Обсуждение: I/O error during autovacuum

I recently started seeing the following in the logs:

ERROR:  could not read block 46 of relation
pg_tblspc/16385/8578554/11521501: Input/output error
CONTEXT:  automatic vacuum of table "foo.pg_toast.pg_toast_8578780"

This is a Slony-I 1.2.21 slave, which runs 8.4.3 on RHEL 5.5.

This toast relation is the toast of an index of one of the most
frequently used tables.

Is this *definitely* a disk error, or could it be related to PostgreSQL
itself? I'm in the process of replacing disks, but I wanted to make
sure.

Regards,
--
Devrim GÜNDÜZ
PostgreSQL Danışmanı/Consultant, Red Hat Certified Engineer
PostgreSQL RPM Repository: http://yum.pgrpms.org
Community: devrim~PostgreSQL.org, devrim.gunduz~linux.org.tr
http://www.gunduz.org  Twitter: http://twitter.com/devrimgunduz

Devrim =?ISO-8859-1?Q?G=DCND=DCZ?= <devrim@gunduz.org> writes:
> I recently started seeing the following in the logs:

> ERROR:  could not read block 46 of relation
> pg_tblspc/16385/8578554/11521501: Input/output error
> CONTEXT:  automatic vacuum of table "foo.pg_toast.pg_toast_8578780"

> Is this *definitely* a disk error, or could it be related to PostgreSQL
> itself?

"Input/output error" is pretty much always indicative of a hardware
problem.  If it's always associated with attempts to access a specific
disk block then it's definitely a bad-sector error.  There might be more
details in the kernel log.

            regards, tom lane

Devrim GÜNDÜZ wrote:
>
> I recently started seeing the following in the logs:
>
> ERROR:  could not read block 46 of relation
> pg_tblspc/16385/8578554/11521501: Input/output error
> CONTEXT:  automatic vacuum of table "foo.pg_toast.pg_toast_8578780"
>
> This is a Slony-I 1.2.21 slave, which runs 8.4.3 on RHEL 5.5.
>
> This toast relation is the toast of an index of one of the most
> frequently used tables.
>
> Is this *definitely* a disk error, or could it be related to PostgreSQL
> itself? I'm in the process of replacing disks, but I wanted to make
> sure.

The error comes straight from the kernel, so the problem is certainly
not in PostgreSQL.

--
Alvaro Herrera                                http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc.

On Thu, 2010-04-22 at 18:47 -0400, Tom Lane wrote:
> > Is this *definitely* a disk error, or could it be related to
> PostgreSQL
> > itself?
>
> "Input/output error" is pretty much always indicative of a hardware
> problem.  If it's always associated with attempts to access a specific
> disk block then it's definitely a bad-sector error.  T

Thanks. I just wanted to make sure that I did not hit a PostgreSQL bug.

> There might be more details in the kernel log.

That is another part that confuses me. cciss driver *may* report false
positives under some circumstances, per a bug report that I saw at
bugzilla (I cannot remember bz id right now). That's why I asked to
list, too.

Regards,
--
Devrim GÜNDÜZ
PostgreSQL Danışmanı/Consultant, Red Hat Certified Engineer
PostgreSQL RPM Repository: http://yum.pgrpms.org
Community: devrim~PostgreSQL.org, devrim.gunduz~linux.org.tr
http://www.gunduz.org  Twitter: http://twitter.com/devrimgunduz