Re: Vacuum looping? - Mailing list pgsql-performance

From Decibel!
Subject Re: Vacuum looping?
Date
Msg-id D494F6AB-F9C7-4A61-B050-181BEE320312@decibel.org
Whole thread Raw
In response to Re: Vacuum looping?  ("Steven Flatt" <steven.flatt@gmail.com>)
List pgsql-performance
On Jul 30, 2007, at 9:04 AM, Steven Flatt wrote:
> On 7/28/07, Jim C. Nasby <decibel@decibel.org> wrote: What are your
> vacuum_cost_* settings? If you set those too aggressively
> you'll be in big trouble.
>
> autovacuum_vacuum_cost_delay = 100

Wow, that's *really* high. I don't think I've ever set it higher than
25. I'd cut it way back.

> autovacuum_vacuum_cost_limit = 200
>
> These are generally fine, autovacuum keeps up, and there is minimal
> impact on the system.
>
> vacuum_cost_delay = 100
> vacuum_cost_limit = 1000
>
> We set this cost_limit a little higher so that, in the few cases
> where we have to intervene manually, vacuum runs faster.

IIRC, when the cost delay was initially introduced (8.0), someone did
testing and decided that the cost limit of 200 was optimal, so I
wouldn't go changing it like that without good reason.

Normally, I'll use a delay of 10ms on good disk hardware, and 20ms on
slower hardware.
--
Decibel!, aka Jim Nasby                        decibel@decibel.org
EnterpriseDB      http://enterprisedb.com      512.569.9461 (cell)



pgsql-performance by date:

Previous
From: Tom Lane
Date:
Subject: Re: disk filling up
Next
From: Henrik Zagerholm
Date:
Subject: Seq scan on join table despite index and high statistics