Dne 6.4.2011 17:33, Kevin Grittner napsal(a):
> "Maria L. Wilson" <Maria.L.Wilson-1@nasa.gov> wrote:
>
>> Autovacuum is not running - but regular vacuums are being done
>> twice daily.
>
> Is the ANALYZE keyword used on those VACUUM runs? What version of
> PostgreSQL is this. If it's enough to need fsm settings, do you run
> with the VERBOSE option and check the end of the output to make sure
> they are set high enough?
Why do you think the problem is related to stale stats? It seems to me
fairly accurate - see the explain analyze in the first post). All the
nodes are less than 1% off (which is great), except for the last hash
join that returns 336949 rows instead of 924675 expected rows.
Maybe I'm missing something, but the stats seem to be quite accurate and
there is just very little dead tuples I guess.
regards
Tomas