Thank you for your fix. It is really seems that this bug appears only when some autovacuum jobs are active.
Andrey Zhidenkov.
On Wed, Jul 30, 2014 at 7:09 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
I wrote: > However ... it looks to me like if the autovac launcher gets launched > forcibly because of a possible wraparound issue, it will keep running > normally after that even if autovacuum is off :-(.
Nope, scratch that: I missed the bit before the launcher's main loop, where it fires off one child and exits if it's launched when autovacuum is already false.
Still, if it does fire one child process, it looks like that child will vacuum everything that seems to need vacuuming in the target database. We should probably adjust that so that tables not in danger of wraparound get skipped.