Comment 12 for bug 723999

Revision history for this message
Robert Collins (lifeless) wrote : Re: [Bug 723999] Re: BugNomination:+edit-form structural subscriptions taking 4.8 seconds during nomination editing POST

Ok, this suggests it is moderately cheap (4000 estimated cost) - note
that in a cold cache situation that is several seconds on its own.

The plan is very complex, but wgrant pointed out a key element: there
are only -4- subscription filters on qastaging at the moment, so we're
not seeing anything matching what prod is doing.

e.g.
                                       -> Hash (cost=1.02..1.02
rows=2 width=8)
                                             -> Seq Scan on
bugsubscriptionfilter (cost=0.00..1.02 rows=2 width=8)

is reading every bugsubscription filter and hashing it with no
condition - its then filtering *after* doing the hash join. We should
expect this to change radically if we put the filters we have on prod
into the staging environment - 24K worth.