On Jul 3, 2007, at 3:36 PM, Tom Lane wrote:
> "Jim C. Nasby" <decibel@decibel.org> writes:
>> On Mon, Jul 02, 2007 at 11:19:12PM -0400, Tom Lane wrote:
>>> Is there a reason to say anything beyond "use autovac"?
>> There is; I know that things like web session tables aren't
>> handled very
>> well by autovacuum if there are any moderately large tables (anything
>> that will take more than a few minutes to vacuum). Eventually we
>> should
>> be able to accommodate that case with multiple workers, but we'll
>> need a
>> mechanism to ensure that at least one worker doesn't get tied up in
>> large vacuums.
>
> And which part of that do you think isn't resolved in 8.3?
It's still possible to tie up all autovac workers in large tables,
though of course it's now far less likely.
BTW, +1 to dropping the thresholds to a very low value. 0 might be
pushing it, but 10 or 20 certainly doesn't sound absurd.
--
Jim Nasby jim@nasby.net
EnterpriseDB http://enterprisedb.com 512.569.9461 (cell)