Re: growing disk usage problem: alternative solution? - Mailing list pgsql-general

From Vivek Khera
Subject Re: growing disk usage problem: alternative solution?
Date
Msg-id DED7310D-6DE3-468C-8687-1304522D3AE0@khera.org
Whole thread Raw
In response to Re: growing disk usage problem: alternative solution?  (Bill Moran <wmoran@potentialtech.com>)
List pgsql-general
On Jun 26, 2007, at 3:31 PM, Bill Moran wrote:

> VACUUM FULL and REINDEX are not required to maintain disk usage.
> Good old-
> fashoned VACUUM will do this as long as your FSM settings are high
> enough.
>

I find this true for the data but not necessarily for indexes.  The
other week I reindexed a couple of O(100,000,000) row tables and
shaved about 20Gb of index bloat.  Those tables are vacuumed
regularly, but we do a large data purge every few weeks.  I think
that causes some issues.  I'm running 8.1.

To mitigate the downtime, we make use of the fact that we have live
replica of the database on similarly capable hardware so we bring the
replica offline, reindex it, bring it back up, move the "master" to
it, then reindex the other server, and move the master back.


pgsql-general by date:

Previous
From: Alvaro Herrera
Date:
Subject: Re: AutoVacuum Behaviour Question
Next
From: "Gurjeet Singh"
Date:
Subject: Re: a JOIN on same table, but 'slided over'