Re: Incomplete freezing when truncating a relation during vacuum

On 2013-12-01 13:33:42 +0100, Andres Freund wrote:
> On 2013-11-27 14:53:27 -0500, Noah Misch wrote:
> > How would you characterize the chances of this happening with default
> > *vacuum_freeze_*_age settings?  Offhand, it seems you would need to encounter
> > this bug during each of ~10 generations of autovacuum_freeze_max_age before
> > the old rows actually become invisible.
> 
> On second thought, it's quite possible to see problems before that
> leading to more problems. A single occurance of such a illegitimate
> increase in relfrozenxid can be enough to cause problems of a slightly
> different nature.
> As relfrozenxid has been updated we might now, or after vacuuming some
> other tables, become elegible to truncate the clog. In that case we'll
> get ERRORs about "could not access status of transaction" if the tuple
> hasn't been fully hinted when scanning it later.

And indeed, a quick search shows up some threads that might suffer from
it:
BD7EE863F673A14EBF99D95562AEE15E44B1DA71@digi-pdc.digitilitiprod.int
CAAzPmNxfDrV72wDmBEv5tcQOByE_wvGSeqRkQj0FizXmCYyaPQ@mail.gmail.com
CAK9oVJwvAZLmdMrHMPg1+s37z16j+BZ8FbarZSpmrHsXxH-4GQ@mail.gmail.com

Greetings,

Andres Freund

-- Andres Freund                       http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services