Phoenix Kiula <phoenix.kiula@gmail.com> writes:
> Thanks, Gregory and Simon, for the very useful posts.
>
> I have increased the vacuum_cost_limit to 2000 for now, just to see if
> that has an impact. Hopefully positive.
Note that that was offhand speculation. Conventional wisdom is that it should
make things *worse* -- you're saying to process more pages between sleeping so
it'll use more i/o. I was speculating that you increased both
vacuum_cost_limit and vacuum_cost_delay proportionally it might use the i/o
more efficiently even though it's using the same amount of total bandwidth.
The more normal suggestion is to increase *vacuum_cost_delay* which tells it
to sleep longer between bits of work. Don't increase it too much or vacuum
will take forever. But if you increase it from 20 to 40 it should use half as
much i/o as bandwidth as now.
--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com
Ask me about EnterpriseDB's 24x7 Postgres support!