To: Bjorn Halvor Solberg <bhso@pacbell.net>
cc: pgsql-interfaces@postgresql.org
Subject: Re: Backend dies for DBI 
In-reply-to: <m3k85ycvq6.fsf@pacbell.net> 
References: <m3k862d2nk.fsf@pacbell.net> <22575.984012545@sss.pgh.pa.us>
	<m33dcpqcyq.fsf@pacbell.net> <24449.984016287@sss.pgh.pa.us>
	<m3k85ycvq6.fsf@pacbell.net>
Comments: In-reply-to Bjorn Halvor Solberg <bhso@pacbell.net>
	message dated "09 Mar 2001 16:29:21 -0800"
Date: Fri, 09 Mar 2001 22:39:32 -0500
Message-ID: <13541.984195572@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Bjorn Halvor Solberg <bhso@pacbell.net> writes:
> What confuses me is that it ran fine for a couple of months.  Then, after
> the power outage, it crashes every few hours, resulting in having to
> restart the web server to regain the DB handle.

>> Hmm.  Possibly some corrupted data in the database now?

> That thought did come to mind, but manual queries go through just fine.
> I did vacuum the database, but that didn't help.

This is reaching a little bit, but something to think about: maybe you
are not running exactly the same software or configuration as you were
before the power outage.  A well known gotcha on machines that stay up
for a long time between boots is that you may have changed something
without realizing that the effects would disappear at the next boot
(or alternatively, only come into effect at the next boot).  For
example, starting a daemon without remembering to arrange for the boot
scripts to start it, or changing a daemon's config file and neglecting
to test the results, or installing a new version of some shlib that's
used by a daemon.  When a reboot does finally happen, you've forgotten
that there was anything to worry about, and so the sudden change in
behavior is mysterious.

I realize this doesn't give anything concrete to go on, but maybe it
will give you an idea of where to look.

			regards, tom lane