"Mason Hale" <masonhale@gmail.com> writes:
>> It looks like "read a page, sleep for 80 milliseconds, repeat".
That's what it looks like to me too.
>> I'd look at your settings for autovacuum_vacuum_cost_limit /
>> autovacuum_vacuum_cost_delay, and maybe the contents of pg_autovacuum.
> autovacuum_vacuum_cost_limit = -1
> autovacuum_vacuum_cost_delay = 20ms
That process is *clearly* not using those vacuum cost parameters ---
it's evidently using a delay of 80ms and some completely over-aggressive
cost settings that're making it sleep for each single page read.
So you need to find out where those whacked-out values are coming from.
pg_autovacuum might be a likely source. Or maybe you just forgot a
SIGHUP after a recent change to postgresql.conf?
regards, tom lane
