Re: pg_upgrade and wraparound

On 2018-06-09 15:52:26 -0400, Tom Lane wrote:
> Adrian Klaver <adrian(dot)klaver(at)aklaver(dot)com> writes:
> > On 06/09/2018 03:46 AM, Alexander Shutyaev wrote:
> >> The upgrade operation failed after several hours with the following error:
> >> database is not accepting commands to avoid wraparound data loss in
> >> database with OID 0
>
> > Do you know which database has an OID of 0?
>
> Well, none do, so the correct question is what is passing an invalid
> database OID to the code that's complaining. This sure looks like a
> bug, though I'm not sure we have enough info to locate it.

It sure looks like ShmemVariableCache->oldestXidDB isn't initialized.
As far as I remember we just initialize that from a checkpoint at
startup. I suspect the issue is that pg_resetwal does:
if (set_xid != 0)
{
ControlFile.checkPointCopy.nextXid = set_xid;

/*
* For the moment, just set oldestXid to a value that will force
* immediate autovacuum-for-wraparound. It's not clear whether adding
* user control of this is useful, so let's just do something that's
* reasonably safe. The magic constant here corresponds to the
* maximum allowed value of autovacuum_freeze_max_age.
*/
ControlFile.checkPointCopy.oldestXid = set_xid - 2000000000;
if (ControlFile.checkPointCopy.oldestXid < FirstNormalTransactionId)
ControlFile.checkPointCopy.oldestXid += FirstNormalTransactionId;
ControlFile.checkPointCopy.oldestXidDB = InvalidOid;
}

but we have codepath that doesn't check for oldestXidDB being
InvalidOid. Not great.

Greetings,

Andres Freund