From: | Andy Fan <zhihui(dot)fan1213(at)gmail(dot)com> |
---|---|
To: | PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
Subject: | initscan for MVCC snapshot |
Date: | 2020-12-07 12:26:53 |
Message-ID: | CAKU4AWrfUNtfxyq=ApJ-K4WeuhCC_6wekwETJhTCOTMBTzN+eA@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Hi:
I see initscan calls RelationGetwNumberOfBlocks every time and rescan calls
initscan as well. In my system, RelationGetNumberOfBlocks is expensive
(the reason
doesn't deserve a talk.. ), so in a nest loop + Bitmap heap scan case,
the
impact will be huge. The comments of initscan are below.
/*
* Determine the number of blocks we have to scan.
*
* It is sufficient to do this once at scan start, since any tuples added
* while the scan is in progress will be invisible to my snapshot anyway.
* (That is not true when using a non-MVCC snapshot. However, we couldn't
* guarantee to return tuples added after scan start anyway, since they
* might go into pages we already scanned. To guarantee consistent
* results for a non-MVCC snapshot, the caller must hold some higher-level
* lock that ensures the interesting tuple(s) won't change.)
*/
I still do not fully understand the comments. Looks we only need to call
multi times for non-MVCC snapshot, IIUC, does the following change
reasonable?
===
diff --git a/src/backend/access/heap/heapam.c
b/src/backend/access/heap/heapam.c
index 1b2f70499e..8238eabd8b 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -211,6 +211,7 @@ initscan(HeapScanDesc scan, ScanKey key, bool
keep_startblock)
ParallelBlockTableScanDesc bpscan = NULL;
bool allow_strat;
bool allow_sync;
+ bool is_mvcc = scan->rs_base.rs_snapshot &&
IsMVCCSnapshot(scan->rs_base.rs_snapshot);
/*
* Determine the number of blocks we have to scan.
@@ -229,7 +230,8 @@ initscan(HeapScanDesc scan, ScanKey key, bool
keep_startblock)
scan->rs_nblocks = bpscan->phs_nblocks;
}
else
- scan->rs_nblocks =
RelationGetNumberOfBlocks(scan->rs_base.rs_rd);
+ if (scan->rs_nblocks == -1 || !is_mvcc)
+ scan->rs_nblocks =
RelationGetNumberOfBlocks(scan->rs_base.rs_rd);
/*
* If the table is large relative to NBuffers, use a bulk-read
access
@@ -1210,6 +1212,7 @@ heap_beginscan(Relation relation, Snapshot snapshot,
else
scan->rs_base.rs_key = NULL;
+ scan->rs_nblocks = -1;
initscan(scan, key, false);
--
Best Regards
Andy Fan
From | Date | Subject | |
---|---|---|---|
Next Message | Ashutosh Bapat | 2020-12-07 12:27:22 | Re: [PATCH] Keeps tracking the uniqueness with UniqueKey |
Previous Message | Andy Fan | 2020-12-07 12:14:56 | Re: [PATCH] Keeps tracking the uniqueness with UniqueKey |