mirror/postgresql
2
0
Fork 0
mirror of https://git.postgresql.org/git/postgresql.git synced 2025-02-17 19:30:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

Use the buffer cache when initializing an unlogged index.

Browse Source 
Some of the ambuildempty functions used smgrwrite() directly, followed
by smgrimmedsync(). A few small problems with that:

Firstly, one is supposed to use smgrextend() when extending a
relation, not smgrwrite(). It doesn't make much difference in
production builds. smgrextend() updates the relation size cache, so
you miss that, but that's harmless because we never use the cached
relation size of an init fork. But if you compile with
CHECK_WRITE_VS_EXTEND, you get an assertion failure.

Secondly, the smgrwrite() calls were performed before WAL-logging, so
the page image written to disk had 0/0 as the LSN, not the LSN of the
WAL record. That's also harmless in practice, but seems sloppy.

Thirdly, it's better to use the buffer cache, because then you don't
need to smgrimmedsync() the relation to disk, which adds latency.
Bypassing the cache makes sense for bulk operations like index
creation, but not when you're just initializing an empty index.
Creation of unlogged tables is hardly performance bottleneck in any
real world applications, but nevertheless.

Backpatch to v16, but no further. These issues should be harmless in
practice, so better to not rock the boat in older branches.

Reviewed-by: Robert Haas
Discussion: https://www.postgresql.org/message-id/6e5bbc08-cdfc-b2b3-9e23-1a914b9850a9@iki.fi
This commit is contained in:
Heikki Linnakangas 2023-08-23 17:21:31 +03:00
parent ee99330a0b
commit ccadf73163
5 changed files with 58 additions and 85 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
contrib/bloom/blinsert.c
View File

@ -129,7 +129,7 @@ blbuild(Relation heap, Relation index, IndexInfo *indexInfo)
RelationGetRelationName(index));
/* Initialize the meta page */
BloomInitMetapage(index);
BloomInitMetapage(index, MAIN_FORKNUM);
/* Initialize the bloom build state */
memset(&buildstate, 0, sizeof(buildstate));
@ -163,31 +163,8 @@ blbuild(Relation heap, Relation index, IndexInfo *indexInfo)
void
blbuildempty(Relation index)
{
Page metapage;
/* Construct metapage. */
metapage = (Page) palloc_aligned(BLCKSZ, PG_IO_ALIGN_SIZE, 0);
BloomFillMetapage(index, metapage);
/*
* Write the page and log it. It might seem that an immediate sync would
* be sufficient to guarantee that the file exists on disk, but recovery
* itself might remove it while replaying, for example, an
* XLOG_DBASE_CREATE* or XLOG_TBLSPC_CREATE record. Therefore, we need
* this even when wal_level=minimal.
*/
PageSetChecksumInplace(metapage, BLOOM_METAPAGE_BLKNO);
smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, BLOOM_METAPAGE_BLKNO,
metapage, true);
log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
BLOOM_METAPAGE_BLKNO, metapage, true);
/*
* An immediate sync is required even if we xlog'd the page, because the
* write did not go through shared_buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
/* Initialize the meta page */
BloomInitMetapage(index, INIT_FORKNUM);
}
/*

2
contrib/bloom/bloom.h
View File

@ -177,7 +177,7 @@ typedef BloomScanOpaqueData *BloomScanOpaque;
/* blutils.c */
extern void initBloomState(BloomState *state, Relation index);
extern void BloomFillMetapage(Relation index, Page metaPage);
extern void BloomInitMetapage(Relation index);
extern void BloomInitMetapage(Relation index, ForkNumber forknum);
extern void BloomInitPage(Page page, uint16 flags);
extern Buffer BloomNewBuffer(Relation index);
extern void signValue(BloomState *state, BloomSignatureWord *sign, Datum value, int attno);

8
contrib/bloom/blutils.c
View File

@ -443,7 +443,7 @@ BloomFillMetapage(Relation index, Page metaPage)
* Initialize metapage for bloom index.
*/
void
BloomInitMetapage(Relation index)
BloomInitMetapage(Relation index, ForkNumber forknum)
{
Buffer metaBuffer;
Page metaPage;
@ -451,9 +451,11 @@ BloomInitMetapage(Relation index)
/*
* Make a new page; since it is first page it should be associated with
* block number 0 (BLOOM_METAPAGE_BLKNO).
* block number 0 (BLOOM_METAPAGE_BLKNO). No need to hold the extension
* lock because there cannot be concurrent inserters yet.
*/
metaBuffer = BloomNewBuffer(index);
metaBuffer = ReadBufferExtended(index, forknum, P_NEW, RBM_NORMAL, NULL);
LockBuffer(metaBuffer, BUFFER_LOCK_EXCLUSIVE);
Assert(BufferGetBlockNumber(metaBuffer) == BLOOM_METAPAGE_BLKNO);
/* Initialize contents of meta page */

