Re: Will Postgres ever lock with read only queries? - Mailing list pgsql-performance

From Nikolas Everett
Subject Re: Will Postgres ever lock with read only queries?
Date
Msg-id d4e11e980907280725x2067a914oa73bf294b9e4178b@mail.gmail.com
Whole thread Raw
In response to Re: Will Postgres ever lock with read only queries?  (Robert James <srobertjames@gmail.com>)
List pgsql-performance
Can you run those two queries with psql?

I remember having some trouble running multiple queries in the same pgadmin process.  Both would get stuck until both finished I think.  I went to running a pgadmin process per query.

On Tue, Jul 28, 2009 at 9:17 AM, Robert James <srobertjames@gmail.com> wrote:
Thanks for the replies.  I'm running Postgres 8.2 on Windows XP, Intel Core Duo (though Postgres seems to use only one 1 core).
The queries are self joins on very large tables, with lots of nested loops.

On Mon, Jul 27, 2009 at 9:40 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
Chris <dmagick@gmail.com> writes:
> Robert James wrote:
>> Hi.  I'm seeing some weird behavior in Postgres.  I'm running read only
>> queries (SELECT that is - no UPDATE or DELETE or INSERT is happening at
>> all).  I can run one rather complicated query and the results come
>> back... eventually.  Likewise with another.  But, when I run both
>> queries at the same time, Postgres seems to ground to a halt.

> They're probably not blocking each other but more likely you're
> exhausting your servers resources. If they return "eventually"
> individually, then running both at the same time will take at least
> "eventually x2".

It could be a lot more than x2.  If the two queries together eat enough
RAM to drive the machine into swapping, where it didn't swap while
doing one at a time, the slowdown could be orders of magnitude.

Watching vmstat output might be informative --- it would at least give
an idea if the bottleneck is CPU, I/O, or swap.

                       regards, tom lane


pgsql-performance by date:

Previous
From: Robert James
Date:
Subject: Re: Will Postgres ever lock with read only queries?
Next
From: Dave Youatt
Date:
Subject: Re: hyperthreaded cpu still an issue in 8.4?