public inbox for [email protected]
help / color / mirror / Atom feed[PATCH 01/12] Allow alternate compression methods for wal_compression
105+ messages / 4 participants
[nested] [flat]
* [PATCH 01/10] Allow alternate compression methods for wal_compression
@ 2021-02-27 04:03 Andrey Borodin <[email protected]>
0 siblings, 0 replies; 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 01/10] 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 01/10] 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 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 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 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 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 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 | 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 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 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 01/10] 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 01/10] 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 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 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 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 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 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 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 | 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 01/10] 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 01/10] 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 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 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 | 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 01/10] 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 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 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 01/12] 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 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 | 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 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 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 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 | 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 | 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 | 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 01/10] 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 01/10] 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
* [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 01/10] 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 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 | 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 01/10] 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 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 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 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 01/10] 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 01/10] 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 01/10] 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 01/10] 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 | 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 | 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 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 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 | 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 01/10] 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 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 | 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 01/10] 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 | 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 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 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 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 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 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 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 | 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 | 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 | 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 | 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 01/10] 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 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 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 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 | 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 01/10] 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 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 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 01/10] 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 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
* Re: a wrong index choose when statistics is out of date
@ 2024-03-04 05:33 David Rowley <[email protected]>
2024-03-04 11:20 ` Re: a wrong index choose when statistics is out of date Andy Fan <[email protected]>
2024-03-04 17:03 ` Re: a wrong index choose when statistics is out of date Tomas Vondra <[email protected]>
0 siblings, 2 replies; 105+ messages in thread
From: David Rowley @ 2024-03-04 05:33 UTC (permalink / raw)
To: Andy Fan <[email protected]>; +Cc: pgsql-hackers
On Sun, 3 Mar 2024 at 20:08, Andy Fan <[email protected]> wrote:
> The issue can be reproduced with the following steps:
>
> create table x_events (.., created_at timestamp, a int, b int);
>
> create index idx_1 on t(created_at, a);
> create index idx_2 on t(created_at, b);
>
> query:
> select * from t where create_at = current_timestamp and b = 1;
>
> index (created_at, a) rather than (created_at, b) may be chosen for the
> above query if the statistics think "create_at = current_timestamp" has
> no rows, then both index are OK, actually it is true just because
> statistics is out of date.
I don't think there's really anything too special about the fact that
the created_at column is always increasing. We commonly get 1-row
estimates after multiplying the selectivities from individual stats.
Your example just seems like yet another reason that this could
happen.
I've been periodically talking about introducing "risk" as a factor
that the planner should consider. I did provide some detail in [1]
about the design that was in my head at that time. I'd not previously
thought that it could also solve this problem, but after reading your
email, I think it can.
I don't think it would be right to fudge the costs in any way, but I
think the risk factor for IndexPaths could take into account the
number of unmatched index clauses and increment the risk factor, or
"certainty_factor" as it is currently in my brain-based design. That
way add_path() would be more likely to prefer the index that matches
the most conditions.
The exact maths to calculate the certainty_factor for this case I
don't quite have worked out yet. I plan to work on documenting the
design of this and try and get a prototype patch out sometime during
this coming southern hemisphere winter so that there's at least a full
cycle of feedback opportunity before the PG18 freeze.
We should do anything like add column options in the meantime. Those
are hard to remove once added.
David
[1] https://www.postgresql.org/message-id/CAApHDvo2sMPF9m%3Di%2BYPPUssfTV1GB%3DZ8nMVa%2B9Uq4RZJ8sULeQ%40...
^ permalink raw reply [nested|flat] 105+ messages in thread
* Re: a wrong index choose when statistics is out of date
2024-03-04 05:33 Re: a wrong index choose when statistics is out of date David Rowley <[email protected]>
@ 2024-03-04 11:20 ` Andy Fan <[email protected]>
2024-03-04 23:13 ` Re: a wrong index choose when statistics is out of date David Rowley <[email protected]>
1 sibling, 1 reply; 105+ messages in thread
From: Andy Fan @ 2024-03-04 11:20 UTC (permalink / raw)
To: David Rowley <[email protected]>; +Cc: pgsql-hackers
David Rowley <[email protected]> writes:
> On Sun, 3 Mar 2024 at 20:08, Andy Fan <[email protected]> wrote:
>> The issue can be reproduced with the following steps:
>>
>> create table x_events (.., created_at timestamp, a int, b int);
>>
>> create index idx_1 on t(created_at, a);
>> create index idx_2 on t(created_at, b);
>>
>> query:
>> select * from t where create_at = current_timestamp and b = 1;
>>
>> index (created_at, a) rather than (created_at, b) may be chosen for the
>> above query if the statistics think "create_at = current_timestamp" has
>> no rows, then both index are OK, actually it is true just because
>> statistics is out of date.
