Re: Odd db lockup - investigation advice wanted

From: Douglas McNaught <doug(at)mcnaught(dot)org>
To: Marc Munro <marc(at)bloodnok(dot)com>
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: Odd db lockup - investigation advice wanted
Date: 2005-11-07 22:21:15
Message-ID: m2hdaohzfo.fsf@Douglas-McNaughts-Powerbook.local
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Marc Munro <marc(at)bloodnok(dot)com> writes:

> PGDATA is installed on a Netapp network storage device.

This is generally not recommended--it should be on a local disk (SAN,
etc) rather than NFS.

> We are using slony 1.1.0 for replication.
>
> The (provider) database locked-up after I killed a slony client process
> (kill -9) on the subscriber. Psql connections would not respond to \d
> and simply locked up. I was able to run a query to check for blocking
> locks - this returned no rows. There was a significant test load on the
> database at the time.
>
> I stopped the database but was unable to restart it. I was unable to
> kill a number of postgres processes and could not release postgres
> shared memory. Having decided that the database was toast, I discovered
> that I could not even delete the database files, and eventually the only
> solution was a full reboot.

This sounds like you had kernel problems, not PG problems. Linux NFS
support is continually improving but still seems to have issues. I
would strongly recommend using local disk for your database storage.

-Doug

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Cristian Prieto 2005-11-07 22:51:51 Help with Array Function in C language...
Previous Message Aly Dharshi 2005-11-07 21:29:12 Re: DTrace?