Dinesh Bhandary wrote:
> What you are saying is vacuum without tx wraparound wonât cause the
> blocking issue we had experienced.
Yep, that's correct. You're only seeing an issue because vacuum (manual
or auto) is trying to freeze the tuple. If the tuple is not considered
old enough to need freezing it will not cause the system to stall. Once
you get the oldestMulti value fixed, the freezing routine will be able
to look at the tuple, notice that the multixact in it is older than the
upgrade, and decide it can just rip it out without trying to "resolve"
it.
> If thatâs the case I can just change the auto vacuum on ( which can be
> changed on the fly), and I had already bumped up max_freeze_settings to
> 500 mill for now. We will change that to lower number in the next
> maintenance window.
Right -- AFAICS that should work.
--
Ãlvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services