Ok, I'll keep track of how many pages change between updates and play with the max_fsm_pages value. I'll post my findings.
--Michael
Tom Lane wrote:
"Michael G. Martin" <michael@vpmonline.com> writes:
That's what I was beginning to think from reading. I just wasn't sure
of the correlation between MAX_FSM_PAGES and the actual pages occupied
by the table after a full vacuum. 1:1 ?? Any general rules?
1:1 would be worst case, probably. The real question is how many pages
get dirtied between vacuums --- so you can trade off more frequent
vacuums for a smaller free space map.
We don't have a lot of experience yet with tuning that parameter (if
we did, there'd be advice in the manual). Please report anything
you learn from experimentation.
regards, tom lane