From 4035cd5d4eee4dae797bfc77ab07f8dcd8781b41 Mon Sep 17 00:00:00 2001 From: Michael Paquier Date: Tue, 29 Jun 2021 11:17:55 +0900 Subject: [PATCH] Add support for LZ4 with compression of full-page writes in WAL The logic is implemented so as there can be a choice in the compression used when building a WAL record, and an extra per-record bit is used to track down if a block is compressed with PGLZ, LZ4 or nothing. wal_compression, the existing parameter, is changed to an enum with support for the following backward-compatible values: - "off", the default, to not use compression. - "pglz" or "on", to compress FPWs with PGLZ. - "lz4", the new mode, to compress FPWs with LZ4. Benchmarking has showed that LZ4 outclasses easily PGLZ. ZSTD would be also an interesting choice, but going just with LZ4 for now makes the patch minimalistic as toast compression is already able to use LZ4, so there is no need to worry about any build-related needs for this implementation. Author: Andrey Borodin, Justin Pryzby Reviewed-by: Dilip Kumar, Michael Paquier Discussion: https://postgr.es/m/3037310D-ECB7-4BF1-AF20-01C10BB33A33@yandex-team.ru --- doc/src/sgml/config.sgml | 14 ++-- doc/src/sgml/install-windows.sgml | 2 +- doc/src/sgml/installation.sgml | 5 +- doc/src/sgml/standalone-profile.xsl | 4 + src/backend/access/transam/xlog.c | 2 +- src/backend/access/transam/xloginsert.c | 73 ++++++++++++++++--- src/backend/access/transam/xlogreader.c | 58 ++++++++++++--- src/backend/utils/misc/guc.c | 36 ++++++--- src/backend/utils/misc/postgresql.conf.sample | 3 +- src/bin/pg_waldump/pg_waldump.c | 19 ++++- src/include/access/xlog.h | 10 ++- src/include/access/xlogrecord.h | 16 ++-- src/tools/pgindent/typedefs.list | 1 + 13 files changed, 191 insertions(+), 52 deletions(-) diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml index 3eee988359..6098f6b020 100644 --- a/doc/src/sgml/config.sgml +++ b/doc/src/sgml/config.sgml @@ -3128,23 +3128,27 @@ include_dir 'conf.d' - wal_compression (boolean) + wal_compression (enum) wal_compression configuration parameter - When this parameter is on, the PostgreSQL + This parameter enables compression of WAL using the specified + compression method. + When enabled, the PostgreSQL server compresses full page images written to WAL when is on or during a base backup. A compressed page image will be decompressed during WAL replay. - The default value is off. - Only superusers can change this setting. + The supported methods are pglz and + lz4 (if PostgreSQL was + compiled with ). The default value is + off. Only superusers can change this setting. - 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/install-windows.sgml b/doc/src/sgml/install-windows.sgml index 312edc6f7a..ba794b8c93 100644 --- a/doc/src/sgml/install-windows.sgml +++ b/doc/src/sgml/install-windows.sgml @@ -299,7 +299,7 @@ $ENV{MSBFLAGS}="/m"; LZ4 Required for supporting LZ4 compression - method for compressing the table data. Binaries and source can be + method for compressing table or WAL data. Binaries and source can be downloaded from . diff --git a/doc/src/sgml/installation.sgml b/doc/src/sgml/installation.sgml index 3c0aa118c7..61d0bc8c43 100644 --- a/doc/src/sgml/installation.sgml +++ b/doc/src/sgml/installation.sgml @@ -270,7 +270,8 @@ su - postgres You need LZ4, if you want to support compression of data with this method; see - . + and + . @@ -980,7 +981,7 @@ build-postgresql: Build with LZ4 compression support. This allows the use of LZ4 for - compression of table data. + compression of table and WAL data. diff --git a/doc/src/sgml/standalone-profile.xsl b/doc/src/sgml/standalone-profile.xsl index 8bdf58632c..d748076a05 100644 --- a/doc/src/sgml/standalone-profile.xsl +++ b/doc/src/sgml/standalone-profile.xsl @@ -52,6 +52,10 @@ variant without links and references to the main documentation. the configuration parameter default_toast_compression + + the configuration parameter wal_compression + + the documentation diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 2c6e21bea5..9cbca6392d 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..10b3b09053 100644 --- a/src/backend/access/transam/xloginsert.c +++ b/src/backend/access/transam/xloginsert.c @@ -33,8 +33,20 @@ #include "storage/proc.h" #include "utils/memutils.h" -/* Buffer size required to store a compressed version of backup block image */ -#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) +/* + * Guess the maximum buffer size required to store a compressed version of + * backup block image. + */ +#ifdef USE_LZ4 +#include +#define