Re: Postgresql 'eats' all mi data partition - Mailing list pgsql-bugs

From Tomas Szepe
Subject Re: Postgresql 'eats' all mi data partition
Date
Msg-id 20030926184830.GD26641@louise.pinerecords.com
Whole thread Raw
In response to Re: Postgresql 'eats' all mi data partition  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: Postgresql 'eats' all mi data partition  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-bugs
> [tgl@sss.pgh.pa.us]
>
> Tomas Szepe <szepe@pinerecords.com> writes:
> > We have a production database that happens to receive several
> > thousand row updates per minute.  We VACUUM ANALYZE every four
> > hours with max_fsm_pages set to 2100000, and it's no use.
>
> Have you spent any time determining exactly where the problem is?
> I'm suspicious that it's an index-bloat issue.

I'm afraid so.  The weird thing is, we once tried running
VACUUM FULL followed by REINDEX and although the process
took ages to complete, it didn't seem to help either. :(

I'll post whatever debug data I'm asked for, just don't make
me run VACUUM FULL or REINDEX again please.  (A full redump
is 10-20 times faster. :O)

Thanks,
--
Tomas Szepe <szepe@pinerecords.com>

pgsql-bugs by date:

Previous
From: Tomas Szepe
Date:
Subject: Re: Postgresql 'eats' all mi data partition
Next
From: Tom Lane
Date:
Subject: Re: Postgresql 'eats' all mi data partition