"Joel Fradkin" <jfradkin(at)wazagua(dot)com> writes:
> Running this explain on windows box, but production on linux both 8.0.1
Are you using any nondefault optimizer settings? The vast bulk of the
time is going into the indexscan on tblassociate (almost 42 out of the
48 seconds), and I'm a bit surprised it didn't choose a seqscan and sort
instead. Or even more likely, forget the merge joins altogether and use
hash joins --- the other tables are plenty small enough to fit in hash
tables.
regards, tom lane