I wrote this patch last year in response to a customer issue and I
thought I had submitted it here, but evidently I didn't. So here it is.
The short story is: in commit 5364b357fb11 we increased the size of
pg_commit (née pg_clog) but we didn't increase the size of pg_commit_ts
to match. When commit_ts is in use, this can lead to significant buffer
thrashing and thus poor performance.
Since commit_ts entries are larger than pg_commit, my proposed patch uses
twice as many buffers.
Suffice it to say once we did this the customer problem went away.
(Andrey Borodin already has a patch to change the behavior for
multixact, which is something we should perhaps also do. I now notice
that they're also reporting a bug in that thread ... sigh)
--
Álvaro Herrera 39°49'30"S 73°17'W
"The problem with the future is that it keeps turning into the present"
(Hobbes)
diff --git a/src/backend/access/transam/commit_ts.c b/src/backend/access/transam/commit_ts.c
index b786eeff7a..712f1eff01 100644
--- a/src/backend/access/transam/commit_ts.c
+++ b/src/backend/access/transam/commit_ts.c
@@ -530,7 +530,7 @@ pg_xact_commit_timestamp_origin(PG_FUNCTION_ARGS)
Size
CommitTsShmemBuffers(void)
{
- return Min(16, Max(4, NBuffers / 1024));
+ return Min(256, Max(4, NBuffers / 512));
}
/*
