[00:00] Er, 23000 [00:01] 4000 queries to 10 for trusty amd64 [00:01] Score [00:04] Tempting to get this deployed first thing tomorrow and find out. [00:04] The actual query time is negligible, so we're really down to Python optimisation now if we want to go further. [00:04] And at that point you might as well skip the Storm objects, since they could easily be 30% of the remaining overhead. [00:10] The SSDs should also help a lot with the cold cache case, but if not we can pull some tricks with index-only scans. [00:12] Yeah, I guess my 17043 runs weren't really cold cache to start with [00:15] cron.ppa's two publish-distro runs combined averaged 656 seconds in the 99 runs in today's log file [01:41] cprov: https://code.launchpad.net/~wgrant/launchpad/bug-1083709/+merge/222895 [01:57] wgrant: on it [02:02] wgrant: now, isDriver(IPerson) works for IHasDriver and IHasDrivers ? [02:03] cprov: Right. isOneOfDrivers always did, but there's no reason for isDriver to not also consider .drivers. [02:03] wgrant: looks sensible, I don't get the old isOneOfDrivers() ... [02:05] cprov: Thanks [02:09] cprov, cjwatson: I'm tagging possibly-RTM-relevant bugs "phone-rtm" [02:10] wgrant: okay [02:10] wgrant: will we consider the ppa publisher scale a problem for rtm ? [02:11] cprov: No, but it' [02:11] s mostly fixed now anyway. [02:11] RTM isn't going to use any significantly sized PPAs. [02:12] right, the derived-distros is a way of avoiding huge PPAs [02:13] but it kind of hold us a little more to a-f [02:13] No more than Ubuntu does. [02:17] true [05:19] lifeless: pytz.FixedOffset(minutes), which never seemed to make it into the README [05:19] stub: AHHA [05:19] stub: <3 === jamesh_ is now known as jamesh [08:55] cprov: We identified PPA publishing performance as an issue when figuring out what to do for RTM, but it wasn't the reason we decided to use a derived distribution instead. The reason for that was that PPAs are insufficiently isolated from their parent distribution: they still use the parent in builds, and chroots aren't archive-specific. [08:56] The model works better if we split out an entire new series. [12:53] cjwatson: I see, thanks for contextualising it. === Ursinha is now known as Ursinha-afk === Ursinha-afk is now known as Ursinha === timrc-afk is now known as timrc === timrc is now known as timrc-afk === timrc-afk is now known as timrc === timrc is now known as timrc-afk === timrc-afk is now known as timrc [18:35] cprov: any chance, if I file a bug, that adding specs via the API would be done soon ? [18:35] cprov: it looks like bugs are adding via the /bugs collection, but there is no /specifications top level collection, so the createSpecification method isn't exposed [18:36] wgrant: ^ if you happen to be in an awake state before cprov answers :)) [18:36] lifeless: I can certainly look at it and figure out how much effort would be involved [18:37] cprov: that would rock :) [18:37] lifeless: cool, *bug* me and I will discuss it later with wgrant [18:39] cprov: wicked - thanks! https://bugs.launchpad.net/launchpad/+bug/1329424 [18:39] <_mup_> Bug #1329424: cannot create specification via API === BradCrittenden is now known as bac