LZ4_MAX_BLCKSZ LZ4_COMPRESSBOUND(BLCKSZ) +#else +#define LZ4_MAX_BLCKSZ 0 +#endif + +#define PGLZ_MAX_BLCKSZ PGLZ_MAX_OUTPUT(BLCKSZ) + +#define COMPRESS_BUFSIZE Max(PGLZ_MAX_BLCKSZ, LZ4_MAX_BLCKSZ) /* * For each block reference registered with XLogRegisterBuffer, we fill in @@ -58,7 +70,7 @@ 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; @@ -628,7 +640,7 @@ 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, @@ -665,8 +677,29 @@ XLogRecordAssemble(RmgrId rmid, uint8 info, if (is_compressed) { + /* The current compression is stored in the WAL record */ bimg.length = compressed_len; - bimg.bimg_info |= BKPIMAGE_IS_COMPRESSED; + + /* Set the compression method used for this block */ + switch ((WalCompression) wal_compression) + { + case WAL_COMPRESSION_PGLZ: + bimg.bimg_info |= BKPIMAGE_COMPRESS_PGLZ; + break; + + case WAL_COMPRESSION_LZ4: +#ifdef USE_LZ4 + bimg.bimg_info |= BKPIMAGE_COMPRESS_LZ4; +#else + elog(ERROR, "LZ4 is not supported by this build"); +#endif + break; + + case WAL_COMPRESSION_NONE: + Assert(false); /* cannot happen */ + break; + /* no default case, so that compiler will warn */ + } rdt_datas_last->data = regbuf->compressed_page; rdt_datas_last->len = compressed_len; @@ -853,12 +886,34 @@ XLogCompressBackupBlock(char *page, uint16 hole_offset, uint16 hole_length, else source = page; + switch ((WalCompression) wal_compression) + { + case WAL_COMPRESSION_PGLZ: + len = pglz_compress(source, orig_len, dest, PGLZ_strategy_default); + break; + + case WAL_COMPRESSION_LZ4: +#ifdef USE_LZ4 + len = LZ4_compress_default(source, dest, orig_len, + COMPRESS_BUFSIZE); + if (len <= 0) + len = -1; /* failure */ +#else + elog(ERROR, "LZ4 is not supported by this build"); +#endif + break; + + case WAL_COMPRESSION_NONE: + Assert(false); /* cannot happen */ + break; + /* no default case, so that compiler will warn */ + } + /* - * We recheck the actual size even if pglz_compress() 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. + * 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..9a2cdf888e 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -18,6 +18,9 @@ #include "postgres.h" #include +#ifdef USE_LZ4 +#include +#endif #include "access/transam.h" #include "access/xlog_internal.h" @@ -1290,7 +1293,7 @@ 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_COMPRESSED(blk->bimg_info)) { if (blk->bimg_info & BKPIMAGE_HAS_HOLE) COPY_HEADER_FIELD(&blk->hole_length, sizeof(uint16)); @@ -1335,29 +1338,28 @@ 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 is compressed. */ - if ((blk->bimg_info & BKPIMAGE_IS_COMPRESSED) && + if (BKPIMAGE_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_COMPRESSED set, 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 COMPRESSED(). */ if (!(blk->bimg_info & BKPIMAGE_HAS_HOLE) && - !(blk->bimg_info & BKPIMAGE_IS_COMPRESSED) && + !BKPIMAGE_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_COMPRESSED set, but block image length is %u at %X/%X", (unsigned int) blk->data_len, LSN_FORMAT_ARGS(state->ReadRecPtr)); goto err; @@ -1555,17 +1557,49 @@ 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_COMPRESSED(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) + bool decomp_success = true; + + if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_PGLZ) != 0) + { + if (pglz_decompress(ptr, bkpb->bimg_len, tmp.data, + BLCKSZ - bkpb->hole_length, true) < 0) + decomp_success = false; + } + else if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_LZ4) != 0) + { +#ifdef USE_LZ4 + if (LZ4_decompress_safe(ptr, tmp.data, + bkpb->bimg_len, BLCKSZ - bkpb->hole_length) <= 0) + decomp_success = false; +#else + report_invalid_record(record, "image at %X/%X compressed with %s not supported by build, block %d", + (uint32) (record->ReadRecPtr >> 32), + (uint32) record->ReadRecPtr, + "LZ4", + block_id); + return false; +#endif + } + else + { + report_invalid_record(record, "image at %X/%X compressed with unknown method, block %d", + (uint32) (record->ReadRecPtr >> 32), + (uint32) record->ReadRecPtr, + block_id); + return false; + } + + if (!decomp_success) { report_invalid_record(record, "invalid compressed image at %X/%X, block %d", LSN_FORMAT_ARGS(record->ReadRecPtr), block_id); return false; } + ptr = tmp.data; } diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 297e705b80..480e8cd199 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -540,6 +540,22 @@ static struct config_enum_entry default_toast_compression_options[] = { {NULL, 0, false} }; +static const struct config_enum_entry