Re: [HACKERS] Re: PD_ALL_VISIBLE flag was incorrectly set happend during repeatable vacuum

daveg <daveg@sonic.net>

From: daveg <daveg@sonic.net>

To: bricklen <bricklen@gmail.com>

Cc: pgsql-admin <pgsql-admin@postgresql.org>, PostgreSQL-development <pgsql-hackers@postgresql.org>

Date: 2011-03-03T00:29:52Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

Fix a violation of WAL coding rules in the recent patch to include an
- 7fc7a7c4d082 9.0.0 cited

On Wed, Mar 02, 2011 at 04:20:24PM -0800, bricklen wrote:
> On Wed, Mar 2, 2011 at 3:53 PM, daveg <daveg@sonic.net> wrote:
> >> > Postgresql version is 8.4.4.
> >>
> >> I don't see how this could be related, but since you're running on NFS,
> >> maybe it is, somehow:
> >> http://archives.postgresql.org/message-id/4D40DDB7.1010000@credativ.com
> >> (for example what if the visibility map fork's last page is overwritten?)
> >
> > Running on ISCSI, not nfs. But it is still a Netapp, so who knows. I'll look.
> > Also, we are not seeing any of the "unexpected data beyond EOF" errors,
> > just thousands per day of the PD_ALL_VISIBLE error.
> >
> > -dg
> 
> FWIW, we had a couple occurrences of that message about a month ago on 9.0.2
> 
> http://archives.postgresql.org/pgsql-general/2011-01/msg00887.php
> 
> Haven't seen it since we ran a cluster-wide vacuum.

We did a shutdown and restart to clear the buffer cache (but did not reboot
the host) and a vacuum on all dbs in the cluster last night. That cleared it
up for a couple hours, but we are still getting lots of these messages.

Most of them are pg_statistic and we create and drop hundreds of thousands of
temp tables daily, so there is a good chance there is a concurrancy issue.

-dg
 

-- 
David Gould       daveg@sonic.net      510 536 1443    510 282 0869
If simplicity worked, the world would be overrun with insects.