Re: perfromance impact of vacuum - Mailing list pgsql-general

From Tom Lane
Subject Re: perfromance impact of vacuum
Date
Msg-id 14243.1058217319@sss.pgh.pa.us
Whole thread Raw
In response to perfromance impact of vacuum  ("Jay O'Connor" <joconnor@cybermesa.com>)
List pgsql-general
"Jay O'Connor" <joconnor@cybermesa.com> writes:
> What impact in performance does vacuum have on an active database?

I've seen varying reports on this; some say it's not a problem, and some
say it is.  I suspect (without real proof) that the critical issue is
whether you have adequate I/O bandwidth.  If your disk is near
saturation already, the extra I/O load for vacuum can drive you over the
knee of the performance curve and cause serious degradation.  But if
you've got spare I/O bandwidth then it's not a big problem.

You can reduce the I/O needs for routine vacuums by only vacuuming the
tables you need to, and selecting an appropriate vacuum frequency for
each one.  (You might want to try the "autovacuum" daemon that's being
worked on, as a substitute for doing this tuning by hand.  Not sure if
the daemon is ready for prime time though.)

BTW, I suggest moving any followup discussion to pgsql-performance.

            regards, tom lane

pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: Reverse compatibility
Next
From: Zengfa Gao
Date:
Subject: Lock file "/tmp/.s.PGSQL.5432.lock" already exists.