wal_compression_options[] = { + {"pglz", WAL_COMPRESSION_PGLZ, false}, +#ifdef USE_LZ4 + {"lz4", WAL_COMPRESSION_LZ4, false}, +#endif + {"on", WAL_COMPRESSION_PGLZ, false}, + {"off", WAL_COMPRESSION_NONE, false}, + {"true", WAL_COMPRESSION_PGLZ, true}, + {"false", WAL_COMPRESSION_NONE, true}, + {"yes", WAL_COMPRESSION_PGLZ, true}, + {"no", WAL_COMPRESSION_NONE, true}, + {"1", WAL_COMPRESSION_PGLZ, true}, + {"0", WAL_COMPRESSION_NONE, true}, + {NULL, 0, false} +}; + /* * Options for enum values stored in other modules */ @@ -1304,16 +1320,6 @@ 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."), @@ -4816,6 +4822,16 @@ static struct config_enum ConfigureNamesEnum[] = NULL, NULL, NULL }, + { + {"wal_compression", PGC_SUSET, WAL_SETTINGS, + gettext_noop("Compresses full-page writes written in WAL file with specified method."), + NULL + }, + &wal_compression, + WAL_COMPRESSION_NONE, wal_compression_options, + NULL, NULL, NULL + }, + { {"wal_level", PGC_POSTMASTER, WAL_SETTINGS, gettext_noop("Sets the level of information written to the WAL."), diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample index af04ec3c74..b696abfe54 100644 --- a/src/backend/utils/misc/postgresql.conf.sample +++ b/src/backend/utils/misc/postgresql.conf.sample @@ -218,7 +218,8 @@ #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 # enables compression of full-page writes; + # off, pglz, lz4, or on #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..d83847b276 100644 --- a/src/bin/pg_waldump/pg_waldump.c +++ b/src/bin/pg_waldump/pg_waldump.c @@ -537,18 +537,29 @@ XLogDumpDisplayRecord(XLogDumpConfig *config, XLogReaderState *record) blk); if (XLogRecHasBlockImage(record, block_id)) { - if (record->blocks[block_id].bimg_info & - BKPIMAGE_IS_COMPRESSED) + uint8 bimg_info = record->blocks[block_id].bimg_info; + + if (BKPIMAGE_COMPRESSED(bimg_info)) { + const char *method; + + if ((bimg_info & BKPIMAGE_COMPRESS_PGLZ) != 0) + method = "pglz"; + else if ((bimg_info & BKPIMAGE_COMPRESS_LZ4) != 0) + method = "lz4"; + else + method = "unknown"; + printf(" (FPW%s); hole: offset: %u, length: %u, " - "compression saved: %u", + "compression saved: %u, method: %s", XLogRecBlockImageApply(record, block_id) ? "" : " for WAL verification", record->blocks[block_id].hole_offset, record->blocks[block_id].hole_length, BLCKSZ - record->blocks[block_id].hole_length - - record->blocks[block_id].bimg_len); + record->blocks[block_id].bimg_len, + method); } else { diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h index 7510e88228..ccfcf43d62 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; @@ -167,6 +167,14 @@ typedef enum WalLevel WAL_LEVEL_LOGICAL } WalLevel; +/* Compression algorithms for WAL */ +typedef enum WalCompression +{ + WAL_COMPRESSION_NONE = 0, + WAL_COMPRESSION_PGLZ, + WAL_COMPRESSION_LZ4 +} WalCompression; + /* Recovery states */ typedef enum RecoveryState { diff --git a/src/include/access/xlogrecord.h b/src/include/access/xlogrecord.h index 80c92a2498..e06ee92a5e 100644 --- a/src/include/access/xlogrecord.h +++ b/src/include/access/xlogrecord.h @@ -114,8 +114,8 @@ typedef struct XLogRecordBlockHeader * present is (BLCKSZ - ). * * Additionally, when wal_compression is enabled, we will try to compress full - * page images using the PGLZ compression algorithm, after removing the "hole". - * This can reduce the WAL volume, but at some extra cost of CPU spent + * page images using one of the supported algorithms, 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 * from BLCKSZ, basically it needs to be stored as an extra information. @@ -134,7 +134,7 @@ typedef struct XLogRecordBlockImageHeader uint8 bimg_info; /* flag bits, see below */ /* - * If BKPIMAGE_HAS_HOLE and BKPIMAGE_IS_COMPRESSED, an + * If BKPIMAGE_HAS_HOLE and BKPIMAGE_COMPRESSED(), an * XLogRecordBlockCompressHeader struct follows. */ } XLogRecordBlockImageHeader; @@ -144,9 +144,13 @@ 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 - * replay */ +#define BKPIMAGE_APPLY 0x02 /* page image should be restored + * during replay */ +/* compression methods supported */ +#define BKPIMAGE_COMPRESS_PGLZ 0x04 +#define BKPIMAGE_COMPRESS_LZ4 0x08 +#define BKPIMAGE_COMPRESSED(info) \ + ((info & (BKPIMAGE_COMPRESS_PGLZ | BKPIMAGE_COMPRESS_LZ4)) != 0) /* * Extra header information used when page image has "hole" and diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 1b3da85421..64c06cf952 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -2843,6 +2843,7 @@ WaitEventSet WaitEventTimeout WaitPMResult WalCloseMethod +WalCompression WalLevel WalRcvData WalRcvExecResult