On March 22, 2017 3:54:07 PM GMT+01:00, Tom Lane <t...@sss.pgh.pa.us> wrote: >Andres Freund <and...@anarazel.de> writes: >> On 2017-03-22 10:14:14 -0400, Tom Lane wrote: >>> (2) How you gonna update this in vacuum? It cannot do a >transactional >>> update. > >> I think we can just do that in a separate transaction, at the tail >end >> of vacuum_rel() - if we crash just before that, not that much is >lost. > >Blech. What if someone's queued an exclusive lock request on the >table?
Point - could do that with a session lock if needed. >Actually though, maybe you could get away with updating the last-vacuum >field nontransactionally. I think we're already updating some of the >other pg_class fields nontransactionally. We do, reltuples/pages, relfrozenxid and such. That should work too, the whole inplace business ain't pretty, but one more column won't make it much worse. Andres -- Sent from my Android device with K-9 Mail. Please excuse my brevity.