Re: vacuum full questions - Mailing list pgsql-admin

From Alvaro Herrera
Subject Re: vacuum full questions
Date
Msg-id 20090818224859.GA5890@alvh.no-ip.org
Whole thread Raw
In response to Re: vacuum full questions  ("Kevin Grittner" <Kevin.Grittner@wicourts.gov>)
List pgsql-admin
Kevin Grittner wrote:
> Kevin Kempter <kevink@consistentstate.com> wrote:

> > 2) can I safely kill the vacuum full and do a dump, drop table,
> > restore instead?
>
> Killing a VACUUM FULL can leave the table or its indexes more bloated
> than when you started, but it should have no other negative impact.  I
> have generally had to kill my attempts to VACUUM FULL and resort to
> other techniques to deal with extreme bloat.

Note that cancelling a VACUUM FULL is known to cause a database restart
(a.k.a. PANIC) if the circumstances are right, viz. that the initial
transaction committed.  This is an unfixable bug (short of rewriting
vacuum full from scratch) and has been reported in the wild several
times.

--
Alvaro Herrera                                http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc.

pgsql-admin by date:

Previous
From: "Kevin Grittner"
Date:
Subject: Re: vacuum full questions
Next
From: "Donald Fraser"
Date:
Subject: Partial WAL file replay before fail-over startup