error in vacuum

Started by Dave Peticolasover 7 years ago9 messagesgeneral
Jump to latest
#1Dave Peticolas
dave@krondo.com

I'm seeing this error on a table repeatedly:

found update xid 3433096759 <(343)%20309-6759> from before relfrozenxid
3498422955

What can I do to repair this table?

thanks,
dave

#2Dave Peticolas
dave@krondo.com
In reply to: Dave Peticolas (#1)
Re: error in vacuum

On Sat, Sep 1, 2018 at 4:47 PM Dave Peticolas <dave@krondo.com> wrote:

Show quoted text

I'm seeing this error on a table repeatedly:

found update xid 3433096759 <(343)%20309-6759> from before relfrozenxid
3498422955

What can I do to repair this table?

This is version 9.6.

#3Adrian Klaver
adrian.klaver@aklaver.com
In reply to: Dave Peticolas (#1)
Re: error in vacuum

On 09/01/2018 04:47 PM, Dave Peticolas wrote:

I'm seeing this error on a table repeatedly:

found update xid 3433096759 <tel:%28343%29%20309-6759> from before
relfrozenxid 3498422955

Where is this, your primary or the instance you are running from the backup?

What can I do to repair this table?

thanks,
dave

--
Adrian Klaver
adrian.klaver@aklaver.com

#4Dave Peticolas
dave@krondo.com
In reply to: Adrian Klaver (#3)
Re: error in vacuum

On Sat, Sep 1, 2018, 5:10 PM Adrian Klaver <adrian.klaver@aklaver.com>
wrote:

On 09/01/2018 04:47 PM, Dave Peticolas wrote:

I'm seeing this error on a table repeatedly:

found update xid 3433096759 <tel:%28343%29%20309-6759> from before
relfrozenxid 3498422955

Where is this, your primary or the instance you are running from the
backup?

This is a primary, 9.6.8 to be exact.

Show quoted text

What can I do to repair this table?

thanks,
dave

--
Adrian Klaver
adrian.klaver@aklaver.com

#5Dave Peticolas
dave@krondo.com
In reply to: Dave Peticolas (#4)
Re: error in vacuum

On Sat, Sep 1, 2018 at 5:28 PM Dave Peticolas <dave@krondo.com> wrote:

On Sat, Sep 1, 2018, 5:10 PM Adrian Klaver <adrian.klaver@aklaver.com>
wrote:

On 09/01/2018 04:47 PM, Dave Peticolas wrote:

I'm seeing this error on a table repeatedly:

found update xid 3433096759 <(343)%20309-6759>

<tel:%28343%29%20309-6759> from before

relfrozenxid 3498422955

Where is this, your primary or the instance you are running from the
backup?

This is a primary, 9.6.8 to be exact.

What can I do to repair this table?

One thing I tried on a database restored from a primary backup that
exhibits the same "found update xid" error is this:

1. pg_dump the table and its referencing tables.
2. truncate the table and its referencing tables.
3. Load the dump file.

Does that seem reasonable?

#6Tom Lane
tgl@sss.pgh.pa.us
In reply to: Dave Peticolas (#5)
Re: error in vacuum

Dave Peticolas <dave@krondo.com> writes:

I'm seeing this error on a table repeatedly:
found update xid 3433096759 from before relfrozenxid 3498422955

This is a primary, 9.6.8 to be exact.

This message is known to be associated with several data-corruption bugs
we've fixed recently. You should update to current (9.6.10) sooner not
later. I don't think that will cure existing corruption for you, but
it should prevent things from getting worse.

regards, tom lane

#7Dave Peticolas
dave@krondo.com
In reply to: Tom Lane (#6)
Re: error in vacuum

On Sat, Sep 1, 2018 at 6:56 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

Dave Peticolas <dave@krondo.com> writes:

I'm seeing this error on a table repeatedly:
found update xid 3433096759 <(343)%20309-6759> from before

relfrozenxid 3498422955

This is a primary, 9.6.8 to be exact.

This message is known to be associated with several data-corruption bugs
we've fixed recently. You should update to current (9.6.10) sooner not
later. I don't think that will cure existing corruption for you, but
it should prevent things from getting worse.

Ok, thank you for that information I will do so. The dump/truncate/restore
I mentioned in a later post seems to fix the corruption in a restored DB.
Given the nature of the bug do you think that would fix that table?

#8Andres Freund
andres@anarazel.de
In reply to: Dave Peticolas (#7)
Re: error in vacuum

On 2018-09-01 19:02:18 -0700, Dave Peticolas wrote:

On Sat, Sep 1, 2018 at 6:56 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

Dave Peticolas <dave@krondo.com> writes:

I'm seeing this error on a table repeatedly:
found update xid 3433096759 <(343)%20309-6759> from before

relfrozenxid 3498422955

This is a primary, 9.6.8 to be exact.

This message is known to be associated with several data-corruption bugs
we've fixed recently. You should update to current (9.6.10) sooner not
later. I don't think that will cure existing corruption for you, but
it should prevent things from getting worse.

Ok, thank you for that information I will do so. The dump/truncate/restore
I mentioned in a later post seems to fix the corruption in a restored DB.
Given the nature of the bug do you think that would fix that table?

Yes, it's fairly likely that it does. You might even get around it by
just doing a no-op update of the whole table. Nevertheless you need to
upgrade, or more of that corruption can occur.

Greetings,

Andres Freund

#9Dave Peticolas
dave@krondo.com
In reply to: Andres Freund (#8)
Re: error in vacuum

On Sat, Sep 1, 2018 at 7:24 PM Andres Freund <andres@anarazel.de> wrote:

On 2018-09-01 19:02:18 -0700, Dave Peticolas wrote:

On Sat, Sep 1, 2018 at 6:56 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

Dave Peticolas <dave@krondo.com> writes:

I'm seeing this error on a table repeatedly:
found update xid 3433096759 <(343)%20309-6759> <(343)%20309-6759>

from before

relfrozenxid 3498422955

This is a primary, 9.6.8 to be exact.

This message is known to be associated with several data-corruption

bugs

we've fixed recently. You should update to current (9.6.10) sooner not
later. I don't think that will cure existing corruption for you, but
it should prevent things from getting worse.

Ok, thank you for that information I will do so. The

dump/truncate/restore

I mentioned in a later post seems to fix the corruption in a restored DB.
Given the nature of the bug do you think that would fix that table?

Yes, it's fairly likely that it does. You might even get around it by
just doing a no-op update of the whole table. Nevertheless you need to
upgrade, or more of that corruption can occur.

Indeed, that does seem to fix the issue and is much simpler, thank you. I'm
going to schedule this upgrade with all due haste.