[02:07]  * wgrant throws rocks at SKS
[02:10] <StevenK> wgrant: Oh?
[02:10] <wgrant> Its 404 message has changed
[02:10] <wgrant> That's why we're getting oopses from it
[02:10] <wgrant> But it doesn't actually crash the page
[02:10] <StevenK> What's one of the OOPSes?
[02:10] <wgrant> OOPS-7d16107fdc9d638d26333ce4668d524e
[02:12] <StevenK> I thought we dealt with both 404 and 500?
[02:12] <wgrant> We do
[02:12] <wgrant> But if the error message isn't plausible, we raise a soft OOPS
[02:12] <StevenK> Ah
[02:13] <StevenK> And they changed it
[02:13] <wgrant> Yes
[02:24] <StevenK> wgrant: So I can't find any OOPSes for "[^:]DistroSeries:+localpackagediffs"
[02:24] <wgrant> StevenK: Probably because nobody uses it, potentially because it doesn't work
[02:31] <wgrant> StevenK: https://code.launchpad.net/~wgrant/launchpad/sks-404-msg/+merge/155647
[02:49] <StevenK> wgrant: No fair landing something
[02:50] <StevenK> wgrant: Meh, that was probably fair enough to be a self-review. r=me
[02:52] <wgrant> Probably
[02:52] <wgrant> But the r=wgrant in the log was getting a bit boring
[02:53] <StevenK> Haha
[02:53] <StevenK> Only because the last 9 landings on the trot have been mine
[02:56] <StevenK> wgrant: It does load, though
[02:57] <StevenK> Of course, now it times out for me.
[02:57] <wgrant> StevenK: What or who loads?
[02:57]  * StevenK shoots Murphy in the face.
[02:57] <wgrant> Oh, +localpackagediffs?
[02:57] <StevenK> https://launchpad.net/ubuntu/raring/+localpackagediffs
[02:58] <wgrant> 7.87s
[02:58] <wgrant> Nice
[02:59] <StevenK> 52 queries/external actions issued in 7.43 seconds
[02:59] <StevenK> Yeah
[03:00] <StevenK> It spends 300ms looking up the DSDs, and then 3.5seconds grabbing the publications
[03:00] <wgrant> Yeah
[03:00] <wgrant> Explaining now
[03:00] <wgrant> It's using SPR.spn, which is probably the main problem
[03:00] <wgrant> I missed another one :(
[03:01] <StevenK> Can you also explain why it runs the query four times?
[03:01] <StevenK> It's Launchpad, not Santa Claus.
[03:01] <wgrant> Hey, you were on the team that wrote that page, not me :)
[03:01] <wgrant> But I'd get an OOPS and check what calls it
[03:01] <wgrant> Because that's hilarious
[03:02] <wgrant>  Total runtime: 90667.543 ms
[03:02] <wgrant> ooh yeah
[03:02] <wgrant> 350ms->65ms by dropping the SPR join
[03:03] <StevenK> wgrant: http://pastebin.ubuntu.com/5651202/
[03:03] <wgrant> Oh right
[03:03] <StevenK> wgrant: Just check that I'm not going insane, and it is
[03:03] <wgrant> This is the derpquery
[03:03] <wgrant> That has the completely strange anti-optimisation technique
[03:03] <wgrant> To try to prevent it from using the index that is probably now optimal
[03:03] <wgrant> AND SourcePackagePublishingHistory.archive+0 = Archive.id
[03:04] <wgrant> 3ms
[03:04]  * wgrant enbranchinates
[03:04] <StevenK> Ah
[03:04] <StevenK> I was going to do that
[03:04] <StevenK> Beh eh
[03:04] <wgrant> No'
[03:04] <wgrant> I get the trivial ones today, to let me escape from translations crap :)_
[03:05] <StevenK> And all four of them are different
[03:05] <wgrant> Meh
[03:05] <wgrant> I'll make them not suck
[03:05] <StevenK> derived_series vs parent_series
[03:05] <wgrant> Then we can separately eliminate the dupes
[03:05] <wgrant> Ah
[03:05] <wgrant> Anyway
[03:05] <wgrant> Not so bad once they're a direct index scan
[03:05] <wgrant> What's the traceback?
[03:05] <wgrant> Not sure which func it is
[03:06] <wgrant> I guess I could just grep for +0...
[03:06] <StevenK> That +0 might be a red herring
[03:06] <StevenK> I was pulling it from the query log
[03:07] <wgrant> No
[03:07] <wgrant> I remember the query
[03:07] <StevenK> eager_load_dsds
[03:07] <wgrant> It was deliberately used as an anti-optimisation technique
[03:08] <wgrant> And the archive join is just pointless
[03:08] <StevenK> Twitch
[03:08] <StevenK> It's limited to purpose PRIMARY, though
[03:08] <wgrant>         # XXX: GavinPanella 2011-06-23 bug=801097: The + 0 in the condition
[03:09] <wgrant>         # below prevents PostgreSQL from using the (archive, status) index on
[03:09] <wgrant>         # SourcePackagePublishingHistory, the use of which results in a
[03:09] <wgrant> get the ID :)
[03:09] <wgrant>         # terrible query plan.
[03:09] <wgrant> Yeah
[03:09] <wgrant> But you know what we can do instead...
