From: | Mark Stosberg <mark(at)summersault(dot)com> |
---|---|
To: | pgsql-performance(at)postgresql(dot)org |
Subject: | Using the 8.2 autovacuum values with 8.1 |
Date: | 2007-02-22 21:53:52 |
Message-ID: | erl3ed$1mj9$1@news.hub.org |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
Thanks to everyone for the feedback about vacuuming. It's been very
useful. The pointers to the pgstattuple and Pgfouine tools were also
helpful.
I'm now considering the following plan for trying Autovacuuming again
with 8.1. I'd like any peer review you have to offer of the following:
1. First, I'll move the settings to match the defaults in 8.2. The ones
I noticed in particular were:
autovacuum_vacuum_threshold changes: 1000 -> 500
autovacuum_anayze_threshold changes: 500 -> 250
autovacuum_scale_factor changes: .4 -> .2
autovacuum_analyze_scale_factor changes .2 -> .1
2. Try the vacuum cost delay feature, starting with a 20ms value:
autovacuum_vacuum_cost_delay = 20
3. Immediately add a row to pg_autovacuum for a huge logging table that
would be too slow to vacuum usually. We'll still vacuum it once a week
for good measure by cron.
4. For good measure, I think I still keep the nightly cron entry that
does a complete vacuum analyze (except for that large table...).
Seem like a reasonable plan?
Mark
From | Date | Subject | |
---|---|---|---|
Next Message | ismo.tuononen | 2007-02-23 06:18:20 | Re: slow update on 1M rows (worse with indexes) |
Previous Message | Steinar H. Gunderson | 2007-02-22 20:29:20 | Re: Vacuum full very slow due to nonremovable dead rows...What makes the dead rows non-removable? |