database/postgres
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2025-07-28 23:42:10 +03:00
Code Activity

Arrange to cache btree metapage data in the relcache entry for the index,

Browse Source 
thereby saving a visit to the metapage in most index searches/updates.
This wouldn't actually save any I/O (since in the old regime the metapage
generally stayed in cache anyway), but it does provide a useful decrease
in bufmgr traffic in high-contention scenarios.  Per my recent proposal.
This commit is contained in:
Tom Lane
2006-04-25 22:46:05 +00:00
parent 89083876c9
commit d2896a9ed1
6 changed files with 122 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
src/backend/access/nbtree/nbtree.c
View File

@ -12,7 +12,7 @@
* Portions Copyright (c) 1994, Regents of the University of California
*
* IDENTIFICATION
* $PostgreSQL: pgsql/src/backend/access/nbtree/nbtree.c,v 1.144 2006/04/01 03:03:37 tgl Exp $
* $PostgreSQL: pgsql/src/backend/access/nbtree/nbtree.c,v 1.145 2006/04/25 22:46:05 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -26,6 +26,7 @@
#include "miscadmin.h"
#include "storage/freespace.h"
#include "storage/smgr.h"
#include "utils/inval.h"
#include "utils/memutils.h"
@ -127,6 +128,17 @@ btbuild(PG_FUNCTION_ARGS)
}
#endif /* BTREE_BUILD_STATS */
/*
* If we are reindexing a pre-existing index, it is critical to send out
* a relcache invalidation SI message to ensure all backends re-read the
* index metapage. In most circumstances the update-stats operation will
* cause that to happen, but at the moment there are corner cases where
* no pg_class update will occur, so force an inval here. XXX FIXME:
* the upper levels of CREATE INDEX should handle the stats update as
* well as guaranteeing relcache inval.
*/
CacheInvalidateRelcache(index);
/* since we just counted the # of tuples, may as well update stats */
IndexCloseAndUpdateStats(heap, reltuples, index, buildstate.indtuples);