Re: does VACUUM ANALYZE complete with this error?

From: Susan Russo <russo(at)morgan(dot)harvard(dot)edu>
To: russo(at)morgan(dot)harvard(dot)edu, tgl(at)sss(dot)pgh(dot)pa(dot)us
Cc: harvsys(at)morgan(dot)harvard(dot)edu, pgsql-performance(at)postgresql(dot)org
Subject: Re: does VACUUM ANALYZE complete with this error?
Date: 2007-05-23 19:03:57
Message-ID: 200705231903.l4NJ3vuT014318@larrys.harvard.edu
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Hi Tom - thanks for the additional/confirming info.

>So you definitely have got a problem of needing more vacuuming...

Yes, we're going to nightly, as I said in last message, however,
it worse than this.....

I found that *1* vacuum analyze works well in many instances to
help optimize query performance (which in one example was running
in lightening speed on 2 of our 5 identical software/hardware/configs
Pg 8.1.4 servers). However, in several cases, a *2nd* vacuum
analyze was necessary. (btw - first vacuum was after adjusting
max_fsm_pages, and getting no error msgs from vacuum).

I *think* - please advise, I may be able to affect configs
for a more effective vacuum analyze the first time around (??)
Perhaps an increase to deafult_statistics_target (set to 100??).

I'd read that when performing a vacuum analyze, Pg doesn't actually
go through all values in each table and update statistics, rather,
it samples some of the values and uses that statistical sample.
Thus, different runs of the vacuum analyze might generate different
statistics (on different dbs on different servers) since the same db
may be used differently on a different server. Is this correct??

Thanks for any advice....I'm hoping regular duplicate vacuum
analyze isn't the solution...

Susan

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Chris Mair 2007-05-23 20:24:23 Re: Drop table vs Delete record
Previous Message Y Sidhu 2007-05-23 18:58:06 Simulate database fragmentation