[20:55] <lifeless> wheee
[20:55] <lifeless> 17921  OOPS-10c6e2453422218f59fb3e8e621ea198  Person:+contactuser
[21:30] <lifeless> wgrant: https://lp-oops.canonical.com/oops.py/?oopsid=OOPS-bcba588b8f33bfa5da7c88e680763bf8
[23:17] <wgrant> lifeless: Interesting, obviously a cold cache somehow.
[23:17] <wgrant> lifeless: I suspect that GIN will be nice -- it's usually the FTI index that ends up cold.
[23:17] <wgrant> Since the entire set of Ubuntu bugs is kept very hot by the regular full index scans.
[23:21] <wgrant> lifeless: The only major repeatable timeouts are Person:+commentedbugs and Person:+bugs and their API equivalents for particularly prolific people (better than they used to be, but still bad), and things that use last_patch_uploaded, which I'm about to add to bugtaskflat.
[23:21] <lifeless> cool
[23:21] <lifeless> yesterday had 1.6K timeouts still
[23:21] <wgrant> Yeah, mostly not bug-related, though.
[23:21] <wgrant> Only a few hundred.
[23:21] <wgrant> Most of them being +commentedbugs
[23:22] <lifeless> kk
[23:22] <lifeless> is the root bugs api one fixed ?
[23:22] <wgrant> Ja
[23:22] <wgrant> Overrode it to [], timeouts went away and nobody has complained.
[23:24] <lifeless> hah
[23:24] <wgrant> Hopefully DF will be back today so I can fix Distribution:+search's query
[23:25] <wgrant> lifeless: All the +patches timeouts and most of Distribution:EntryResource:searchTasks will go away when I denorm latest_patch_uploaded.
[23:30] <wgrant> StevenK: You didn't revert that bad rev..?
[23:48] <StevenK> wgrant: I figured that since jml's QA was going to block it I could fix it better
[23:50] <wgrant> StevenK: jml's thing will be deployable in 10 hours, so you'd best be swift.
[23:51] <StevenK> wgrant: How best to fix it? self.findBranches().limit() ?
[23:51] <wgrant> That won't work.
[23:51] <StevenK> I didn't think so.
[23:51] <wgrant> The Stormiest way to do it is to use a subselect.
[23:51] <wgrant> Well, that's the SQLiest way to do it too
[23:51] <wgrant> UPDATE doesn't support LIMIT
[23:52] <StevenK> Ah, so it has to be WHERE id IS IN ...