41
src/backend/access/nbtree/nbtree.c
View File

@ -151,31 +151,32 @@ bthandler(PG_FUNCTION_ARGS)
void
btbuildempty(Relation index)
{
Buffer metabuf;
Page metapage;
/* Construct metapage. */
metapage = (Page) palloc_aligned(BLCKSZ, PG_IO_ALIGN_SIZE, 0);
/*
* Initalize the metapage.
*
* Regular index build bypasses the buffer manager and uses smgr functions
* directly, with an smgrimmedsync() call at the end. That makes sense
* when the index is large, but for an empty index, it's better to use the
* buffer cache to avoid the smgrimmedsync().
*/
metabuf = ReadBufferExtended(index, INIT_FORKNUM, P_NEW, RBM_NORMAL, NULL);
Assert(BufferGetBlockNumber(metabuf) == BTREE_METAPAGE);
_bt_lockbuf(index, metabuf, BT_WRITE);
START_CRIT_SECTION();
metapage = BufferGetPage(metabuf);
_bt_initmetapage(metapage, P_NONE, 0, _bt_allequalimage(index, false));
MarkBufferDirty(metabuf);
log_newpage_buffer(metabuf, true);
/*
* Write the page and log it. It might seem that an immediate sync would
* be sufficient to guarantee that the file exists on disk, but recovery
* itself might remove it while replaying, for example, an
* XLOG_DBASE_CREATE* or XLOG_TBLSPC_CREATE record. Therefore, we need
* this even when wal_level=minimal.
*/
PageSetChecksumInplace(metapage, BTREE_METAPAGE);
smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, BTREE_METAPAGE,
metapage, true);
log_newpage(&RelationGetSmgr(index)->smgr_rlocator.locator, INIT_FORKNUM,
BTREE_METAPAGE, metapage, true);
END_CRIT_SECTION();
/*
* An immediate sync is required even if we xlog'd the page, because the
* write did not go through shared_buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
_bt_unlockbuf(index, metabuf);
ReleaseBuffer(metabuf);
}
/*

63
src/backend/access/spgist/spginsert.c
View File

@ -155,49 +155,42 @@ spgbuild(Relation heap, Relation index, IndexInfo *indexInfo)
void
spgbuildempty(Relation index)
{
Page page;
/* Construct metapage. */
page = (Page) palloc_aligned(BLCKSZ, PG_IO_ALIGN_SIZE, 0);
SpGistInitMetapage(page);
Buffer metabuffer,
rootbuffer,
nullbuffer;
/*
* Write the page and log it unconditionally. This is important
* particularly for indexes created on tablespaces and databases whose
* creation happened after the last redo pointer as recovery removes any
* of their existing content when the corresponding create records are
* replayed.
* Initialize the meta page and root pages
*/
PageSetChecksumInplace(page, SPGIST_METAPAGE_BLKNO);
smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_METAPAGE_BLKNO,
page, true);
log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
SPGIST_METAPAGE_BLKNO, page, true);
metabuffer = ReadBufferExtended(index, INIT_FORKNUM, P_NEW, RBM_NORMAL, NULL);
LockBuffer(metabuffer, BUFFER_LOCK_EXCLUSIVE);
rootbuffer = ReadBufferExtended(index, INIT_FORKNUM, P_NEW, RBM_NORMAL, NULL);
LockBuffer(rootbuffer, BUFFER_LOCK_EXCLUSIVE);
nullbuffer = ReadBufferExtended(index, INIT_FORKNUM, P_NEW, RBM_NORMAL, NULL);
LockBuffer(nullbuffer, BUFFER_LOCK_EXCLUSIVE);
/* Likewise for the root page. */
SpGistInitPage(page, SPGIST_LEAF);
Assert(BufferGetBlockNumber(metabuffer) == SPGIST_METAPAGE_BLKNO);
Assert(BufferGetBlockNumber(rootbuffer) == SPGIST_ROOT_BLKNO);
Assert(BufferGetBlockNumber(nullbuffer) == SPGIST_NULL_BLKNO);
PageSetChecksumInplace(page, SPGIST_ROOT_BLKNO);
smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_ROOT_BLKNO,
page, true);
log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
SPGIST_ROOT_BLKNO, page, true);
START_CRIT_SECTION();
/* Likewise for the null-tuples root page. */
SpGistInitPage(page, SPGIST_LEAF | SPGIST_NULLS);
SpGistInitMetapage(BufferGetPage(metabuffer));
MarkBufferDirty(metabuffer);
SpGistInitBuffer(rootbuffer, SPGIST_LEAF);
MarkBufferDirty(rootbuffer);
SpGistInitBuffer(nullbuffer, SPGIST_LEAF | SPGIST_NULLS);
MarkBufferDirty(nullbuffer);
PageSetChecksumInplace(page, SPGIST_NULL_BLKNO);
smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_NULL_BLKNO,
page, true);
log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
SPGIST_NULL_BLKNO, page, true);
log_newpage_buffer(metabuffer, true);
log_newpage_buffer(rootbuffer, true);
log_newpage_buffer(nullbuffer, true);
/*
* An immediate sync is required even if we xlog'd the pages, because the
* writes did not go through shared buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
END_CRIT_SECTION();
UnlockReleaseBuffer(metabuffer);
UnlockReleaseBuffer(rootbuffer);
UnlockReleaseBuffer(nullbuffer);
}
/*