Re: BUG #7546: Backups on hot standby cancelled despite hot_standby_feedback=on - Mailing list pgsql-bugs

From Stuart Bishop
Subject Re: BUG #7546: Backups on hot standby cancelled despite hot_standby_feedback=on
Date
Msg-id CADmi=6NTPeRrBq-F=FxteGG_254aJ-bvaLHJ2d5O3erMfB=_ZQ@mail.gmail.com
Whole thread Raw
In response to Re: BUG #7546: Backups on hot standby cancelled despite hot_standby_feedback=on  (Andres Freund <andres@2ndquadrant.com>)
Responses Re: BUG #7546: Backups on hot standby cancelled despite hot_standby_feedback=on  (Stuart Bishop <stuart@stuartbishop.net>)
List pgsql-bugs
On Wed, Sep 19, 2012 at 6:56 PM, Andres Freund <andres@2ndquadrant.com> wrote:

> You see buffer pin errors or you see the relation errors? Those arent the same.
> So if you see no buffer pin errors but relation lock ones in production your
> way to reproduce the issue isn't telling us very much ;)

On production I have seen both. I have only been able to confirm and
isolate the trigger of the buffer pin errors, and thus this bug
report. I mentioned the rarer relation lock error just in case it was
relevant.

> Do you have DDL logging enabled so you could investigate that?
>
> I really wonder why we don't bother logging the dbOid/relOid pair before we
> cancel transactions due to lock conflicts. That should make investigating such
> issues considerably easier...

I do have DDL logging enabled. I've waded through a two hour period on
the primary where the problem must have occurred and can see no DDL
except for the creation and dropping of temporary tables. My parser is
unfortunately grep and my eyeballs to filter out the noise.


--
Stuart Bishop <stuart@stuartbishop.net>
http://www.stuartbishop.net/

pgsql-bugs by date:

Previous
From: Andres Freund
Date:
Subject: Re: BUG #7546: Backups on hot standby cancelled despite hot_standby_feedback=on
Next
From: max@suse.de
Date:
Subject: BUG #7561: syslogger doesn't close stdout and stderr