Steven Flatt escribió:
> Thanks everyone. It appears that we had hacked the 502.pgsql script for our
> 8.1 build to disable the daily vacuum. I was not aware of this when
> building and upgrading to 8.2.
>
> So it looks like for the past two weeks, that 36 hour db-wide vacuum has
> been running every 24 hours. Good for it for being reasonably non-intrusive
> and going unnoticed until now. :)
Looks like you have plenty of spare I/O ;-)
> Although apparently not related anymore, I still think it was a good move to
> change autovacuum_freeze_max_age from 200 million to 2 billion.
Absolutely not related. Also note that
1. autovacuum is not able (in 8.2 or older) to have more than one task
running
2. autovacuum in 8.2 doesn't ever launch database-wide vacuums. As of
8.2 it only vacuums tables that are in actual danger of Xid
wraparound (according to pg_class.relfrozenxid); tables that had been
vacuumed the day before would not need another vacuum for Xid
purposes (though if you had modified the table to the point that it
needed another vacuum, that would be another matter). Unless you
consumed 200 million (or 2 billion) transactions during the day, that
is.
--
Alvaro Herrera http://www.amazon.com/gp/registry/5ZYLFMCVHXC
"The only difference is that Saddam would kill you on private, where the
Americans will kill you in public" (Mohammad Saleh, 39, a building contractor)