Re: found xmin from before relfrozenxid on pg_catalog.pg_authid

On Tue, Jun 19, 2018 at 12:53 PM, Andres Freund <andres(at)anarazel(dot)de> wrote:

> Hi,
>
> On 2018-06-19 10:26:26 -0300, Matheus de Oliveira wrote:
> > Hello friends.
> >
> > On Tue, Jun 12, 2018 at 3:31 PM, Andres Freund <andres(at)anarazel(dot)de>
> wrote:
> >
> > >
> > > On 2018-06-11 17:39:14 -0700, Andres Freund wrote:
> > > > I plan to go over the change again tomorrow, and then push. Unless
> > > > somebody has comments before then, obviously.
> > >
> > > Done.
> > >
> > >
> > Sorry to bother about this, but do you have any plan to do the minor
> > release before planned due to this bug?
>
> Unclear at this point. There's been discussion about it, without coming
> to a conclusion.
>
>
Ok. Thank you for the information.

I really hope you decide to release it soon, I'm very afraid about the
users that have hit the bug but haven't noticed the issue.

>
> > I'm pondering what is the best option to avoid a forced shutdown of this
> > server:
> > - should I just wait for a release (if it is soon, I would be fine)?
> > - build PG from the git version by myself?
> > - or is there a safer workaround to the problem? (not clear to me if
> > deleting the `global/pg_internal.init` file is really the way to go, and
> > the details, is it safe? Should I stop the server, delete, start?)
>
> You should first make sure it's actually this problem - which tables are
> holding back the xmin horizon?

How can I be sure? The tables are `pg_authid` and `pg_auth_members`, the
following message is logged every minute (which I belive is due to
`autovacuum_naptime`, which is using the default of 1 minute):

ERROR: found xmin 3460221635 from before relfrozenxid 1245633870
CONTEXT: automatic vacuum of table "template0.pg_catalog.pg_authid"
ERROR: found xmin 3460221635 from before relfrozenxid 1245633870
CONTEXT: automatic vacuum of table
"template0.pg_catalog.pg_auth_members"

Do you need controldata or more info to validate it?

> After that, yes, deleting the
> global/pg_internal.init file is the way to go, and I can't think of a
> case where it's problematic, even without stopping the server.
>
>
I'll try that and get back to you if it worked or not. Thank you for the
confirmation.

And thank you for all clarifications.

Best regards,
--
Matheus de Oliveira