Re: Help with duration of statement: EXECUTE [PREPARE: COMMIT] - Mailing list pgsql-performance

From Tom Lane
Subject Re: Help with duration of statement: EXECUTE [PREPARE: COMMIT]
Date
Msg-id 9954.1287498334@sss.pgh.pa.us
Whole thread Raw
In response to Re: Help with duration of statement: EXECUTE [PREPARE: COMMIT]  (Eric Comeau <ecomeau@signiant.com>)
List pgsql-performance
Eric Comeau <ecomeau@signiant.com> writes:
> Okay - I'm starting to see other stmts other than just commits taking
> longer than 5 secs sometimes as well now - stress test has been running
> for 3 days now...some commits 17 and 15 secs ouch...

If it's not just commits then some of the stranger theories go away.
I think you're almost certainly looking at overstressed disk.

It'd be worth your while to turn on checkpoint logging and see if
the slow operations occur during checkpoints.  If so, you might be
able to ameliorate things by tweaking the checkpoint parameters
to spread out checkpoint I/O load some more.  But the bottom line
here is that you haven't got much headroom between your application's
needs and the max throughput available from your disk.  Better disk
hardware will be the only real cure.

            regards, tom lane

pgsql-performance by date:

Previous
From: Mladen Gogala
Date:
Subject: Re: HashJoin order, hash the large or small table? Postgres likes to hash the big one, why?
Next
From: Tom Lane
Date:
Subject: Re: odd postgresql performance (excessive lseek)