Re: auto vaccum is dying

On Thu, Oct 2, 2014 at 12:34 PM, Jeff Janes <jeff(dot)janes(at)gmail(dot)com> wrote:

> On Wed, Oct 1, 2014 at 9:43 PM, Rodrigo Barboza <rodrigombufrj(at)gmail(dot)com>
> wrote:
>
>> Hello, I have a table that receives lots of updates and inserts.
>> Auto vaccum is always being cancelled on that table.
>>
>
> Do you have a scheduled task that clusters or reindexes the table?
>
> Newer versions of PostgreSQL will log the conflicting statement that
> caused the vacuum to cancel.
>
>
I have nothing scheduled, only auto vacuum, but with the default parameters.

>
>
>> One day the database went on standby and I had to act manually to recover.
>>
>
> I'm not sure what that means. Do you mean it stopped accepting commands
> to prevent "wrap around" data loss? Once autovacuum starts running on a
> table in "prevent wrap around", then it no longer voluntarily yields to
> other processes trying to take a conflicting lock.
>
>

Exactly, stopped to prevent wrap around. I think it was because auto vacuum
is being canceled.

>
>> What should I do to avoid auto vaccum cancel?
>>
>
> If you have scheduled jobs that do something on the table that requires a
> lock which conflicts with autovac, then you might want to include a manual
> VACUUM in that job.
>
> Also, what full version are you running?
>
>
I am running postgres 9.1.4 with default auto vacuum parameters. I have
only a scheduled job that runs delete for old tuples. Sometimes it a lot of
tuples. Beside that, no other tasks.

Cheers,
>
> Jeff
>