Generate less WAL during GiST, GIN and SP-GiST index build.

Instead of WAL-logging every modification during the build separately,
first build the index without any WAL-logging, and make a separate pass
through the index at the end, to write all pages to the WAL. This
significantly reduces the amount of WAL generated, and is usually also
faster, despite the extra I/O needed for the extra scan through the index.
WAL generated this way is also faster to replay.

For GiST, the LSN-NSN interlock makes this a little tricky. All pages must
be marked with a valid (i.e. non-zero) LSN, so that the parent-child
LSN-NSN interlock works correctly. We now use magic value 1 for that during
index build. Change the fake LSN counter to begin from 1000, so that 1 is
safely smaller than any real or fake LSN. 2 would've been enough for our
purposes, but let's reserve a bigger range, in case we need more special
values in the future.

Author: Anastasia Lubennikova, Andrey V. Lepikhov
Reviewed-by: Heikki Linnakangas, Dmitry Dolgov

3 files changed, 17 insertions, 61 deletions

diff --git a/src/backend/access/spgist/spgdoinsert.c b/src/backend/access/spgist/spgdoinsert.c
index 0d07b8b2915..c34c44cd8b5 100644
--- a/src/backend/access/spgist/spgdoinsert.c
+++ b/src/backend/access/spgist/spgdoinsert.c
@@ -289,7 +289,7 @@ addLeafTuple(Relation index, SpGistState *state, SpGistLeafTuple leafTuple,
 
 	MarkBufferDirty(current->buffer);
 
-	if (RelationNeedsWAL(index))
+	if (RelationNeedsWAL(index) && !state->isBuild)
 	{
 		XLogRecPtr	recptr;
 		int			flags;
@@ -516,7 +516,7 @@ moveLeafs(Relation index, SpGistState *state,
 	MarkBufferDirty(current->buffer);
 	MarkBufferDirty(nbuf);
 
-	if (RelationNeedsWAL(index))
+	if (RelationNeedsWAL(index) && !state->isBuild)
 	{
 		XLogRecPtr	recptr;
 
@@ -1334,7 +1334,7 @@ doPickSplit(Relation index, SpGistState *state,
 		saveCurrent.buffer = InvalidBuffer;
 	}
 
-	if (RelationNeedsWAL(index))
+	if (RelationNeedsWAL(index) && !state->isBuild)
 	{
 		XLogRecPtr	recptr;
 		int			flags;
@@ -1531,7 +1531,7 @@ spgAddNodeAction(Relation index, SpGistState *state,
 
 		MarkBufferDirty(current->buffer);
 
-		if (RelationNeedsWAL(index))
+		if (RelationNeedsWAL(index) && !state->isBuild)
 		{
 			XLogRecPtr	recptr;
 
@@ -1644,7 +1644,7 @@ spgAddNodeAction(Relation index, SpGistState *state,
 
 		MarkBufferDirty(saveCurrent.buffer);
 
-		if (RelationNeedsWAL(index))
+		if (RelationNeedsWAL(index) && !state->isBuild)
 		{
 			XLogRecPtr	recptr;
 			int			flags;
@@ -1840,7 +1840,7 @@ spgSplitNodeAction(Relation index, SpGistState *state,
 
 	MarkBufferDirty(current->buffer);
 
-	if (RelationNeedsWAL(index))
+	if (RelationNeedsWAL(index) && !state->isBuild)
 	{
 		XLogRecPtr	recptr;
 
diff --git a/src/backend/access/spgist/spginsert.c b/src/backend/access/spgist/spginsert.c
index b06feafdc24..b40bd440cf0 100644
--- a/src/backend/access/spgist/spginsert.c
+++ b/src/backend/access/spgist/spginsert.c
@@ -105,26 +105,6 @@ spgbuild(Relation heap, Relation index, IndexInfo *indexInfo)
 	SpGistInitBuffer(nullbuffer, SPGIST_LEAF | SPGIST_NULLS);
 	MarkBufferDirty(nullbuffer);
 
