On Tue, Feb 27, 2007 at 12:12:22PM -0500, Matthew T. O'Connor wrote:
> Jim C. Nasby wrote:
> >On Tue, Feb 27, 2007 at 12:00:41AM -0300, Alvaro Herrera wrote:
> >>Jim C. Nasby wrote:
> >>
> >>>The advantage to keying this to autovac_naptime is that it means we
> >>>don't need another GUC, but after I suggested that before I realized
> >>>that's probably not the best idea. For example, I've seen clusters that
> >>>are running dozens-hundreds of databases; in that environment you really
> >>>need to turn naptime way down (to like a second). In that case you
> >>>wouldn't want to key to naptime.
> >>Actually, I've been thinking that it would be a good idea to change the
> >>semantics of autovacuum_naptime so that it means the average time to
> >>start a worker in any given database. That way, the time between
> >>autovac runs is not dependent on the number of databases you have.
> >
> >BTW, another issue that I don't think we can ignore: we actually need to
> >do this on a per-tablespace level, or at least have the ability to
> >disable or somehow limit it. While it's not common, there are users that
> >run a hundred or more databases in a single cluster; it would be ugly if
> >we suddenly had 100 vacuums trying to run on the same set of drives
> >concurrently.
>
> I think we all agree that autovacuum needs to become tablespace aware at
> some point, but I think that is further down the line, we're having
> enough trouble figuring things out without that additional complication.
Sure, we just need a way to disable the multiple autovac daemon stuff
then.
--
Jim Nasby jim@nasby.net
EnterpriseDB http://enterprisedb.com 512.569.9461 (cell)
