From: | Scott Marlowe <scott(dot)marlowe(at)gmail(dot)com> |
---|---|
To: | Jordan Tomkinson <jordan(at)moodle(dot)com> |
Cc: | pgsql-general(at)postgresql(dot)org |
Subject: | Re: High cpu usage after many inserts |
Date: | 2009-02-23 07:20:29 |
Message-ID: | dcc563d10902222320t74f66c26jcbde700044a5fb84@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
On Mon, Feb 23, 2009 at 12:18 AM, Jordan Tomkinson <jordan(at)moodle(dot)com> wrote:
>
>
> On Mon, Feb 23, 2009 at 4:08 PM, Scott Marlowe <scott(dot)marlowe(at)gmail(dot)com>
> wrote:
>>
>> One last thing. You were doing vacuum fulls but NOT reindexing, right?
>>
>> I quote from the document at google docs:
>> 13:50:00 vacuum full & analyze on all databases through pgadmin
>>
>> 1: Do you have evidence that regular autovacuum isn't keeping up?
>> 2: If you have such evidence, and you have to vacuum full, vacuum full
>> doesn't really shrink indexes all that well.
>>
>> For a heavily updated database, the 1, 2, 3 punch of autovacuum
>> (adjusted properly!), the background writer (adjusted properly)
>> smoothing things out, and the HOT updates reusing all that space
>> autovacuum is constantly reclaiming, meaning you should be able to
>> avoid routine vacuum fulls. It's made a huge difference in db
>> maintenance for me.
>>
>> Still I do find myself in vacuum full territory once or twice a year
>> (rogue update or something like that on a live database). If you do
>> have to vacuum full then reindex. OR cluster on your favorite index.
>
> I have no evidence of autovacuum not working, the manual full was done for
> purpose of elimination.
Oh, ok. If you're trying to make a fair benchmark, you should
probably reindex after vacuum full.
From | Date | Subject | |
---|---|---|---|
Next Message | Scott Marlowe | 2009-02-23 07:29:30 | Re: High cpu usage after many inserts |
Previous Message | Gerd König | 2009-02-23 07:19:13 | Re: PostgreSQL clustering with DRBD |