-	if (RelationNeedsWAL(index))
-	{
-		XLogRecPtr	recptr;
-
-		XLogBeginInsert();
-
-		/*
-		 * Replay will re-initialize the pages, so don't take full pages
-		 * images.  No other data to log.
-		 */
-		XLogRegisterBuffer(0, metabuffer, REGBUF_WILL_INIT | REGBUF_STANDARD);
-		XLogRegisterBuffer(1, rootbuffer, REGBUF_WILL_INIT | REGBUF_STANDARD);
-		XLogRegisterBuffer(2, nullbuffer, REGBUF_WILL_INIT | REGBUF_STANDARD);
-
-		recptr = XLogInsert(RM_SPGIST_ID, XLOG_SPGIST_CREATE_INDEX);
-
-		PageSetLSN(BufferGetPage(metabuffer), recptr);
-		PageSetLSN(BufferGetPage(rootbuffer), recptr);
-		PageSetLSN(BufferGetPage(nullbuffer), recptr);
-	}
 
 	END_CRIT_SECTION();
 
@@ -151,6 +131,17 @@ spgbuild(Relation heap, Relation index, IndexInfo *indexInfo)
 
 	SpGistUpdateMetaPage(index);
 
+	/*
+	 * We didn't write WAL records as we built the index, so if WAL-logging is
+	 * required, write all pages to the WAL now.
+	 */
+	if (RelationNeedsWAL(index))
+	{
+		log_newpage_range(index, MAIN_FORKNUM,
+						  0, RelationGetNumberOfBlocks(index),
+						  true);
+	}
+
 	result = (IndexBuildResult *) palloc0(sizeof(IndexBuildResult));
 	result->heap_tuples = reltuples;
 	result->index_tuples = buildstate.indtuples;
diff --git a/src/backend/access/spgist/spgxlog.c b/src/backend/access/spgist/spgxlog.c
index 71836ee8a5b..ebe6ae8715b 100644
--- a/src/backend/access/spgist/spgxlog.c
+++ b/src/backend/access/spgist/spgxlog.c
@@ -73,38 +73,6 @@ addOrReplaceTuple(Page page, Item tuple, int size, OffsetNumber offset)
 }
 
 static void
-spgRedoCreateIndex(XLogReaderState *record)
-{
-	XLogRecPtr	lsn = record->EndRecPtr;
-	Buffer		buffer;
-	Page		page;
-
-	buffer = XLogInitBufferForRedo(record, 0);
-	Assert(BufferGetBlockNumber(buffer) == SPGIST_METAPAGE_BLKNO);
-	page = (Page) BufferGetPage(buffer);
-	SpGistInitMetapage(page);
-	PageSetLSN(page, lsn);
-	MarkBufferDirty(buffer);
-	UnlockReleaseBuffer(buffer);
-
-	buffer = XLogInitBufferForRedo(record, 1);
-	Assert(BufferGetBlockNumber(buffer) == SPGIST_ROOT_BLKNO);
-	SpGistInitBuffer(buffer, SPGIST_LEAF);
-	page = (Page) BufferGetPage(buffer);
-	PageSetLSN(page, lsn);
-	MarkBufferDirty(buffer);
-	UnlockReleaseBuffer(buffer);
-
-	buffer = XLogInitBufferForRedo(record, 2);
-	Assert(BufferGetBlockNumber(buffer) == SPGIST_NULL_BLKNO);
-	SpGistInitBuffer(buffer, SPGIST_LEAF | SPGIST_NULLS);
-	page = (Page) BufferGetPage(buffer);
-	PageSetLSN(page, lsn);
-	MarkBufferDirty(buffer);
-	UnlockReleaseBuffer(buffer);
-}
-
-static void
 spgRedoAddLeaf(XLogReaderState *record)
 {
 	XLogRecPtr	lsn = record->EndRecPtr;
@@ -976,9 +944,6 @@ spg_redo(XLogReaderState *record)
 	oldCxt = MemoryContextSwitchTo(opCtx);
 	switch (info)
 	{
-		case XLOG_SPGIST_CREATE_INDEX:
-			spgRedoCreateIndex(record);
-			break;
 		case XLOG_SPGIST_ADD_LEAF:
 			spgRedoAddLeaf(record);
 			break;