[03:09] <StevenK> Fetch self.context.main_archive.id ?
[03:09] <StevenK> wgrant: Anyway, damn you for stealing my bug
[03:10] <wgrant> Hmm
[03:10] <StevenK> Now I have to find another.
[03:10] <wgrant> May not be able to do this,a ctually
[03:10] <wgrant> Because it cares about the parent
[03:10] <StevenK> "#899003 deadlock detected when importing .po file" is likely going to die as a case of your chainsawing ?
[03:10] <_mup_> Bug #899003: deadlock detected when importing .po file <lp-translations> <oops> <rosetta-imports> <Launchpad itself:Triaged> < https://launchpad.net/bugs/899003 >
[03:11] <wgrant> Unlikely
[03:11] <wgrant> Unless I rewrite rosetta
[03:11] <StevenK> Pity
[03:11] <StevenK> Maybe I should destroy the puller
[03:11] <StevenK> That will keep me out of trouble for a long while
[03:11] <wgrant> Heh
[03:11] <StevenK> But I have no idea where jelmer got up to
[03:11] <wgrant> A very very long while
[03:12] <StevenK> We can't just transition the branches of type MIRRORED to code imports and destroy it?
[03:12] <wgrant> You can get rid of mirrored branches just by migrating the existing ones, really, but that won't obviate the need for the puller.
[03:12] <wgrant> The puller is still used for imports
[03:13] <StevenK> I'm thinking of the broken netstrings critical
[03:13] <StevenK> Migrating MIRRORED branches and dropping the type might close it, or at least drop it to Low
[03:14] <wgrant> I don't think so
[03:32] <wgrant> StevenK: https://code.launchpad.net/~wgrant/launchpad/bug-1160461/+merge/155651
[03:33] <StevenK> wgrant: v = ? Really?
[03:33] <wgrant> Really!
[03:33] <wgrant> I can change it if you want
[03:34] <StevenK> I would prefer view =
[03:34] <StevenK> And you're not trying to force a long line to be short enough
[03:36] <wgrant> wtf
[03:36] <wgrant> dsd indices are terrible
[03:40] <wgrant> StevenK: Fixed.
[03:41] <StevenK> I just saw
[03:41] <StevenK> You're a terrible person
[03:41] <StevenK> r=me
[03:42] <wgrant> Thanks
[03:44] <wgrant> And the 1s DSD-finding query is now mysteriously 2ms
[03:44] <wgrant> (* may actually be index, not actual mysteriousness)
[03:45] <StevenK> Haha
[03:59] <wgrant> StevenK: https://dogfood.launchpad.net/ubuntu/quantal/+localpackagediffs
[03:59] <wgrant> And that's without the DSD index
[03:59] <wgrant> Still slow, but that's hopefully just because DF is DF
[04:10] <wgrant> StevenK: https://code.launchpad.net/~wgrant/launchpad/localpackagediffs-rargh/+merge/155657
[04:17] <StevenK> wgrant: What's the index out of interest?
[04:19] <wgrant> (derived_series, status, difference_type)
[04:19] <wgrant> Not going to add it for now
[04:19] <wgrant> But it's something we might want to look at later
[04:20] <StevenK> wgrant: Going to link 801097 to the MP?
[04:20] <wgrant> A good point.
[04:21] <StevenK> wgrant: r=me
[04:21] <wgrant> Thanks.
[04:26]  * StevenK glares at the critical list
[04:27] <StevenK> wgrant: I do wonder about a BugJob that does notification
[04:27] <StevenK> But I worry that I'd get murdered
[04:30] <wgrant> StevenK: Sort of
[04:31] <StevenK> I'd get sort of murdered?
[04:31] <wgrant> Deferring recipient calculation by storing the criteria in a table would work
[04:31] <wgrant> But it's a fair bit of work
[04:32] <StevenK> wgrant: Suggestions for what I should look at, then?
[04:32] <wgrant> Um
[04:32] <wgrant> Stuff.
[04:32] <wgrant> And things :)
[04:32] <StevenK> Thanks for being so very helpful. :-P
[04:33] <StevenK> I do wonder how txlongpoll is run, WRT 960868
[04:33] <wgrant> It might almost be worth making another pass at the loggerhead criticals
[04:33] <wgrant> You can probably fix several of those in one day
[04:33] <StevenK> But it's loggerhead ...
[05:06] <StevenK> wgrant: How do I land stuff against loggerhead?
[05:07] <StevenK> Hmmm, we're 7 revs behind
[05:17] <wgrant> Should just need to commit directly to trunk
[05:17] <wgrant> Check the recent history
[05:31] <StevenK> wgrant: We have no codebounce on staging/qas ?
[05:32] <wgrant> StevenK: Yes, it's there
[13:18]  * jelmer waves
[13:18] <czajkowski> jelmer: boo
[13:18] <jelmer> StevenK: migrating MIRRORED branches to code imports should do the trick
[13:19] <jelmer> StevenK: I don't think there was a reason for not doing that, apart from the time it would take to manage that transition
[13:19] <jelmer> czajkowski: hey hey :)
[23:54] <wgrant> StevenK: Can you QA your loggerhead thing?
[23:56] <wgrant> StevenK: And shouldn't that LH task be Fix Committed?