>
> I don't think there's really anything too special about the fact that
> the created_at column is always increasing. We commonly get 1-row
> estimates after multiplying the selectivities from individual stats.
> Your example just seems like yet another reason that this could
> happen.
You are right about there are more cases which lead this happen. However
this is the only case where the created_at = $1 trick can works, which
was the problem I wanted to resove when I was writing.
> I've been periodically talking about introducing "risk" as a factor
> that the planner should consider. I did provide some detail in [1]
> about the design that was in my head at that time. I'd not previously
> thought that it could also solve this problem, but after reading your
> email, I think it can.
Haha, I remeber you were against "risk factor" before at [1], and at
that time we are talking about the exact same topic as here, and I
proposaled another risk factor. Without an agreement, I did it in my
own internal version and get hurted then, something like I didn't pay
enough attention to Bitmap Index Scan and Index scan. Then I forget the
"risk factor".
>
> I don't think it would be right to fudge the costs in any way, but I
> think the risk factor for IndexPaths could take into account the
> number of unmatched index clauses and increment the risk factor, or
> "certainty_factor" as it is currently in my brain-based design. That
> way add_path() would be more likely to prefer the index that matches
> the most conditions.
This is somehow similar with my proposal at [1]? What do you think
about the treat 'col op const' as 'col op $1' for the marked column?
This could just resolve a subset of questions in your mind, but the
method looks have a solid reason.
Currently I treat the risk factor as what you did before, but this maybe
another time for me to switch my mind again.
[1] https://www.postgresql.org/message-id/CAApHDvovVWCbeR4v%2BA4Dkwb%3DYS_GuJG9OyCm8jZu%2B%2BcP2xsY_A%40...
--
Best Regards
Andy Fan
^ permalink raw reply [nested|flat] 105+ messages in thread
* Re: a wrong index choose when statistics is out of date
2024-03-04 05:33 Re: a wrong index choose when statistics is out of date David Rowley <[email protected]>
2024-03-04 11:20 ` Re: a wrong index choose when statistics is out of date Andy Fan <[email protected]>
@ 2024-03-04 23:13 ` David Rowley <[email protected]>
0 siblings, 0 replies; 105+ messages in thread
From: David Rowley @ 2024-03-04 23:13 UTC (permalink / raw)
To: Andy Fan <[email protected]>; +Cc: pgsql-hackers
On Tue, 5 Mar 2024 at 00:37, Andy Fan <[email protected]> wrote:
>
> David Rowley <[email protected]> writes:
> > I don't think it would be right to fudge the costs in any way, but I
> > think the risk factor for IndexPaths could take into account the
> > number of unmatched index clauses and increment the risk factor, or
> > "certainty_factor" as it is currently in my brain-based design. That
> > way add_path() would be more likely to prefer the index that matches
> > the most conditions.
>
> This is somehow similar with my proposal at [1]? What do you think
> about the treat 'col op const' as 'col op $1' for the marked column?
> This could just resolve a subset of questions in your mind, but the
> method looks have a solid reason.
Do you mean this?
> + /*
> + * To make the planner more robust to handle some inaccurate statistics
> + * issue, we will add a extra cost to qpquals so that the less qpquals
> + * the lower cost it has.
> + */
> + cpu_run_cost += 0.01 * list_length(qpquals);
I don't think it's a good idea to add cost penalties like you proposed
there. This is what I meant by "I don't think it would be right to
fudge the costs in any way".
If you modify the costs to add some small penalty so that the planner
is more likely to favour some other plan, what happens if we then
decide the other plan has some issue and we want to penalise that for
some other reason? Adding the 2nd penalty might result in the original
plan choice again. Which one should be penalised more? I think the
uncertainty needs to be tracked separately.
Fudging the costs like this is also unlikely to play nicely with
add_path's use of STD_FUZZ_FACTOR. There'd be an incentive to do
things like total_cost *= STD_FUZZ_FACTOR; to ensure we get a large
enough penalty.
David
> [1] https://www.postgresql.org/message-id/CAApHDvovVWCbeR4v%2BA4Dkwb%3DYS_GuJG9OyCm8jZu%2B%2BcP2xsY_A%40...
