From: | Alvaro Herrera <alvherre(at)commandprompt(dot)com> |
---|---|
To: | Ron Mayer <rm_pg(at)cheapcomplexdevices(dot)com> |
Cc: | jd(at)commandprompt(dot)com, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Simon Riggs <simon(at)2ndQuadrant(dot)com>, Pg Hackers <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: new GUC var: autovacuum_process_all_tables |
Date: | 2009-02-06 18:04:47 |
Message-ID: | 20090206180447.GC3089@alvh.no-ip.org |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Ron Mayer wrote:
> Joshua D. Drake wrote:
> >> The main remaining use-case seems to me to make vacuuming work adhere
> >> to some business-determined schedule, hence maintenance windows seem
> >> like the next thing to do.
> > Also agreed.
>
> Somewhat agreed - since in many cases the business-determined schedule
> is just a rough estimate of measurable attributes of the machine. When
> we say "vacuum between midnight and 5am" we often actually mean "vacuum
> when the I/O subsystem has bandwidth to spare and the machine's otherwise
> lightly loaded, and we guess that means late at night".
The current state of the system is not necessarily a good indicator of
the immediately future state. If we were to collect history (I/O load
versus time of day and day of week) that would be another matter, but
I'm not sure that's a productive use of our development time.
If we could guess what's going to be the load of the server in the near
future, that would help scheduling considerable. Since we can't, we
have to trust that the user can provide an educated guess, which is what
"maintenance windows" are all about.
--
Alvaro Herrera http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support
From | Date | Subject | |
---|---|---|---|
Next Message | Robert Haas | 2009-02-06 18:08:01 | Re: A single escape required for log_filename |
Previous Message | Bruce Momjian | 2009-02-06 17:44:48 | Re: A single escape required for log_filename |