Re: performance sol10 zone (fup) - Mailing list pgsql-performance

From Scott Marlowe
Subject Re: performance sol10 zone (fup)
Date
Msg-id AANLkTik7W3qcttmDwrgujXsn9eraVoMGL2ADXHB994X2@mail.gmail.com
Whole thread Raw
In response to performance sol10 zone (fup)  ("Heiko L." <heikol@fh-lausitz.de>)
List pgsql-performance
On Wed, Aug 4, 2010 at 1:16 AM, Heiko L. <heikol@fh-lausitz.de> wrote:
> Hallo,
>
> Im running pg-8,pgpoolII on sol10-zone.

I noticed late you mention 8.3.1.  Two points, you're missing > 1 year
of updates, bug fixes, security patches etc.  Assuming this version
was fast before, we'll assume it's not the cause of this problem,
however, you're asking for trouble with a version that old.  There are
bugs that might not bite you today, but may well in the future.
Please upgrade to 8.3.11.

> After update sol10u7, queries on coltype timestamp are very slow.
> System: sparc, 2GB RAM

Is it possible you had an index that was working that now isn't?  Are
the queries you included the real ones or approximations?

It looks like you have a bunch of seq scans happening.  If they're all
happening on the same table or small set of them, then a lot of
queries should be able to access them in any order together in 8.3

Are sequential scans normal for this query when it runs fast?

What does vmstat 10 and / or iostat -xd 10 have to say while this is running?

--
To understand recursion, one must first understand recursion.

pgsql-performance by date:

Previous
From: Robert Haas
Date:
Subject: Re: performance sol10 zone (fup)
Next
From: Greg Smith
Date:
Subject: Re: Completely un-tuned Postgresql benchmark results: SSD vs desktop HDD