
* [PATCH 01/12] Allow alternate compression methods for wal_compression @ 2021-02-27 04:03 Andrey Borodin <[email protected]> 0 siblings, 0 replies; 105+ 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 | 67 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 64 +++++++++++++++++- 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 | 16 +++++ src/include/access/xlogrecord.h | 11 ++- 10 files changed, 187 insertions(+), 13 deletions(-) diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index ee4925d6d9..bed89d055d 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 6f8810e149..5b657ec724 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..a93b33464f 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,18 @@ #include "storage/proc.h" #include "utils/memutils.h" +#ifdef HAVE_LIBZ +#include <zlib.h> +/* zlib compressBound is not a macro */ +#define ZLIB_MAX_BLCKSZ BLCKSZ + (BLCKSZ>>12) + (BLCKSZ>>14) + (BLCKSZ>>25) + 13 +#else +#define ZLIB_MAX_BLCKSZ 0 +#endif + /* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, ZLIB_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +68,7 @@ typedef struct * backup block data in XLogRecordAssemble() */ /* buffer to store a compressed version of backup block image */ - char compressed_page[PGLZ_MAX_BLCKSZ]; + char compressed_page[COMPRESS_BUFSIZE]; } registered_buffer; static registered_buffer *registered_buffers; @@ -113,7 +123,8 @@ 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 @@ -625,16 +636,26 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, cbimg.hole_length = 0; } + bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; + /* * Try to compress a block image if wal_compression is enabled */ if (wal_compression) { + int compression; + /* The current compression is stored in the WAL record */ + wal_compression_name(wal_compression_method); /* Range check */ + compression = walmethods[wal_compression_method].walmethod; + Assert(compression < (1 << BKPIMAGE_COMPRESS_BITS)); + bimg.bimg_info |= + compression << BKPIMAGE_COMPRESS_OFFSET_BITS; is_compressed = XLogCompressBackupBlock(page, bimg.hole_offset, cbimg.hole_length, regbuf->compressed_page, - &compressed_len); + &compressed_len, + wal_compression_method); } /* @@ -652,8 +673,6 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, rdt_datas_last->next = ®buf->bkp_rdatas[0]; rdt_datas_last = rdt_datas_last->next; - bimg.bimg_info = (cbimg.hole_length == 0) ? 0 : BKPIMAGE_HAS_HOLE; - /* * If WAL consistency checking is enabled for the resource manager * of this WAL record, a full-page image is included in the record @@ -827,7 +846,7 @@ 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 +872,42 @@ 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 = COMPRESS_BUFSIZE; + int ret; + ret = compress2((Bytef*)dest, &len_l, (Bytef*)source, orig_len, 1); + if (ret != Z_OK) + 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..0d8830fc50 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); @@ -1535,6 +1539,30 @@ XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len) } } +/* This is a mapping indexed by wal_compression */ +// XXX: maybe this is better done as a GUC hook to assign the 1) method; and 2) level +struct walcompression walmethods[] = { + {"pglz", WAL_COMPRESSION_PGLZ}, + {"zlib", WAL_COMPRESSION_ZLIB}, +}; + +/* + * Return a statically allocated string associated with the given compression + * method. + * This is here to be visible to frontend tools like pg_rewind. + */ +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. + */ + if (compression < sizeof(walmethods)/sizeof(*walmethods)) + return walmethods[compression].name; + return "???"; +} + /* * Restore a full-page image from a backup block attached to an XLOG record. * @@ -1557,9 +1585,41 @@ RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page) if (bkpb->bimg_info & BKPIMAGE_IS_COMPRESSED) { + int compression_method = BKPIMAGE_COMPRESSION(bkpb->bimg_info); /* 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 (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, + compression_method, + wal_compression_name(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 3b36a31a47..47d30d6cda 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -517,6 +517,7 @@ 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 @@ -4760,6 +4761,16 @@ 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 86425965d0..530007f55a 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 77187c12be..2f3886431a 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..b000a21557 100644 --- a/src/include/access/xlog_internal.h +++ b/src/include/access/xlog_internal.h @@ -324,4 +324,20 @@ extern bool InArchiveRecovery; extern bool StandbyMode; extern char *recoveryRestoreCommand; +struct walcompression +{ + char *name; + int walmethod; /* Compression method to be stored in WAL */ +}; + +extern struct walcompression walmethods[]; + +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/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..7107cf6186 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 @@ -147,6 +147,15 @@ typedef struct XLogRecordBlockImageHeader #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */ #define BKPIMAGE_APPLY 0x04 /* page image should be restored during * replay */ +#define BKPIMAGE_COMPRESS_METHOD1 0x08 /* bits to encode compression method */ +#define BKPIMAGE_COMPRESS_METHOD2 0x10 /* 0=pglz; 1=zlib; */ + +/* How many bits to shift to extract compression */ +#define BKPIMAGE_COMPRESS_OFFSET_BITS 3 +/* How many bits are for compression */ +#define BKPIMAGE_COMPRESS_BITS 2 +/* Extract the compression from the bimg_info */ +#define BKPIMAGE_COMPRESSION(info) ((info >> BKPIMAGE_COMPRESS_OFFSET_BITS) & ((1<<BKPIMAGE_COMPRESS_BITS) - 1)) /* * Extra header information used when page image has "hole" and -- 2.17.0 --45Z9DzgjV8m4Oswq 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] 105+ messages in thread
* Re: Emitting JSON to file using COPY TO @ 2024-03-09 01:13 jian he <[email protected]> 0 siblings, 1 reply; 105+ messages in thread From: jian he @ 2024-03-09 01:13 UTC (permalink / raw) To: Joe Conway <[email protected]>; +Cc: Andrey M. Borodin <[email protected]>; Dean Rasheed <[email protected]>; Daniel Verite <[email protected]>; Andrew Dunstan <[email protected]>; Davin Shearer <[email protected]>; pgsql-hackers On Sat, Mar 9, 2024 at 2:03 AM Joe Conway <[email protected]> wrote: > > On 3/8/24 12:28, Andrey M. Borodin wrote: > > Hello everyone! > > > > Thanks for working on this, really nice feature! > > > >> On 9 Jan 2024, at 01:40, Joe Conway <[email protected]> wrote: > >> > >> Thanks -- will have a look > > > > Joe, recently folks proposed a lot of patches in this thread that seem like diverted from original way of implementation. > > As an author of CF entry [0] can you please comment on which patch version needs review? > > > I don't know if I agree with the proposed changes, but I have also been > waiting to see how the parallel discussion regarding COPY extensibility > shakes out. > > And there were a couple of issues found that need to be tracked down. > > Additionally I have had time/availability challenges recently. > > Overall, chances seem slim that this will make it into 17, but I have > not quite given up hope yet either. Hi. summary changes I've made in v9 patches at [0] meta: rebased. Now you need to use `git apply` or `git am`, previously copyto_json.007.diff, you need to use GNU patch. at [1], Dean Rasheed found some corner cases when the returned slot's tts_tupleDescriptor from ` ExecutorRun(cstate->queryDesc, ForwardScanDirection, 0, true); processed = ((DR_copy *) cstate->queryDesc->dest)->processed; ` cannot be used for composite_to_json. generally DestReceiver->rStartup is to send the TupleDesc to the DestReceiver, The COPY TO DestReceiver's rStartup function is copy_dest_startup, however copy_dest_startup is a no-op. That means to make the final TupleDesc of COPY TO (FORMAT JSON) operation bullet proof, we need to copy the tupDesc from CopyToState's queryDesc. This only applies to when the COPY TO source is a query (example: copy (select 1) to stdout), not a table. The above is my interpretation. at [2], Masahiko Sawada made several points. Mainly split the patch to two, one for format json, second is for options force_array. Splitting into two is easier to review, I think. My changes also addressed all the points Masahiko Sawada had mentioned. [0] https://postgr.es/m/CACJufxHd6ZRmJJBsDOGpovaVAekMS-u6AOrcw0Ja-Wyi-0kGtA@mail.gmail.com [1] https://postgr.es/m/CAEZATCWh29787xf=4NgkoixeqRHrqi0Qd33Z6_-F8t2dZ0yLCQ@mail.gmail.com [2] https://postgr.es/m/CAD21AoCb02zhZM3vXb8HSw8fwOsL+iRdEFb--Kmunv8PjPAWjw@mail.gmail.com ^ permalink raw reply [nested|flat] 105+ messages in thread
