Re: Query times change by orders of magnitude as DB ages - Mailing list pgsql-performance

From Richard Neill
Subject Re: Query times change by orders of magnitude as DB ages
Date
Msg-id 4B0D1EDD.3070909@cam.ac.uk
Whole thread Raw
In response to Re: Query times change by orders of magnitude as DB ages  (Matthew Wakeling <matthew@flymine.org>)
Responses Re: Query times change by orders of magnitude as DB ages  (Matthew Wakeling <matthew@flymine.org>)
List pgsql-performance

Matthew Wakeling wrote:
> On Sun, 22 Nov 2009, Richard Neill wrote:
>> Worse still, doing a cluster of most of the tables and vacuum full
>> analyze
>
> Why are you doing a vacuum full? That command is not meant to be used
> except in the most unusual of circumstances, as it causes bloat to indexes.

We'd left it too long, and the DB was reaching 90% of disk space. I
didn't realise that vacuum full was ever actively bad, only sometimes
unneeded. I do now - thanks for the tip.

>
> If you have run a cluster command, then running vacuum full will make
> the table and index layout worse, not better.
>

So, having managed to bloat the indexes in this way, what can I do to
fix it? Will a regular vacuum do the job?

Richard


pgsql-performance by date:

Previous
From: Richard Neill
Date:
Subject: Re: Query times change by orders of magnitude as DB ages
Next
From: Matthew Wakeling
Date:
Subject: Re: Query times change by orders of magnitude as DB ages