Re: BUG #5759: Autovacuum cost limit trends to 1, using default config

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: "Simon Poole" <postgresql(at)simon(dot)themalago(dot)net>
Cc: pgsql-bugs(at)postgresql(dot)org, Alvaro Herrera <alvherre(at)commandprompt(dot)com>
Subject: Re: BUG #5759: Autovacuum cost limit trends to 1, using default config
Date: 2010-11-20 02:28:11
Message-ID: 28302.1290220091@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-bugs

"Simon Poole" <postgresql(at)simon(dot)themalago(dot)net> writes:
> With the default autovacuum settings (3 workers, and default cost_limit of
> 200), all three workers start with a cost-limit of 66 each, but after each
> table is completed, the cost-limit is reduced until they're all running with
> a cost-limit of 1.

Yeah, I was able to reproduce this, more or less. It looks like the
problem is that the cost rebalancing algorithm uses VacuumCostLimit
as an input, and supposes that's constant ... but after the first table
it's actually been reduced by a previous iteration (assuming there's
more than one worker). So you can end up computing a smaller value
on the next round. Lather, rinse, repeat.

Will fix. Thanks for the report!

regards, tom lane

In response to

Browse pgsql-bugs by date

  From Date Subject
Next Message dp 2010-11-22 08:04:03 BUG #5761: In 'dblink' function connection string truncated
Previous Message Kevin Grittner 2010-11-19 19:10:55 Re: BUG #5758: Error during backup