* Re: Emitting JSON to file using COPY TO @ 2024-04-01 12:00 jian he <[email protected]> parent: jian he <[email protected]> 0 siblings, 1 reply; 105+ messages in thread From: jian he @ 2024-04-01 12:00 UTC (permalink / raw) To: Joe Conway <[email protected]>; +Cc: Andrey M. Borodin <[email protected]>; Dean Rasheed <[email protected]>; Daniel Verite <[email protected]>; Andrew Dunstan <[email protected]>; Davin Shearer <[email protected]>; pgsql-hackers On Sat, Mar 9, 2024 at 9:13 AM jian he <[email protected]> wrote: > > On Sat, Mar 9, 2024 at 2:03 AM Joe Conway <[email protected]> wrote: > > > > On 3/8/24 12:28, Andrey M. Borodin wrote: > > > Hello everyone! > > > > > > Thanks for working on this, really nice feature! > > > > > >> On 9 Jan 2024, at 01:40, Joe Conway <[email protected]> wrote: > > >> > > >> Thanks -- will have a look > > > > > > Joe, recently folks proposed a lot of patches in this thread that seem like diverted from original way of implementation. > > > As an author of CF entry [0] can you please comment on which patch version needs review? > > > > > > I don't know if I agree with the proposed changes, but I have also been > > waiting to see how the parallel discussion regarding COPY extensibility > > shakes out. > > > > And there were a couple of issues found that need to be tracked down. > > > > Additionally I have had time/availability challenges recently. > > > > Overall, chances seem slim that this will make it into 17, but I have > > not quite given up hope yet either. > > Hi. > summary changes I've made in v9 patches at [0] > > meta: rebased. Now you need to use `git apply` or `git am`, previously > copyto_json.007.diff, you need to use GNU patch. > > > at [1], Dean Rasheed found some corner cases when the returned slot's > tts_tupleDescriptor > from > ` > ExecutorRun(cstate->queryDesc, ForwardScanDirection, 0, true); > processed = ((DR_copy *) cstate->queryDesc->dest)->processed; > ` > cannot be used for composite_to_json. > generally DestReceiver->rStartup is to send the TupleDesc to the DestReceiver, > The COPY TO DestReceiver's rStartup function is copy_dest_startup, > however copy_dest_startup is a no-op. > That means to make the final TupleDesc of COPY TO (FORMAT JSON) > operation bullet proof, > we need to copy the tupDesc from CopyToState's queryDesc. > This only applies to when the COPY TO source is a query (example: > copy (select 1) to stdout), not a table. > The above is my interpretation. > trying to simplify the explanation. first refer to the struct DestReceiver. COPY TO (FORMAT JSON), we didn't send the preliminary Tupdesc to the DestReceiver via the rStartup function pointer within struct _DestReceiver. `CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot)` the slot is the final slot returned via query execution. but we cannot use the tupdesc (slot->tts_tupleDescriptor) to do composite_to_json. because the final return slot Tupdesc may change during the query execution. so we need to copy the tupDesc from CopyToState's queryDesc. aslo rebased, now we can apply it cleanly. Attachments: [text/x-patch] v10-0001-introduce-json-format-for-COPY-TO-operation.patch (14.1K, ../../CACJufxGjJXKfPZoyFqgvYTBYW3GmB5WA+0H0zXkfYjrc7au7qg@mail.gmail.com/2-v10-0001-introduce-json-format-for-COPY-TO-operation.patch) download | inline diff: From 17d9f3765bb74d863e229afed59af7dd923f5379 Mon Sep 17 00:00:00 2001 From: jian he <[email protected]> Date: Mon, 1 Apr 2024 19:47:40 +0800 Subject: [PATCH v10 1/2] introduce json format for COPY TO operation. json format is only allowed in COPY TO operation. also cannot be used with header option. discussion: https://postgr.es/m/CALvfUkBxTYy5uWPFVwpk_7ii2zgT07t3d-yR_cy4sfrrLU%3Dkcg%40mail.gmail.com discussion: https://postgr.es/m/[email protected] --- doc/src/sgml/ref/copy.sgml | 5 ++ src/backend/commands/copy.c | 13 +++ src/backend/commands/copyto.c | 125 ++++++++++++++++++++--------- src/backend/parser/gram.y | 8 ++ src/backend/utils/adt/json.c | 5 +- src/include/commands/copy.h | 1 + src/include/utils/json.h | 2 + src/test/regress/expected/copy.out | 54 +++++++++++++ src/test/regress/sql/copy.sql | 38 +++++++++ 9 files changed, 208 insertions(+), 43 deletions(-) diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml index 33ce7c4e..add84dbb 100644 --- a/doc/src/sgml/ref/copy.sgml +++ b/doc/src/sgml/ref/copy.sgml @@ -218,9 +218,14 @@ COPY { <replaceable class="parameter">table_name</replaceable> [ ( <replaceable Selects the data format to be read or written: <literal>text</literal>, <literal>csv</literal> (Comma Separated Values), + <literal>json</literal> (JavaScript Object Notation), or <literal>binary</literal>. The default is <literal>text</literal>. </para> + <para> + The <literal>json</literal> option is allowed only in + <command>COPY TO</command>. + </para> </listitem> </varlistentry> diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c index f75e1d70..02f16d9e 100644 --- a/src/backend/commands/copy.c +++ b/src/backend/commands/copy.c @@ -497,6 +497,8 @@ ProcessCopyOptions(ParseState *pstate, /* default format */ ; else if (strcmp(fmt, "csv") == 0) opts_out->csv_mode = true; + else if (strcmp(fmt, "json") == 0) + opts_out->json_mode = true; else if (strcmp(fmt, "binary") == 0) opts_out->binary = true; else @@ -740,6 +742,11 @@ ProcessCopyOptions(ParseState *pstate, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), errmsg("cannot specify HEADER in BINARY mode"))); + if (opts_out->json_mode && opts_out->header_line) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("cannot specify HEADER in JSON mode"))); + /* Check quote */ if (!opts_out->csv_mode && opts_out->quote != NULL) ereport(ERROR, @@ -817,6 +824,12 @@ ProcessCopyOptions(ParseState *pstate, (errcode(ERRCODE_INVALID_PARAMETER_VALUE), errmsg("COPY FREEZE cannot be used with COPY TO"))); + /* Check json format */ + if (opts_out->json_mode && is_from) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("cannot use JSON mode in COPY FROM"))); + if (opts_out->default_print) { if (!is_from) diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c index ae8b2e36..fe2eb244 100644 --- a/src/backend/commands/copyto.c +++ b/src/backend/commands/copyto.c @@ -24,6 +24,7 @@ #include "executor/execdesc.h" #include "executor/executor.h" #include "executor/tuptable.h" +#include "funcapi.h" #include "libpq/libpq.h" #include "libpq/pqformat.h" #include "mb/pg_wchar.h" @@ -31,6 +32,7 @@ #include "pgstat.h" #include "storage/fd.h" #include "tcop/tcopprot.h" +#include "utils/json.h" #include "utils/lsyscache.h" #include "utils/memutils.h" #include "utils/rel.h" @@ -139,9 +141,20 @@ SendCopyBegin(CopyToState cstate) pq_beginmessage(&buf, PqMsg_CopyOutResponse); pq_sendbyte(&buf, format); /* overall format */ - pq_sendint16(&buf, natts); - for (i = 0; i < natts; i++) - pq_sendint16(&buf, format); /* per-column formats */ + if (!cstate->opts.json_mode) + { + pq_sendint16(&buf, natts); + for (i = 0; i < natts; i++) + pq_sendint16(&buf, format); /* per-column formats */ + } + else + { + /* + * JSON mode is always one non-binary column + */ + pq_sendint16(&buf, 1); + pq_sendint16(&buf, 0); + } pq_endmessage(&buf); cstate->copy_dest = COPY_FRONTEND; } @@ -901,11 +914,7 @@ DoCopyTo(CopyToState cstate) static void CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) { - bool need_delim = false; - FmgrInfo *out_functions = cstate->out_functions; MemoryContext oldcontext; - ListCell *cur; - char *string; MemoryContextReset(cstate->rowcontext); oldcontext = MemoryContextSwitchTo(cstate->rowcontext); @@ -916,53 +925,89 @@ CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) CopySendInt16(cstate, list_length(cstate->attnumlist)); } - /* Make sure the tuple is fully deconstructed */ - slot_getallattrs(slot); - - foreach(cur, cstate->attnumlist) + if (!cstate->opts.json_mode) { - int attnum = lfirst_int(cur); - Datum value = slot->tts_values[attnum - 1]; - bool isnull = slot->tts_isnull[attnum - 1]; + bool need_delim = false; + FmgrInfo *out_functions = cstate->out_functions; + ListCell *cur; + char *string; - if (!cstate->opts.binary) - { - if (need_delim) - CopySendChar(cstate, cstate->opts.delim[0]); - need_delim = true; - } + /* Make sure the tuple is fully deconstructed */ + slot_getallattrs(slot); - if (isnull) - { - if (!cstate->opts.binary) - CopySendString(cstate, cstate->opts.null_print_client); - else - CopySendInt32(cstate, -1); - } - else + foreach(cur, cstate->attnumlist) { + int attnum = lfirst_int(cur); + Datum value = slot->tts_values[attnum - 1]; + bool isnull = slot->tts_isnull[attnum - 1]; + if (!cstate->opts.binary) { - string = OutputFunctionCall(&out_functions[attnum - 1], - value); - if (cstate->opts.csv_mode) - CopyAttributeOutCSV(cstate, string, - cstate->opts.force_quote_flags[attnum - 1]); + if (need_delim) + CopySendChar(cstate, cstate->opts.delim[0]); + need_delim = true; + } + + if (isnull) + { + if (!cstate->opts.binary) + CopySendString(cstate, cstate->opts.null_print_client); else - CopyAttributeOutText(cstate, string); + CopySendInt32(cstate, -1); } else { - bytea *outputbytes; + if (!cstate->opts.binary) + { + string = OutputFunctionCall(&out_functions[attnum - 1], + value); + if (cstate->opts.csv_mode) + CopyAttributeOutCSV(cstate, string, + cstate->opts.force_quote_flags[attnum - 1]); + else + CopyAttributeOutText(cstate, string); + } + else + { + bytea *outputbytes; - outputbytes = SendFunctionCall(&out_functions[attnum - 1], - value); - CopySendInt32(cstate, VARSIZE(outputbytes) - VARHDRSZ); - CopySendData(cstate, VARDATA(outputbytes), - VARSIZE(outputbytes) - VARHDRSZ); + outputbytes = SendFunctionCall(&out_functions[attnum - 1], + value); + CopySendInt32(cstate, VARSIZE(outputbytes) - VARHDRSZ); + CopySendData(cstate, VARDATA(outputbytes), + VARSIZE(outputbytes) - VARHDRSZ); + } } } } + else + { + Datum rowdata; + StringInfo result; + + /* + * if the COPY TO command's source data is from a query, not a table, + * then we need to copy the TupleDesc from the cstate->queryDesc. + * because returning slot's TupleDesc may change during query execution, + * but composite_to_json requires correct TupleDesc. + * so we need copy from cstate->queryDesc to make it