^ permalink raw reply [nested|flat] 105+ messages in thread
* Re: a wrong index choose when statistics is out of date
2024-03-04 05:33 Re: a wrong index choose when statistics is out of date David Rowley <[email protected]>
@ 2024-03-04 17:03 ` Tomas Vondra <[email protected]>
1 sibling, 0 replies; 105+ messages in thread
From: Tomas Vondra @ 2024-03-04 17:03 UTC (permalink / raw)
To: David Rowley <[email protected]>; Andy Fan <[email protected]>; +Cc: pgsql-hackers
On 3/4/24 06:33, David Rowley wrote:
> On Sun, 3 Mar 2024 at 20:08, Andy Fan <[email protected]> wrote:
>> The issue can be reproduced with the following steps:
>>
>> create table x_events (.., created_at timestamp, a int, b int);
>>
>> create index idx_1 on t(created_at, a);
>> create index idx_2 on t(created_at, b);
>>
>> query:
>> select * from t where create_at = current_timestamp and b = 1;
>>
>> index (created_at, a) rather than (created_at, b) may be chosen for the
>> above query if the statistics think "create_at = current_timestamp" has
>> no rows, then both index are OK, actually it is true just because
>> statistics is out of date.
>
> I don't think there's really anything too special about the fact that
> the created_at column is always increasing. We commonly get 1-row
> estimates after multiplying the selectivities from individual stats.
> Your example just seems like yet another reason that this could
> happen.
>
> I've been periodically talking about introducing "risk" as a factor
> that the planner should consider. I did provide some detail in [1]
> about the design that was in my head at that time. I'd not previously
> thought that it could also solve this problem, but after reading your
> email, I think it can.
>
> I don't think it would be right to fudge the costs in any way, but I
> think the risk factor for IndexPaths could take into account the
> number of unmatched index clauses and increment the risk factor, or
> "certainty_factor" as it is currently in my brain-based design. That
> way add_path() would be more likely to prefer the index that matches
> the most conditions.
>
> The exact maths to calculate the certainty_factor for this case I
> don't quite have worked out yet. I plan to work on documenting the
> design of this and try and get a prototype patch out sometime during
> this coming southern hemisphere winter so that there's at least a full
> cycle of feedback opportunity before the PG18 freeze.
>
I've been thinking about this stuff too, so I'm curious to hear what
kind of plan you come up with. Every time I tried to formalize a more
concrete plan, I ended up with a very complex (and possible yet more
fragile) approach.
I think we'd need to consider a couple things:
1) reliability of cardinality estimates
I think this is pretty much the same concept as confidence intervals,
i.e. knowing not just the regular estimate, but also a range where the
actual value lies with high confidence (e.g. 90%).
For a single clauses this might not be terribly difficult, but for more
complex cases (multiple conditions, ...) it seems far more complex :-(
For example, let's say we know confidence intervals for two conditions.
What's the confidence interval when combined using AND or OR?
2) robustness of the paths
Knowing just the confidence intervals does not seem sufficient, though.
The other thing that matters is how this affects the paths, how robust
the paths are. I mean, if we have alternative paths with costs that flip
somewhere in the confidence interval - which one to pick? Surely one
thing to consider is how fast the costs change for each path.
3) complexity of the model
I suppose we'd prefer a systematic approach (and not some ad hoc
solution for one particular path/plan type). So this would be somehow
integrated into the cost model, making it yet more complex. I'm quite
worried about this (not necessarily because of performance reasons).
I wonder if trying to improve the robustness solely by changes in the
planning phase is a very promising approach. I mean - sure, we should
improve that, but by definition it relies on a priori information. And
not only the stats may be stale - it's a very lossy approximation of the
actual data. Even if the stats are perfectly up to date / very detailed,
there's still going to be a lot of uncertainty.
I wonder if we'd be better off if we experimented with more robust
plans, like SmoothScan [1] or g-join [2].
regards
[1]
https://stratos.seas.harvard.edu/sites/scholar.harvard.edu/files/stratos/files/smooth_vldbj.pdf
[2]
http://wwwlgis.informatik.uni-kl.de/cms/fileadmin/users/haerder/2011/JoinAndGrouping.pdf
--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
^ permalink raw reply [nested|flat] 105+ messages in thread
end of thread, other threads:[~2024-03-04 23:13 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 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 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 3/8] 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 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 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 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 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 01/12] 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 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 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 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 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 01/10] 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 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 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 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 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 01/10] Allow alternate compression methods for wal_compression Andrey Borodin <[email protected]>
2024-03-04 05:33 Re: a wrong index choose when statistics is out of date David Rowley <[email protected]>
2024-03-04 11:20 ` Re: a wrong index choose when statistics is out of date Andy Fan <[email protected]>
2024-03-04 23:13 ` Re: a wrong index choose when statistics is out of date David Rowley <[email protected]>
2024-03-04 17:03 ` Re: a wrong index choose when statistics is out of date Tomas Vondra <[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