"Kevin Grittner" <Kevin.Grittner@wicourts.gov> wrote:
> When we first started using PostgreSQL we initially had problems
> with bloat and the autovacuum process started creating significant
> load. Our first reaction was to make autovacuum less aggressive,
> but this just made things worse. The counter-intuitive step of
> making autovacuum very aggressive actually prevented bloat to the
> point where autovacuum wound up creating less load. Don't be shy
> about running it frequently and at low thresholds; but if it does
> have a performance impact, a small autovacuum_cost_delay (I've never
> needed more than 10ms) will probably solve the issue.
Cheers, I might give that a go. I've set the cost_delay to 10 already, so
will decrease the thresholds more and see what happens.
John
