Thread: BUG #4229: Page header error

BUG #4229: Page header error

From
"Valeriy"
Date:
The following bug has been logged online:

Bug reference:      4229
Logged by:          Valeriy
Email address:      tester080606@rambler.ru
PostgreSQL version: 7.4.5
Operating system:   Linux 2.6.17
Description:        Page header error
Details:

We have PostgreSQL v. 7.4.5 and main application
working with PGSQL via libpq. Linux 2.6.17.

Page header error periodically occurs when
main application is working hard.
We repair it automatically using restore
from night dump, no problems.

There are some servers (among more than 20)
that are loaded much less, so this error doesn't occur at all.

Can we assume that this error is connected with
high load of hardware only?

I suppose that  the upgrade up to version 8.x
will not exclude this error.

Thank You
Valeriy

Re: BUG #4229: Page header error

From
Tom Lane
Date:
"Valeriy" <tester080606@rambler.ru> writes:
> Bug reference:      4229
> Logged by:          Valeriy
> Email address:      tester080606@rambler.ru
> PostgreSQL version: 7.4.5
> Operating system:   Linux 2.6.17
> Description:        Page header error
> Details:

> We have PostgreSQL v. 7.4.5 and main application
> working with PGSQL via libpq. Linux 2.6.17.

> Page header error periodically occurs when
> main application is working hard.
> We repair it automatically using restore
> from night dump, no problems.

> There are some servers (among more than 20)
> that are loaded much less, so this error doesn't occur at all.

> Can we assume that this error is connected with
> high load of hardware only?

Theory 1: if your kernel is as old as your PG version, it very likely
has bugs in its disk drivers, which might only manifest under load.
Consider a newer kernel.

Theory 2: you are seeing intermittent hardware errors.  In that case
you need to get new hardware, SOON.  Hardware that has gotten flaky
is likely to stop working entirely at any time --- or at least have
the failure rate go high enough to make it unusable.

Looking into the kernel logs might offer some evidence to determine
which theory is correct.

BTW, you should read the release notes
http://www.postgresql.org/docs/7.4/static/release.html
and think hard about why you are still using 7.4.5 and not
7.4.19 or so.  This particular symptom doesn't sound like any
known Postgres bug, but there are plenty of known bugs in 7.4.5.

            regards, tom lane