bullet proof. + */ + if(!cstate->rel) + { + for (int i = 0; i < slot->tts_tupleDescriptor->natts; i++) + { + memcpy(TupleDescAttr(slot->tts_tupleDescriptor, i), + TupleDescAttr(cstate->queryDesc->tupDesc, i), + 1 * sizeof(FormData_pg_attribute)); + } + BlessTupleDesc(slot->tts_tupleDescriptor); + } + rowdata = ExecFetchSlotHeapTupleDatum(slot); + result = makeStringInfo(); + composite_to_json(rowdata, result, false); + + CopySendData(cstate, result->data, result->len); + } CopySendEndOfRow(cstate); diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y index f1af6147..d5ccdf0b 100644 --- a/src/backend/parser/gram.y +++ b/src/backend/parser/gram.y @@ -3442,6 +3442,10 @@ copy_opt_item: { $$ = makeDefElem("format", (Node *) makeString("csv"), @1); } + | JSON + { + $$ = makeDefElem("format", (Node *) makeString("json"), @1); + } | HEADER_P { $$ = makeDefElem("header", (Node *) makeBoolean(true), @1); @@ -3524,6 +3528,10 @@ copy_generic_opt_elem: { $$ = makeDefElem($1, $2, @1); } + | FORMAT_LA copy_generic_opt_arg + { + $$ = makeDefElem("format", $2, @1); + } ; copy_generic_opt_arg: diff --git a/src/backend/utils/adt/json.c b/src/backend/utils/adt/json.c index d719a61f..fabd4e61 100644 --- a/src/backend/utils/adt/json.c +++ b/src/backend/utils/adt/json.c @@ -83,8 +83,6 @@ typedef struct JsonAggState JsonUniqueBuilderState unique_check; } JsonAggState; -static void composite_to_json(Datum composite, StringInfo result, - bool use_line_feeds); static void array_dim_to_json(StringInfo result, int dim, int ndims, int *dims, Datum *vals, bool *nulls, int *valcount, JsonTypeCategory tcategory, Oid outfuncoid, @@ -507,8 +505,9 @@ array_to_json_internal(Datum array, StringInfo result, bool use_line_feeds) /* * Turn a composite / record into JSON. + * Exported so COPY TO can use it. */ -static void +void composite_to_json(Datum composite, StringInfo result, bool use_line_feeds) { HeapTupleHeader td; diff --git a/src/include/commands/copy.h b/src/include/commands/copy.h index 141fd48d..ff6ecc7a 100644 --- a/src/include/commands/copy.h +++ b/src/include/commands/copy.h @@ -62,6 +62,7 @@ typedef struct CopyFormatOptions bool binary; /* binary format? */ bool freeze; /* freeze rows on loading? */ bool csv_mode; /* Comma Separated Value format? */ + bool json_mode; /* JSON format? */ CopyHeaderChoice header_line; /* header line? */ char *null_print; /* NULL marker string (server encoding!) */ int null_print_len; /* length of same */ diff --git a/src/include/utils/json.h b/src/include/utils/json.h index 6d7f1b38..d5631171 100644 --- a/src/include/utils/json.h +++ b/src/include/utils/json.h @@ -17,6 +17,8 @@ #include "lib/stringinfo.h" /* functions in json.c */ +extern void composite_to_json(Datum composite, StringInfo result, + bool use_line_feeds); extern void escape_json(StringInfo buf, const char *str); extern char *JsonEncodeDateTime(char *buf, Datum value, Oid typid, const int *tzp); diff --git a/src/test/regress/expected/copy.out b/src/test/regress/expected/copy.out index b48365ec..0c5ade47 100644 --- a/src/test/regress/expected/copy.out +++ b/src/test/regress/expected/copy.out @@ -42,6 +42,60 @@ copy copytest3 to stdout csv header; c1,"col with , comma","col with "" quote" 1,a,1 2,b,2 +--- test copying in JSON mode with various styles +copy copytest to stdout json; +{"style":"DOS","test":"abc\r\ndef","filler":1} +{"style":"Unix","test":"abc\ndef","filler":2} +{"style":"Mac","test":"abc\rdef","filler":3} +{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +copy copytest to stdout (format json); +{"style":"DOS","test":"abc\r\ndef","filler":1} +{"style":"Unix","test":"abc\ndef","filler":2} +{"style":"Mac","test":"abc\rdef","filler":3} +{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +-- Error +copy copytest to stdout (format json, header); +ERROR: cannot specify HEADER in JSON mode +-- Error +copy copytest from stdout (format json); +ERROR: cannot use JSON mode in COPY FROM +-- embedded escaped characters +create temp table copyjsontest ( + id bigserial, + f1 text, + f2 timestamptz); +insert into copyjsontest + select g.i, + CASE WHEN g.i % 2 = 0 THEN + 'line with '' in it: ' || g.i::text + ELSE + 'line with " in it: ' || g.i::text + END, + 'Mon Feb 10 17:32:01 1997 PST' + from generate_series(1,5) as g(i); +insert into copyjsontest (f1) values +(E'aaa\"bbb'::text), +(E'aaa\\bbb'::text), +(E'aaa\/bbb'::text), +(E'aaa\bbbb'::text), +(E'aaa\fbbb'::text), +(E'aaa\nbbb'::text), +(E'aaa\rbbb'::text), +(E'aaa\tbbb'::text); +copy copyjsontest to stdout json; +{"id":1,"f1":"line with \" in it: 1","f2":"1997-02-10T17:32:01-08:00"} +{"id":2,"f1":"line with ' in it: 2","f2":"1997-02-10T17:32:01-08:00"} +{"id":3,"f1":"line with \" in it: 3","f2":"1997-02-10T17:32:01-08:00"} +{"id":4,"f1":"line with ' in it: 4","f2":"1997-02-10T17:32:01-08:00"} +{"id":5,"f1":"line with \" in it: 5","f2":"1997-02-10T17:32:01-08:00"} +{"id":1,"f1":"aaa\"bbb","f2":null} +{"id":2,"f1":"aaa\\bbb","f2":null} +{"id":3,"f1":"aaa/bbb","f2":null} +{"id":4,"f1":"aaa\bbbb","f2":null} +{"id":5,"f1":"aaa\fbbb","f2":null} +{"id":6,"f1":"aaa\nbbb","f2":null} +{"id":7,"f1":"aaa\rbbb","f2":null} +{"id":8,"f1":"aaa\tbbb","f2":null} create temp table copytest4 ( c1 int, "colname with tab: " text); diff --git a/src/test/regress/sql/copy.sql b/src/test/regress/sql/copy.sql index 43d2e906..da6b0b0a 100644 --- a/src/test/regress/sql/copy.sql +++ b/src/test/regress/sql/copy.sql @@ -54,6 +54,44 @@ this is just a line full of junk that would error out if parsed copy copytest3 to stdout csv header; +--- test copying in JSON mode with various styles +copy copytest to stdout json; + +copy copytest to stdout (format json); + +-- Error +copy copytest to stdout (format json, header); +-- Error +copy copytest from stdout (format json); + +-- embedded escaped characters +create temp table copyjsontest ( + id bigserial, + f1 text, + f2 timestamptz); + +insert into copyjsontest + select g.i, + CASE WHEN g.i % 2 = 0 THEN + 'line with '' in it: ' || g.i::text + ELSE + 'line with " in it: ' || g.i::text + END, + 'Mon Feb 10 17:32:01 1997 PST' + from generate_series(1,5) as g(i); + +insert into copyjsontest (f1) values +(E'aaa\"bbb'::text), +(E'aaa\\bbb'::text), +(E'aaa\/bbb'::text), +(E'aaa\bbbb'::text), +(E'aaa\fbbb'::text), +(E'aaa\nbbb'::text), +(E'aaa\rbbb'::text), +(E'aaa\tbbb'::text); + +copy copyjsontest to stdout json; + create temp table copytest4 ( c1 int, "colname with tab: " text); base-commit: 7aa00f1360e0c6938fdf32d3fbb8b847b6098b88 -- 2.34.1 [text/x-patch] v10-0002-Add-option-force_array-for-COPY-TO-JSON-fomrat.patch (8.4K, ../../CACJufxGjJXKfPZoyFqgvYTBYW3GmB5WA+0H0zXkfYjrc7au7qg@mail.gmail.com/3-v10-0002-Add-option-force_array-for-COPY-TO-JSON-fomrat.patch) download | inline diff: From c939262f6266cea38761aa89cc6bb495b8d38ccb Mon Sep 17 00:00:00 2001 From: jian he <[email protected]> Date: Mon, 1 Apr 2024 19:53:10 +0800 Subject: [PATCH v10 2/2] Add option force_array for COPY TO JSON fomrat force_array option only apply to COPY TO operation with JSON format. make add opening brackets and close brackets for the whole json output. also, separate each json record with a comma. discussion: https://postgr.es/m/CALvfUkBxTYy5uWPFVwpk_7ii2zgT07t3d-yR_cy4sfrrLU%3Dkcg%40mail.gmail.com discussion: https://postgr.es/m/[email protected] --- doc/src/sgml/ref/copy.sgml | 14 ++++++++++++++ src/backend/commands/copy.c | 17 +++++++++++++++++ src/backend/commands/copyto.c | 29 +++++++++++++++++++++++++++++ src/include/commands/copy.h | 1 + src/test/regress/expected/copy.out | 24 ++++++++++++++++++++++++ src/test/regress/sql/copy.sql | 9 +++++++++ 6 files changed, 94 insertions(+) diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml index add84dbb..f4100cba 100644 --- a/doc/src/sgml/ref/copy.sgml +++ b/doc/src/sgml/ref/copy.sgml @@ -43,6 +43,7 @@ COPY { <replaceable class="parameter">table_name</replaceable> [ ( <replaceable FORCE_QUOTE { ( <replaceable class="parameter">column_name</replaceable> [, ...] ) | * } FORCE_NOT_NULL { ( <replaceable class="parameter">column_name</replaceable> [, ...] ) | * } FORCE_NULL { ( <replaceable class="parameter">column_name</replaceable> [, ...] ) | * } + FORCE_ARRAY [ <replaceable class="parameter">boolean</replaceable> ] ON_ERROR '<replaceable class="parameter">error_action</replaceable>' ENCODING '<replaceable class="parameter">encoding_name</replaceable>' LOG_VERBOSITY <replaceable class="parameter">mode</replaceable> @@ -390,6 +391,19 @@ COPY { <replaceable class="parameter">table_name</replaceable> [ ( <replaceable </listitem> </varlistentry> + <varlistentry> + <term><literal>FORCE_ARRAY</literal></term> + <listitem> + <para> + Force output of square brackets as array decorations at the beginning + and end of output, and commas between the rows. It is allowed only in + <command>COPY TO</command>, and only when using + <literal>JSON</literal> format. The default is + <literal>false</literal>. + </para> + </listitem> + </varlistentry> + <varlistentry> <term><literal>ON_ERROR</literal></term> <listitem> diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c index 02f16d9e..2ffd6978 100644 --- a/src/backend/commands/copy.c +++ b/src/backend/commands/copy.c @@ -473,6 +473,7 @@ ProcessCopyOptions(ParseState *pstate, bool header_specified = false; bool on_error_specified = false; bool log_verbosity_specified = false; + bool force_array_specified = false; ListCell *option; /* Support external use for option sanity checking */ @@ -627,6 +628,13 @@ ProcessCopyOptions(ParseState *pstate, defel->defname), parser_errposition(pstate, defel->location))); } + else if (strcmp(defel->defname, "force_array") == 0) + { + if (force_array_specified) + errorConflictingDefElem(defel, pstate); + force_array_specified = true; + opts_out->force_array = defGetBoolean(defel); + } else if (strcmp(defel->defname, "on_error") == 0) { if (on_error_specified) @@ -830,6 +838,15 @@ ProcessCopyOptions(ParseState *pstate, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), errmsg("cannot use JSON mode in COPY FROM"))); + if (!opts_out->json_mode && opts_out->force_array) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("COPY FORCE_ARRAY requires JSON mode"))); + if (!opts_out->json_mode && force_array_specified) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("COPY FORCE_ARRAY only available in JSON mode"))); + if (opts_out->default_print) { if (!is_from) diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c index fe2eb244..8e4edb6e 100644 --- a/src/backend/commands/copyto.c +++ b/src/backend/commands/copyto.c @@ -81,6 +81,7 @@ typedef struct CopyToStateData List *attnumlist; /* integer list of attnums to copy */ char *filename; /* filename, or NULL for STDOUT */ bool is_program; /* is 'filename' a program to popen? */ + bool json_row_delim_needed; /* need delimiter to start next json array element */ copy_data_dest_cb data_dest_cb; /* function for writing data */ CopyFormatOptions opts; @@ -853,6 +854,16 @@ DoCopyTo(CopyToState cstate) CopySendEndOfRow(cstate); } + /* + * If JSON has been requested, and FORCE_ARRAY has been specified send + * the opening bracket. + */ + if (cstate->opts.json_mode && cstate->opts.force_array) + { + CopySendChar(cstate, '['); + CopySendEndOfRow(cstate); + } + } if (cstate->rel) @@ -900,6 +911,15 @@ DoCopyTo(CopyToState cstate) CopySendEndOfRow(cstate); } + /* + * If JSON has been requested, and FORCE_ARRAY has been specified send the + * closing bracket. + */ + if (cstate->opts.json_mode && cstate->opts.force_array) + { + CopySendChar(cstate, ']'); + CopySendEndOfRow(cstate); + } MemoryContextDelete(cstate->rowcontext); if (fe_copy) @@ -1006,6 +1026,15 @@ CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) result = makeStringInfo(); composite_to_json(rowdata, result, false); + if (cstate->json_row_delim_needed && cstate->opts.force_array) + CopySendChar(cstate, ','); + else if (cstate->opts.force_array) + { + /* first row needs no delimiter */ + CopySendChar(cstate, ' '); + cstate->json_row_delim_needed = true; + } + CopySendData(cstate, result->data, result->len); } diff --git a/src/include/commands/copy.h b/src/include/commands/copy.h index ff6ecc7a..f76fe8fa 100644 --- a/src/include/commands/copy.h +++ b/src/include/commands/copy.h @@ -81,6 +81,7 @@ typedef struct CopyFormatOptions List *force_null; /* list of column names */ bool force_null_all; /* FORCE_NULL *? */ bool *force_null_flags; /* per-column CSV FN flags */ + bool force_array; /* add JSON array decorations */ bool convert_selectively; /* do selective binary conversion? */ CopyOnErrorChoice on_error; /* what to do when error happened */ CopyLogVerbosityChoice log_verbosity; /* verbosity of logged messages */ diff --git a/src/test/regress/expected/copy.out b/src/test/regress/expected/copy.out index 0c5ade47..7812768c 100644 --- a/src/test/regress/expected/copy.out +++ b/src/test/regress/expected/copy.out @@ -59,6 +59,30 @@ ERROR: cannot specify HEADER in JSON mode -- Error copy copytest from stdout (format json); ERROR: cannot use JSON mode in COPY FROM +--Error +copy copytest to stdout (format csv, force_array false); +ERROR: COPY FORCE_ARRAY only available in JSON mode +copy copytest from stdin (format json, force_array true); +ERROR: cannot use JSON mode in COPY FROM +copy copytest to stdout (format json, force_array); +[ + {"style":"DOS","test":"abc\r\ndef","filler":1} +,{"style":"Unix","test":"abc\ndef","filler":2} +,{"style":"Mac","test":"abc\rdef","filler":3} +,{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +] +copy copytest to stdout (format json, force_array true); +[ + {"style":"DOS","test":"abc\r\ndef","filler":1} +,{"style":"Unix","test":"abc\ndef","filler":2} +,{"style":"Mac","test":"abc\rdef","filler":3} +,{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +] +copy copytest to stdout (format json, force_array false); +{"style":"DOS","test":"abc\r\ndef","filler":1} +{"style":"Unix","test":"abc\ndef","filler":2} +{"style":"Mac","test":"abc\rdef","filler":3} +{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} -- embedded escaped characters create temp table copyjsontest ( id bigserial, diff --git a/src/test/regress/sql/copy.sql b/src/test/regress/sql/copy.sql index da6b0b0a..f685193b 100644 --- a/src/test/regress/sql/copy.sql +++ b/src/test/regress/sql/copy.sql @@ -64,6 +64,15 @@ copy copytest to stdout (format json, header); -- Error copy copytest from stdout (format json); +--Error +copy copytest to stdout (format csv, force_array false); +copy copytest from stdin (format json, force_array true); + +copy copytest to stdout (format json, force_array); + +copy copytest to stdout (format json, force_array true); + +copy copytest to stdout (format json, force_array false); -- embedded escaped characters create temp table copyjsontest ( id bigserial, -- 2.34.1 ^ permalink raw reply [nested|flat] 105+ messages in thread
* Re: Emitting JSON to file using COPY TO @ 2024-08-19 00:00 jian he <[email protected]> parent: jian he <[email protected]> 0 siblings, 1 reply; 105+ messages in thread From: jian he @ 2024-08-19 00:00 UTC (permalink / raw) To: Joe Conway <[email protected]>; +Cc: Andrey M. Borodin <[email protected]>; Dean Rasheed <[email protected]>; Daniel Verite <[email protected]>; Andrew Dunstan <[email protected]>; Davin Shearer <[email protected]>; pgsql-hackers On Mon, Apr 1, 2024 at 8:00 PM jian he <[email protected]> wrote: > rebased. minor cosmetic error message change. I think all the issues in this thread have been addressed. Attachments: [text/x-patch] v11-0001-introduce-json-format-for-COPY-TO.patch (12.0K, ../../CACJufxET6LJWbywU_odC_k3vYwEKhNQjiWTDk=37ecQG5mFk5Q@mail.gmail.com/2-v11-0001-introduce-json-format-for-COPY-TO.patch) download | inline diff: From b96dfe41f0935b08b1190f399e29ee2450169529 Mon Sep 17 00:00:00 2001 From: jian he <[email protected]> Date: Sat, 17 Aug 2024 11:08:25 +0800 Subject: [PATCH v11 1/2] introduce json format for COPY TO json format is only allowed in COPY TO operation. also cannot be used with header option. discussion: https://postgr.es/m/CALvfUkBxTYy5uWPFVwpk_7ii2zgT07t3d-yR_cy4sfrrLU%3Dkcg%40mail.gmail.com discussion: https://postgr.es/m/[email protected] --- doc/src/sgml/ref/copy.sgml | 5 +++ src/backend/commands/copy.c | 13 +++++++ src/backend/commands/copyto.c | 51 +++++++++++++++++++++++++--- src/backend/parser/gram.y | 8 +++++ src/backend/utils/adt/json.c | 5 ++- src/include/commands/copy.h | 1 + src/include/utils/json.h | 2 ++ src/test/regress/expected/copy.out | 54 ++++++++++++++++++++++++++++++ src/test/regress/sql/copy.sql | 38 +++++++++++++++++++++ 9 files changed, 169 insertions(+), 8 deletions(-) diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml index 1518af8a04..616abf508e 100644 --- a/doc/src/sgml/ref/copy.sgml +++ b/doc/src/sgml/ref/copy.sgml @@ -218,9 +218,14 @@ COPY { <replaceable class="parameter">table_name</replaceable> [ ( <replaceable Selects the data format to be read or written: <literal>text</literal>, <literal>csv</literal> (Comma Separated Values), + <literal>json</literal> (JavaScript Object Notation), or <literal>binary</literal>. The default is <literal>text</literal>. </para> + <para> + The <literal>json</literal> option is allowed only in + <command>COPY TO</command>. + </para> </listitem> </varlistentry> diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c index 3bb579a3a4..a3ee65d000 100644 --- a/src/backend/commands/copy.c +++ b/src/backend/commands/copy.c @@ -494,6 +494,8 @@ ProcessCopyOptions(ParseState *pstate, /* default format */ ; else if (strcmp(fmt, "csv") == 0) opts_out->csv_mode = true; + else if (strcmp(fmt, "json") == 0) + opts_out->json_mode = true; else if (strcmp(fmt, "binary") == 0) opts_out->binary = true; else @@ -739,6 +741,11 @@ ProcessCopyOptions(ParseState *pstate, /*- translator: %s is the name of a COPY option, e.g. ON_ERROR */ errmsg("cannot specify %s in BINARY mode", "HEADER"))); + if (opts_out->json_mode && opts_out->header_line) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("cannot specify %s in JSON mode", "HEADER"))); + /* Check quote */ if (!opts_out->csv_mode && opts_out->quote != NULL) ereport(ERROR, @@ -837,6 +844,12 @@ ProcessCopyOptions(ParseState *pstate, errmsg("COPY %s cannot be used with %s", "FREEZE", "COPY TO"))); + /* Check json format */ + if (opts_out->json_mode && is_from) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("COPY json mode cannot be used with %s", "COPY FROM"))); + if (opts_out->default_print) { if (!is_from) diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c index 91de442f43..14ba9fde50 100644 --- a/src/backend/commands/copyto.c +++ b/src/backend/commands/copyto.c @@ -24,6 +24,7 @@ #include "executor/execdesc.h" #include "executor/executor.h" #include "executor/tuptable.h" +#include "funcapi.h" #include "libpq/libpq.h" #include "libpq/pqformat.h" #include "mb/pg_wchar.h" @@ -31,6 +32,7 @@ #include "pgstat.h" #include "storage/fd.h" #include "tcop/tcopprot.h" +#include "utils/json.h" #include "utils/lsyscache.h" #include "utils/memutils.h" #include "utils/rel.h" @@ -139,9 +141,20 @@ SendCopyBegin(CopyToState cstate) pq_beginmessage(&buf, PqMsg_CopyOutResponse); pq_sendbyte(&buf, format); /* overall format */ - pq_sendint16(&buf, natts); - for (i = 0; i < natts; i++) - pq_sendint16(&buf, format); /* per-column formats */ + if (!cstate->opts.json_mode) + { + pq_sendint16(&buf, natts); + for (i = 0; i < natts; i++) + pq_sendint16(&buf, format); /* per-column formats */ + } + else + { + /* + * JSON mode is always one non-binary column + */ + pq_sendint16(&buf, 1); + pq_sendint16(&buf, 0); + } pq_endmessage(&buf); cstate->copy_dest = COPY_FRONTEND; } @@ -917,7 +930,7 @@ CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) /* Make sure the tuple is fully deconstructed */ slot_getallattrs(slot); - if (!cstate->opts.binary) + if (!cstate->opts.binary && !cstate->opts.json_mode) { bool need_delim = false; @@ -945,7 +958,7 @@ CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) } } } - else + else if (!cstate->opts.json_mode) { foreach_int(attnum, cstate->attnumlist) { @@ -965,6 +978,34 @@ CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) } } } + else + { + Datum rowdata; + StringInfo result; + + /* + * if COPY TO source data is from a query, not a table, then we need + * copy CopyToState->TupleDesc->attrs to + * slot->tts_tupleDescriptor->attrs because the slot's TupleDesc->attrs + * may change during query execution, but composite_to_json requires + * correct TupleDesc->attrs for constructing the json keys. + * composite_to_json will iterate each TupleDesc->attrs so no need to + * copy other filed in cstate->queryDesc->tupDesc. + */ + if(!cstate->rel) + { + memcpy(TupleDescAttr(slot->tts_tupleDescriptor, 0), + TupleDescAttr(cstate->queryDesc->tupDesc, 0), + cstate->queryDesc->tupDesc->natts * sizeof(FormData_pg_attribute)); + + BlessTupleDesc(slot->tts_tupleDescriptor); + } + rowdata = ExecFetchSlotHeapTupleDatum(slot); + result = makeStringInfo(); + composite_to_json(rowdata, result, false); + + CopySendData(cstate, result->data, result->len); + } CopySendEndOfRow(cstate); diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y index c3f25582c3..76bd6f8949 100644 --- a/src/backend/parser/gram.y +++ b/src/backend/parser/gram.y @@ -3500,6 +3500,10 @@ copy_opt_item: { $$ = makeDefElem("format", (Node *) makeString("csv"), @1); } + | JSON + { + $$ = makeDefElem("format", (Node *) makeString("json"), @1); + } | HEADER_P { $$ = makeDefElem("header", (Node *) makeBoolean(true), @1); @@ -3582,6 +3586,10 @@ copy_generic_opt_elem: { $$ = makeDefElem($1, $2, @1); } + | FORMAT_LA copy_generic_opt_arg + { + $$ = makeDefElem("format", $2, @1); + } ; copy_generic_opt_arg: diff --git a/src/backend/utils/adt/json.c b/src/backend/utils/adt/json.c index 4eeeeaf0a6..b47ae14f20 100644 --- a/src/backend/utils/adt/json.c +++ b/src/backend/utils/adt/json.c @@ -85,8 +85,6 @@ typedef struct JsonAggState JsonUniqueBuilderState unique_check; } JsonAggState; -static void composite_to_json(Datum composite, StringInfo result, - bool use_line_feeds); static void array_dim_to_json(StringInfo result, int dim, int ndims, int *dims, Datum *vals, bool *nulls, int *valcount, JsonTypeCategory tcategory, Oid outfuncoid, @@ -516,8 +514,9 @@ array_to_json_internal(Datum array, StringInfo result, bool use_line_feeds) /* * Turn a composite / record into JSON. + * Exported so COPY TO can use it. */ -static void +void composite_to_json(Datum composite, StringInfo result, bool use_line_feeds) { HeapTupleHeader td; diff --git a/src/include/commands/copy.h b/src/include/commands/copy.h index 141fd48dc1..ff6ecc7ae7 100644 --- a/src/include/commands/copy.h +++ b/src/include/commands/copy.h @@ -62,6 +62,7 @@ typedef struct CopyFormatOptions bool binary; /* binary format? */ bool freeze; /* freeze rows on loading? */ bool csv_mode; /* Comma Separated Value format? */ + bool json_mode; /* JSON format? */ CopyHeaderChoice header_line; /* header line? */ char *null_print; /* NULL marker string (server encoding!) */ int null_print_len; /* length of same */ diff --git a/src/include/utils/json.h b/src/include/utils/json.h index 79c1062e1b..c904ef6c6e 100644 --- a/src/include/utils/json.h +++ b/src/include/utils/json.h @@ -17,6 +17,8 @@ #include "lib/stringinfo.h" /* functions in json.c */ +extern void composite_to_json(Datum composite, StringInfo result, + bool use_line_feeds); extern void escape_json(StringInfo buf, const char *str); extern void escape_json_with_len(StringInfo buf, const char *str, int len); extern void escape_json_text(StringInfo buf, const text *txt); diff --git a/src/test/regress/expected/copy.out b/src/test/regress/expected/copy.out index 44114089a6..77cb36cd61 100644 --- a/src/test/regress/expected/copy.out +++ b/src/test/regress/expected/copy.out @@ -42,6 +42,60 @@ copy copytest3 to stdout csv header; c1,"col with , comma","col with "" quote" 1,a,1 2,b,2 +--- test copying in JSON mode with various styles +copy copytest to stdout json; +{"style":"DOS","test":"abc\r\ndef","filler":1} +{"style":"Unix","test":"abc\ndef","filler":2} +{"style":"Mac","test":"abc\rdef","filler":3} +{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +copy copytest to stdout (format json); +{"style":"DOS","test":"abc\r\ndef","filler":1} +{"style":"Unix","test":"abc\ndef","filler":2} +{"style":"Mac","test":"abc\rdef","filler":3} +{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +-- Error +copy copytest to stdout (format json, header); +ERROR: cannot specify HEADER in JSON mode +-- Error +copy copytest from stdout (format json); +ERROR: COPY json mode cannot be used with COPY FROM +-- embedded escaped characters +create temp table copyjsontest ( + id bigserial, + f1 text, + f2 timestamptz); +insert into copyjsontest + select g.i, + CASE WHEN g.i % 2 = 0 THEN + 'line with '' in it: ' || g.i::text + ELSE + 'line with " in it: ' || g.i::text + END, + 'Mon Feb 10 17:32:01 1997 PST' + from generate_series(1,5) as g(i); +insert into copyjsontest (f1) values +(E'aaa\"bbb'::text), +(E'aaa\\bbb'::text), +(E'aaa\/bbb'::text), +(E'aaa\bbbb'::text), +(E'aaa\fbbb'::text), +(E'aaa\nbbb'::text), +(E'aaa\rbbb'::text), +(E'aaa\tbbb'::text); +copy copyjsontest to stdout json; +{"id":1,"f1":"line with \" in it: 1","f2":"1997-02-10T17:32:01-08:00"} +{"id":2,"f1":"line with ' in it: 2","f2":"1997-02-10T17:32:01-08:00"} +{"id":3,"f1":"line with \" in it: 3","f2":"1997-02-10T17:32:01-08:00"} +{"id":4,"f1":"line with ' in it: 4","f2":"1997-02-10T17:32:01-08:00"} +{"id":5,"f1":"line with \" in it: 5","f2":"1997-02-10T17:32:01-08:00"} +{"id":1,"f1":"aaa\"bbb","f2":null} +{"id":2,"f1":"aaa\\bbb","f2":null} +{"id":3,"f1":"aaa/bbb","f2":null} +{"id":4,"f1":"aaa\bbbb","f2":null} +{"id":5,"f1":"aaa\fbbb","f2":null} +{"id":6,"f1":"aaa\nbbb","f2":null} +{"id":7,"f1":"aaa\rbbb","f2":null} +{"id":8,"f1":"aaa\tbbb","f2":null} create temp table copytest4 ( c1 int, "colname with tab: " text); diff --git a/src/test/regress/sql/copy.sql b/src/test/regress/sql/copy.sql index e2dd24cb35..67f77f6512 100644 --- a/src/test/regress/sql/copy.sql +++ b/src/test/regress/sql/copy.sql @@ -54,6 +54,44 @@ this is just a line full of junk that would error out if parsed copy copytest3 to stdout csv header; +--- test copying in JSON mode with various styles +copy copytest to stdout json; + +copy copytest to stdout (format json); + +-- Error +copy copytest to stdout (format json, header); +-- Error +copy copytest from stdout (format json); + +-- embedded escaped characters +create temp table copyjsontest ( + id bigserial, + f1 text, + f2 timestamptz); + +insert into copyjsontest + select g.i, + CASE WHEN g.i % 2 = 0 THEN + 'line with '' in it: ' || g.i::text + ELSE + 'line with " in it: ' || g.i::text + END, + 'Mon Feb 10 17:32:01 1997 PST' + from generate_series(1,5) as g(i); + +insert into copyjsontest (f1) values +(E'aaa\"bbb'::text), +(E'aaa\\bbb'::text), +(E'aaa\/bbb'::text), +(E'aaa\bbbb'::text), +(E'aaa\fbbb'::text), +(E'aaa\nbbb'::text), +(E'aaa\rbbb'::text), +(E'aaa\tbbb'::text); + +copy copyjsontest to stdout json; + create temp table copytest4 ( c1 int, "colname with tab: " text); base-commit: 03e9b958eef44046c0092f1f34da4b2de0f9071d -- 2.34.1 [text/x-patch] v11-0002-Add-option-force_array-for-COPY-TO.patch (8.0K, ../../CACJufxET6LJWbywU_odC_k3vYwEKhNQjiWTDk=37ecQG5mFk5Q@mail.gmail.com/3-v11-0002-Add-option-force_array-for-COPY-TO.patch) download | inline diff: From cb006c2ad2ca652efa884e058a0cec491fde4b8d Mon Sep 17 00:00:00 2001 From: jian he <[email protected]> Date: Sat, 17 Aug 2024 12:45:35 +0800 Subject: [PATCH v11 2/2] Add option force_array for COPY TO force_array option can only be used in COPY TO with JSON format. it make the output json output like json array type. discussion: https://postgr.es/m/CALvfUkBxTYy5uWPFVwpk_7ii2zgT07t3d-yR_cy4sfrrLU%3Dkcg%40mail.gmail.com discussion: https://postgr.es/m/[email protected] --- doc/src/sgml/ref/copy.sgml | 14 ++++++++++++++ src/backend/commands/copy.c | 13 +++++++++++++ src/backend/commands/copyto.c | 29 +++++++++++++++++++++++++++++ src/include/commands/copy.h | 1 + src/test/regress/expected/copy.out | 23 +++++++++++++++++++++++ src/test/regress/sql/copy.sql | 9 +++++++++ 6 files changed, 89 insertions(+) diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml index 616abf508e..4659e49ec3 100644 --- a/doc/src/sgml/ref/copy.sgml +++ b/doc/src/sgml/ref/copy.sgml @@ -43,6 +43,7 @@ COPY { <replaceable class="parameter">table_name</replaceable> [ ( <replaceable FORCE_QUOTE { ( <replaceable class="parameter">column_name</replaceable> [, ...] ) | * } FORCE_NOT_NULL { ( <replaceable class="parameter">column_name</replaceable> [, ...] ) | * } FORCE_NULL { ( <replaceable class="parameter">column_name</replaceable> [, ...] ) | * } + FORCE_ARRAY [ <replaceable class="parameter">boolean</replaceable> ] ON_ERROR <replaceable class="parameter">error_action</replaceable> ENCODING '<replaceable class="parameter">encoding_name</replaceable>' LOG_VERBOSITY <replaceable class="parameter">verbosity</replaceable> @@ -390,6 +391,19 @@ COPY { <replaceable class="parameter">table_name</replaceable> [ ( <replaceable </listitem> </varlistentry> + <varlistentry> + <term><literal>FORCE_ARRAY</literal></term> + <listitem> + <para> + Force output of square brackets as array decorations at the beginning + and end of output, and commas between the rows. It is allowed only in + <command>COPY TO</command>, and only when using + <literal>JSON</literal> format. The default is + <literal>false</literal>. + </para> + </listitem> + </varlistentry> + <varlistentry> <term><literal>ON_ERROR</literal></term> <listitem> diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c index a3ee65d000..247ff1fcdc 100644 --- a/src/backend/commands/copy.c +++ b/src/backend/commands/copy.c @@ -470,6 +470,7 @@ ProcessCopyOptions(ParseState *pstate, bool header_specified = false; bool on_error_specified = false; bool log_verbosity_specified = false; + bool force_array_specified = false; ListCell *option; /* Support external use for option sanity checking */ @@ -624,6 +625,13 @@ ProcessCopyOptions(ParseState *pstate, defel->defname), parser_errposition(pstate, defel->location))); } + else if (strcmp(defel->defname, "force_array") == 0) + { + if (force_array_specified) + errorConflictingDefElem(defel, pstate); + force_array_specified = true; + opts_out->force_array = defGetBoolean(defel); + } else if (strcmp(defel->defname, "on_error") == 0) { if (on_error_specified) @@ -850,6 +858,11 @@ ProcessCopyOptions(ParseState *pstate, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), errmsg("COPY json mode cannot be used with %s", "COPY FROM"))); + if (!opts_out->json_mode && opts_out->force_array) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("COPY %s can only used with JSON mode", "FORCE_ARRAY"))); + if (opts_out->default_print) { if (!is_from) diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c index 14ba9fde50..649c425dfc 100644 --- a/src/backend/commands/copyto.c +++ b/src/backend/commands/copyto.c @@ -81,6 +81,7 @@ typedef struct CopyToStateData List *attnumlist; /* integer list of attnums to copy */ char *filename; /* filename, or NULL for STDOUT */ bool is_program; /* is 'filename' a program to popen? */ + bool json_row_delim_needed; /* need delimiter to start next json array element */ copy_data_dest_cb data_dest_cb; /* function for writing data */ CopyFormatOptions opts; @@ -854,6 +855,16 @@ DoCopyTo(CopyToState cstate) CopySendEndOfRow(cstate); } + /* + * If JSON has been requested, and FORCE_ARRAY has been specified send + * the opening bracket. + */ + if (cstate->opts.json_mode && cstate->opts.force_array) + { + CopySendChar(cstate, '['); + CopySendEndOfRow(cstate); + } + } if (cstate->rel) @@ -901,6 +912,15 @@ DoCopyTo(CopyToState cstate) CopySendEndOfRow(cstate); } + /* + * If JSON has been requested, and FORCE_ARRAY has been specified send the + * closing bracket. + */ + if (cstate->opts.json_mode && cstate->opts.force_array) + { + CopySendChar(cstate, ']'); + CopySendEndOfRow(cstate); + } MemoryContextDelete(cstate->rowcontext); if (fe_copy) @@ -1004,6 +1024,15 @@ CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) result = makeStringInfo(); composite_to_json(rowdata, result, false); + if (cstate->json_row_delim_needed && cstate->opts.force_array) + CopySendChar(cstate, ','); + else if (cstate->opts.force_array) + { + /* first row needs no delimiter */ + CopySendChar(cstate, ' '); + cstate->json_row_delim_needed = true; + } + CopySendData(cstate, result->data, result->len); } diff --git a/src/include/commands/copy.h b/src/include/commands/copy.h index ff6ecc7ae7..f76fe8fafe 100644 --- a/src/include/commands/copy.h +++ b/src/include/commands/copy.h @@ -81,6 +81,7 @@ typedef struct CopyFormatOptions List *force_null; /* list of column names */ bool force_null_all; /* FORCE_NULL *? */ bool *force_null_flags; /* per-column CSV FN flags */ + bool force_array; /* add JSON array decorations */ bool convert_selectively; /* do selective binary conversion? */ CopyOnErrorChoice on_error; /* what to do when error happened */ CopyLogVerbosityChoice log_verbosity; /* verbosity of logged messages */ diff --git a/src/test/regress/expected/copy.out b/src/test/regress/expected/copy.out index 77cb36cd61..ed1cb6b28e 100644 --- a/src/test/regress/expected/copy.out +++ b/src/test/regress/expected/copy.out @@ -59,6 +59,29 @@ ERROR: cannot specify HEADER in JSON mode -- Error copy copytest from stdout (format json); ERROR: COPY json mode cannot be used with COPY FROM +--Error +copy copytest to stdout (format csv, force_array true); +ERROR: COPY FORCE_ARRAY can only used with JSON mode +--ok +copy copytest to stdout (format json, force_array); +[ + {"style":"DOS","test":"abc\r\ndef","filler":1} +,{"style":"Unix","test":"abc\ndef","filler":2} +,{"style":"Mac","test":"abc\rdef","filler":3} +,{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +] +copy copytest to stdout (format json, force_array true); +[ + {"style":"DOS","test":"abc\r\ndef","filler":1} +,{"style":"Unix","test":"abc\ndef","filler":2} +,{"style":"Mac","test":"abc\rdef","filler":3} +,{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +] +copy copytest to stdout (format json, force_array false); +{"style":"DOS","test":"abc\r\ndef","filler":1} +{"style":"Unix","test":"abc\ndef","filler":2} +{"style":"Mac","test":"abc\rdef","filler":3} +{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} -- embedded escaped characters create temp table copyjsontest ( id bigserial, diff --git a/src/test/regress/sql/copy.sql b/src/test/regress/sql/copy.sql index 67f77f6512..28f698ecc6 100644 --- a/src/test/regress/sql/copy.sql +++ b/src/test/regress/sql/copy.sql @@ -64,6 +64,15 @@ copy copytest to stdout (format json, header); -- Error copy copytest from stdout (format json); +--Error +copy copytest to stdout (format csv, force_array true); + +--ok +copy copytest to stdout (format json, force_array); + +copy copytest to stdout (format json, force_array true); + +copy copytest to stdout (format json, force_array false); -- embedded escaped characters create temp table copyjsontest ( id bigserial, -- 2.34.1 ^ permalink raw reply [nested|flat] 105+ messages in thread
* Re: Emitting JSON to file using COPY TO @ 2024-08-22 05:19 jian he <[email protected]> parent: jian he <[email protected]> 0 siblings, 0 replies; 105+ messages in thread From: jian he @ 2024-08-22 05:19 UTC (permalink / raw) To: Joe Conway <[email protected]>; +Cc: Andrey M. Borodin <[email protected]>; Dean Rasheed <[email protected]>; Daniel Verite <[email protected]>; Andrew Dunstan <[email protected]>; Davin Shearer <[email protected]>; pgsql-hackers On Mon, Aug 19, 2024 at 8:00 AM jian he <[email protected]> wrote: > > On Mon, Apr 1, 2024 at 8:00 PM jian he <[email protected]> wrote: > > > rebased. > minor cosmetic error message change. > > I think all the issues in this thread have been addressed. hi. I did some minor changes based on the v11. mainly changing some error code from ERRCODE_FEATURE_NOT_SUPPORTED to ERRCODE_INVALID_PARAMETER_VALUE. Attachments: [text/x-patch] v12-0002-Add-option-force_array-for-COPY-TO.patch (8.0K, ../../CACJufxFvOQWcGpiaFkOxMGPc=g85h=-LdyB_2wXBSVQCO9RAHQ@mail.gmail.com/2-v12-0002-Add-option-force_array-for-COPY-TO.patch) download | inline diff: From 24c0ac71f0a3fe1f4eb3d558aef34620234b2b35 Mon Sep 17 00:00:00 2001 From: jian he <[email protected]> Date: Thu, 22 Aug 2024 12:15:52 +0800 Subject: [PATCH v12 2/2] Add option force_array for COPY TO force_array option can only be used in COPY TO with JSON format. it make the output json output behave like json array type. discussion: https://postgr.es/m/CALvfUkBxTYy5uWPFVwpk_7ii2zgT07t3d-yR_cy4sfrrLU%3Dkcg%40mail.gmail.com discussion: https://postgr.es/m/[email protected] --- doc/src/sgml/ref/copy.sgml | 14 ++++++++++++++ src/backend/commands/copy.c | 13 +++++++++++++ src/backend/commands/copyto.c | 28 ++++++++++++++++++++++++++++ src/include/commands/copy.h | 1 + src/test/regress/expected/copy.out | 23 +++++++++++++++++++++++ src/test/regress/sql/copy.sql | 9 +++++++++ 6 files changed, 88 insertions(+) diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml index 616abf508e..4659e49ec3 100644 --- a/doc/src/sgml/ref/copy.sgml +++ b/doc/src/sgml/ref/copy.sgml @@ -43,6 +43,7 @@ COPY { <replaceable class="parameter">table_name</replaceable> [ ( <replaceable FORCE_QUOTE { ( <replaceable class="parameter">column_name</replaceable> [, ...] ) | * } FORCE_NOT_NULL { ( <replaceable class="parameter">column_name</replaceable> [, ...] ) | * } FORCE_NULL { ( <replaceable class="parameter">column_name</replaceable> [, ...] ) | * } + FORCE_ARRAY [ <replaceable class="parameter">boolean</replaceable> ] ON_ERROR <replaceable class="parameter">error_action</replaceable> ENCODING '<replaceable class="parameter">encoding_name</replaceable>' LOG_VERBOSITY <replaceable class="parameter">verbosity</replaceable> @@ -390,6 +391,19 @@ COPY { <replaceable class="parameter">table_name</replaceable> [ ( <replaceable </listitem> </varlistentry> + <varlistentry> + <term><literal>FORCE_ARRAY</literal></term> + <listitem> + <para> + Force output of square brackets as array decorations at the beginning + and end of output, and commas between the rows. It is allowed only in + <command>COPY TO</command>, and only when using + <literal>JSON</literal> format. The default is + <literal>false</literal>. + </para> + </listitem> + </varlistentry> + <varlistentry> <term><literal>ON_ERROR</literal></term> <listitem> diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c index 2f65c96dd3..54793fecab 100644 --- a/src/backend/commands/copy.c +++ b/src/backend/commands/copy.c @@ -470,6 +470,7 @@ ProcessCopyOptions(ParseState *pstate, bool header_specified = false; bool on_error_specified = false; bool log_verbosity_specified = false; + bool force_array_specified = false; ListCell *option; /* Support external use for option sanity checking */ @@ -624,6 +625,13 @@ ProcessCopyOptions(ParseState *pstate, defel->defname), parser_errposition(pstate, defel->location))); } + else if (strcmp(defel->defname, "force_array") == 0) + { + if (force_array_specified) + errorConflictingDefElem(defel, pstate); + force_array_specified = true; + opts_out->force_array = defGetBoolean(defel); + } else if (strcmp(defel->defname, "on_error") == 0) { if (on_error_specified) @@ -850,6 +858,11 @@ ProcessCopyOptions(ParseState *pstate, (errcode(ERRCODE_INVALID_PARAMETER_VALUE), errmsg("COPY json mode cannot be used with %s", "COPY FROM"))); + if (!opts_out->json_mode && opts_out->force_array) + ereport(ERROR, + (errcode(ERRCODE_INVALID_PARAMETER_VALUE), + errmsg("COPY %s can only used with JSON mode", "FORCE_ARRAY"))); + if (opts_out->default_print) { if (!is_from) diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c index 14ba9fde50..b56b78a331 100644 --- a/src/backend/commands/copyto.c +++ b/src/backend/commands/copyto.c @@ -81,6 +81,7 @@ typedef struct CopyToStateData List *attnumlist; /* integer list of attnums to copy */ char *filename; /* filename, or NULL for STDOUT */ bool is_program; /* is 'filename' a program to popen? */ + bool json_row_delim_needed; /* need delimiter to start next json array element */ copy_data_dest_cb data_dest_cb; /* function for writing data */ CopyFormatOptions opts; @@ -854,6 +855,15 @@ DoCopyTo(CopyToState cstate) CopySendEndOfRow(cstate); } + /* + * If JSON has been requested, and FORCE_ARRAY has been specified send + * the opening bracket. + */ + if (cstate->opts.json_mode && cstate->opts.force_array) + { + CopySendChar(cstate, '['); + CopySendEndOfRow(cstate); + } } if (cstate->rel) @@ -901,6 +911,15 @@ DoCopyTo(CopyToState cstate) CopySendEndOfRow(cstate); } + /* + * If JSON has been requested, and FORCE_ARRAY has been specified send the + * closing bracket. + */ + if (cstate->opts.json_mode && cstate->opts.force_array) + { + CopySendChar(cstate, ']'); + CopySendEndOfRow(cstate); + } MemoryContextDelete(cstate->rowcontext); if (fe_copy) @@ -1004,6 +1023,15 @@ CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) result = makeStringInfo(); composite_to_json(rowdata, result, false); + if (cstate->json_row_delim_needed && cstate->opts.force_array) + CopySendChar(cstate, ','); + else if (cstate->opts.force_array) + { + /* first row needs no delimiter */ + CopySendChar(cstate, ' '); + cstate->json_row_delim_needed = true; + } + CopySendData(cstate, result->data, result->len); } diff --git a/src/include/commands/copy.h b/src/include/commands/copy.h index ff6ecc7ae7..f76fe8fafe 100644 --- a/src/include/commands/copy.h +++ b/src/include/commands/copy.h @@ -81,6 +81,7 @@ typedef struct CopyFormatOptions List *force_null; /* list of column names */ bool force_null_all; /* FORCE_NULL *? */ bool *force_null_flags; /* per-column CSV FN flags */ + bool force_array; /* add JSON array decorations */ bool convert_selectively; /* do selective binary conversion? */ CopyOnErrorChoice on_error; /* what to do when error happened */ CopyLogVerbosityChoice log_verbosity; /* verbosity of logged messages */ diff --git a/src/test/regress/expected/copy.out b/src/test/regress/expected/copy.out index 77cb36cd61..ed1cb6b28e 100644 --- a/src/test/regress/expected/copy.out +++ b/src/test/regress/expected/copy.out @@ -59,6 +59,29 @@ ERROR: cannot specify HEADER in JSON mode -- Error copy copytest from stdout (format json); ERROR: COPY json mode cannot be used with COPY FROM +--Error +copy copytest to stdout (format csv, force_array true); +ERROR: COPY FORCE_ARRAY can only used with JSON mode +--ok +copy copytest to stdout (format json, force_array); +[ + {"style":"DOS","test":"abc\r\ndef","filler":1} +,{"style":"Unix","test":"abc\ndef","filler":2} +,{"style":"Mac","test":"abc\rdef","filler":3} +,{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +] +copy copytest to stdout (format json, force_array true); +[ + {"style":"DOS","test":"abc\r\ndef","filler":1} +,{"style":"Unix","test":"abc\ndef","filler":2} +,{"style":"Mac","test":"abc\rdef","filler":3} +,{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +] +copy copytest to stdout (format json, force_array false); +{"style":"DOS","test":"abc\r\ndef","filler":1} +{"style":"Unix","test":"abc\ndef","filler":2} +{"style":"Mac","test":"abc\rdef","filler":3} +{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} -- embedded escaped characters create temp table copyjsontest ( id bigserial, diff --git a/src/test/regress/sql/copy.sql b/src/test/regress/sql/copy.sql index 67f77f6512..28f698ecc6 100644 --- a/src/test/regress/sql/copy.sql +++ b/src/test/regress/sql/copy.sql @@ -64,6 +64,15 @@ copy copytest to stdout (format json, header); -- Error copy copytest from stdout (format json); +--Error +copy copytest to stdout (format csv, force_array true); + +--ok +copy copytest to stdout (format json, force_array); + +copy copytest to stdout (format json, force_array true); + +copy copytest to stdout (format json, force_array false); -- embedded escaped characters create temp table copyjsontest ( id bigserial, -- 2.34.1 [text/x-patch] v12-0001-introduce-json-format-for-COPY-TO.patch (12.0K, ../../CACJufxFvOQWcGpiaFkOxMGPc=g85h=-LdyB_2wXBSVQCO9RAHQ@mail.gmail.com/3-v12-0001-introduce-json-format-for-COPY-TO.patch) download | inline diff: From 51f0322ad7ce7159cc46f80380f98d604560a001 Mon Sep 17 00:00:00 2001 From: jian he <[email protected]> Date: Thu, 22 Aug 2024 11:58:12 +0800 Subject: [PATCH v12 1/2] introduce json format for COPY TO json format is only allowed in COPY TO operation. also cannot be used with header option. discussion: https://postgr.es/m/CALvfUkBxTYy5uWPFVwpk_7ii2zgT07t3d-yR_cy4sfrrLU%3Dkcg%40mail.gmail.com discussion: https://postgr.es/m/[email protected] --- doc/src/sgml/ref/copy.sgml | 5 +++ src/backend/commands/copy.c | 13 +++++++ src/backend/commands/copyto.c | 51 +++++++++++++++++++++++++--- src/backend/parser/gram.y | 8 +++++ src/backend/utils/adt/json.c | 5 ++- src/include/commands/copy.h | 1 + src/include/utils/json.h | 2 ++ src/test/regress/expected/copy.out | 54 ++++++++++++++++++++++++++++++ src/test/regress/sql/copy.sql | 38 +++++++++++++++++++++ 9 files changed, 169 insertions(+), 8 deletions(-) diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml index 1518af8a04..616abf508e 100644 --- a/doc/src/sgml/ref/copy.sgml +++ b/doc/src/sgml/ref/copy.sgml @@ -218,9 +218,14 @@ COPY { <replaceable class="parameter">table_name</replaceable> [ ( <replaceable Selects the data format to be read or written: <literal>text</literal>, <literal>csv</literal> (Comma Separated Values), + <literal>json</literal> (JavaScript Object Notation), or <literal>binary</literal>. The default is <literal>text</literal>. </para> + <para> + The <literal>json</literal> option is allowed only in + <command>COPY TO</command>. + </para> </listitem> </varlistentry> diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c index 3bb579a3a4..2f65c96dd3 100644 --- a/src/backend/commands/copy.c +++ b/src/backend/commands/copy.c @@ -494,6 +494,8 @@ ProcessCopyOptions(ParseState *pstate, /* default format */ ; else if (strcmp(fmt, "csv") == 0) opts_out->csv_mode = true; + else if (strcmp(fmt, "json") == 0) + opts_out->json_mode = true; else if (strcmp(fmt, "binary") == 0) opts_out->binary = true; else @@ -739,6 +741,11 @@ ProcessCopyOptions(ParseState *pstate, /*- translator: %s is the name of a COPY option, e.g. ON_ERROR */ errmsg("cannot specify %s in BINARY mode", "HEADER"))); + if (opts_out->json_mode && opts_out->header_line) + ereport(ERROR, + (errcode(ERRCODE_INVALID_PARAMETER_VALUE), + errmsg("cannot specify %s in JSON mode", "HEADER"))); + /* Check quote */ if (!opts_out->csv_mode && opts_out->quote != NULL) ereport(ERROR, @@ -837,6 +844,12 @@ ProcessCopyOptions(ParseState *pstate, errmsg("COPY %s cannot be used with %s", "FREEZE", "COPY TO"))); + /* Check json format */ + if (opts_out->json_mode && is_from) + ereport(ERROR, + (errcode(ERRCODE_INVALID_PARAMETER_VALUE), + errmsg("COPY json mode cannot be used with %s", "COPY FROM"))); + if (opts_out->default_print) { if (!is_from) diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c index 91de442f43..14ba9fde50 100644 --- a/src/backend/commands/copyto.c +++ b/src/backend/commands/copyto.c @@ -24,6 +24,7 @@ #include "executor/execdesc.h" #include "executor/executor.h" #include "executor/tuptable.h" +#include "funcapi.h" #include "libpq/libpq.h" #include "libpq/pqformat.h" #include "mb/pg_wchar.h" @@ -31,6 +32,7 @@ #include "pgstat.h" #include "storage/fd.h" #include "tcop/tcopprot.h" +#include "utils/json.h" #include "utils/lsyscache.h" #include "utils/memutils.h" #include "utils/rel.h" @@ -139,9 +141,20 @@ SendCopyBegin(CopyToState cstate) pq_beginmessage(&buf, PqMsg_CopyOutResponse); pq_sendbyte(&buf, format); /* overall format */ - pq_sendint16(&buf, natts); - for (i = 0; i < natts; i++) - pq_sendint16(&buf, format); /* per-column formats */ + if (!cstate->opts.json_mode) + { + pq_sendint16(&buf, natts); + for (i = 0; i < natts; i++) + pq_sendint16(&buf, format); /* per-column formats */ + } + else + { + /* + * JSON mode is always one non-binary column + */ + pq_sendint16(&buf, 1); + pq_sendint16(&buf, 0); + } pq_endmessage(&buf); cstate->copy_dest = COPY_FRONTEND; } @@ -917,7 +930,7 @@ CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) /* Make sure the tuple is fully deconstructed */ slot_getallattrs(slot); - if (!cstate->opts.binary) + if (!cstate->opts.binary && !cstate->opts.json_mode) { bool need_delim = false; @@ -945,7 +958,7 @@ CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) } } } - else + else if (!cstate->opts.json_mode) { foreach_int(attnum, cstate->attnumlist) { @@ -965,6 +978,34 @@ CopyOneRowTo(CopyToState cstate, TupleTableSlot *slot) } } } + else + { + Datum rowdata; + StringInfo result; + + /* + * if COPY TO source data is from a query, not a table, then we need + * copy CopyToState->TupleDesc->attrs to + * slot->tts_tupleDescriptor->attrs because the slot's TupleDesc->attrs + * may change during query execution, but composite_to_json requires + * correct TupleDesc->attrs for constructing the json keys. + * composite_to_json will iterate each TupleDesc->attrs so no need to + * copy other filed in cstate->queryDesc->tupDesc. + */ + if(!cstate->rel) + { + memcpy(TupleDescAttr(slot->tts_tupleDescriptor, 0), + TupleDescAttr(cstate->queryDesc->tupDesc, 0), + cstate->queryDesc->tupDesc->natts * sizeof(FormData_pg_attribute)); + + BlessTupleDesc(slot->tts_tupleDescriptor); + } + rowdata = ExecFetchSlotHeapTupleDatum(slot); + result = makeStringInfo(); + composite_to_json(rowdata, result, false); + + CopySendData(cstate, result->data, result->len); + } CopySendEndOfRow(cstate); diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y index c3f25582c3..76bd6f8949 100644 --- a/src/backend/parser/gram.y +++ b/src/backend/parser/gram.y @@ -3500,6 +3500,10 @@ copy_opt_item: { $$ = makeDefElem("format", (Node *) makeString("csv"), @1); } + | JSON + { + $$ = makeDefElem("format", (Node *) makeString("json"), @1); + } | HEADER_P { $$ = makeDefElem("header", (Node *) makeBoolean(true), @1); @@ -3582,6 +3586,10 @@ copy_generic_opt_elem: { $$ = makeDefElem($1, $2, @1); } + | FORMAT_LA copy_generic_opt_arg + { + $$ = makeDefElem("format", $2, @1); + } ; copy_generic_opt_arg: diff --git a/src/backend/utils/adt/json.c b/src/backend/utils/adt/json.c index 4eeeeaf0a6..b47ae14f20 100644 --- a/src/backend/utils/adt/json.c +++ b/src/backend/utils/adt/json.c @@ -85,8 +85,6 @@ typedef struct JsonAggState JsonUniqueBuilderState unique_check; } JsonAggState; -static void composite_to_json(Datum composite, StringInfo result, - bool use_line_feeds); static void array_dim_to_json(StringInfo result, int dim, int ndims, int *dims, Datum *vals, bool *nulls, int *valcount, JsonTypeCategory tcategory, Oid outfuncoid, @@ -516,8 +514,9 @@ array_to_json_internal(Datum array, StringInfo result, bool use_line_feeds) /* * Turn a composite / record into JSON. + * Exported so COPY TO can use it. */ -static void +void composite_to_json(Datum composite, StringInfo result, bool use_line_feeds) { HeapTupleHeader td; diff --git a/src/include/commands/copy.h b/src/include/commands/copy.h index 141fd48dc1..ff6ecc7ae7 100644 --- a/src/include/commands/copy.h +++ b/src/include/commands/copy.h @@ -62,6 +62,7 @@ typedef struct CopyFormatOptions bool binary; /* binary format? */ bool freeze; /* freeze rows on loading? */ bool csv_mode; /* Comma Separated Value format? */ + bool json_mode; /* JSON format? */ CopyHeaderChoice header_line; /* header line? */ char *null_print; /* NULL marker string (server encoding!) */ int null_print_len; /* length of same */ diff --git a/src/include/utils/json.h b/src/include/utils/json.h index 79c1062e1b..c904ef6c6e 100644 --- a/src/include/utils/json.h +++ b/src/include/utils/json.h @@ -17,6 +17,8 @@ #include "lib/stringinfo.h" /* functions in json.c */ +extern void composite_to_json(Datum composite, StringInfo result, + bool use_line_feeds); extern void escape_json(StringInfo buf, const char *str); extern void escape_json_with_len(StringInfo buf, const char *str, int len); extern void escape_json_text(StringInfo buf, const text *txt); diff --git a/src/test/regress/expected/copy.out b/src/test/regress/expected/copy.out index 44114089a6..77cb36cd61 100644 --- a/src/test/regress/expected/copy.out +++ b/src/test/regress/expected/copy.out @@ -42,6 +42,60 @@ copy copytest3 to stdout csv header; c1,"col with , comma","col with "" quote" 1,a,1 2,b,2 +--- test copying in JSON mode with various styles +copy copytest to stdout json; +{"style":"DOS","test":"abc\r\ndef","filler":1} +{"style":"Unix","test":"abc\ndef","filler":2} +{"style":"Mac","test":"abc\rdef","filler":3} +{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +copy copytest to stdout (format json); +{"style":"DOS","test":"abc\r\ndef","filler":1} +{"style":"Unix","test":"abc\ndef","filler":2} +{"style":"Mac","test":"abc\rdef","filler":3} +{"style":"esc\\ape","test":"a\\r\\\r\\\n\\nb","filler":4} +-- Error +copy copytest to stdout (format json, header); +ERROR: cannot specify HEADER in JSON mode +-- Error +copy copytest from stdout (format json); +ERROR: COPY json mode cannot be used with COPY FROM +-- embedded escaped characters +create temp table copyjsontest ( + id bigserial, + f1 text, + f2 timestamptz); +insert into copyjsontest + select g.i, + CASE WHEN g.i % 2 = 0 THEN + 'line with '' in it: ' || g.i::text + ELSE + 'line with " in it: ' || g.i::text + END, + 'Mon Feb 10 17:32:01 1997 PST' + from generate_series(1,5) as g(i); +insert into copyjsontest (f1) values +(E'aaa\"bbb'::text), +(E'aaa\\bbb'::text), +(E'aaa\/bbb'::text), +(E'aaa\bbbb'::text), +(E'aaa\fbbb'::text), +(E'aaa\nbbb'::text), +(E'aaa\rbbb'::text), +(E'aaa\tbbb'::text); +copy copyjsontest to stdout json; +{"id":1,"f1":"line with \" in it: 1","f2":"1997-02-10T17:32:01-08:00"} +{"id":2,"f1":"line with ' in it: 2","f2":"1997-02-10T17:32:01-08:00"} +{"id":3,"f1":"line with \" in it: 3","f2":"1997-02-10T17:32:01-08:00"} +{"id":4,"f1":"line with ' in it: 4","f2":"1997-02-10T17:32:01-08:00"} +{"id":5,"f1":"line with \" in it: 5","f2":"1997-02-10T17:32:01-08:00"} +{"id":1,"f1":"aaa\"bbb","f2":null} +{"id":2,"f1":"aaa\\bbb","f2":null} +{"id":3,"f1":"aaa/bbb","f2":null} +{"id":4,"f1":"aaa\bbbb","f2":null} +{"id":5,"f1":"aaa\fbbb","f2":null} +{"id":6,"f1":"aaa\nbbb","f2":null} +{"id":7,"f1":"aaa\rbbb","f2":null} +{"id":8,"f1":"aaa\tbbb","f2":null} create temp table copytest4 ( c1 int, "colname with tab: " text); diff --git a/src/test/regress/sql/copy.sql b/src/test/regress/sql/copy.sql index e2dd24cb35..67f77f6512 100644 --- a/src/test/regress/sql/copy.sql +++ b/src/test/regress/sql/copy.sql @@ -54,6 +54,44 @@ this is just a line full of junk that would error out if parsed copy copytest3 to stdout csv header; +--- test copying in JSON mode with various styles +copy copytest to stdout json; + +copy copytest to stdout (format json); + +-- Error +copy copytest to stdout (format json, header); +-- Error +copy copytest from stdout (format json); + +-- embedded escaped characters +create temp table copyjsontest ( + id bigserial, + f1 text, + f2 timestamptz); + +insert into copyjsontest + select g.i, + CASE WHEN g.i % 2 = 0 THEN + 'line with '' in it: ' || g.i::text + ELSE + 'line with " in it: ' || g.i::text + END, + 'Mon Feb 10 17:32:01 1997 PST' + from generate_series(1,5) as g(i); + +insert into copyjsontest (f1) values +(E'aaa\"bbb'::text), +(E'aaa\\bbb'::text), +(E'aaa\/bbb'::text), +(E'aaa\bbbb'::text), +(E'aaa\fbbb'::text), +(E'aaa\nbbb'::text), +(E'aaa\rbbb'::text), +(E'aaa\tbbb'::text); + +copy copyjsontest to stdout json; + create temp table copytest4 ( c1 int, "colname with tab: " text); base-commit: 9bb842f95ef3384f0822c386a4c569780e613e4e -- 2.34.1 ^ permalink raw reply [nested|flat] 105+ messages in thread
end of thread, other threads:[~2024-08-22 05:19 UTC | newest] Thread overview: 105+ messages (download: mbox mbox.gz follow: Atom feed) -- links below jump to the message on this page -- 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/12] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH v9 1/9] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH v7 1/9] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/12] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH v8 1/9] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 3/8] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2021-02-27 04:03 [PATCH 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]> 2024-03-09 01:13 Re: Emitting JSON to file using COPY TO jian he <[email protected]> 2024-04-01 12:00 ` Re: Emitting JSON to file using COPY TO jian he <[email protected]> 2024-08-19 00:00 ` Re: Emitting JSON to file using COPY TO jian he <[email protected]> 2024-08-22 05:19 ` Re: Emitting JSON to file using COPY TO jian he <[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