[01:11] mgz: can I get on the ppc64el CI machine to dig into this bug? [01:13] mgz: never mind. fails on amd64 using gccgo. [01:43] waigani: looks like you're OCR today. can you please review https://github.com/juju/juju/pull/2892? fixes the critical blocker on master [02:37] axw: sorry, just saw this. Shipit [02:37] thanks [04:00] Bug #1479194 opened: Juju agent hangs if debug-hooks session goes away [04:10] Bug #1479194 changed: Juju agent hangs if debug-hooks session goes away [04:22] Bug #1479194 opened: Juju agent hangs if debug-hooks session goes away [06:26] morning [07:13] Bug #1479024 changed: TestRootDiskTags fails on ppc64el

[07:22] dooferlad, hey [07:56] dimitern: hi. Sorry, that was a very long reboot. [07:57] wasn't the entire 40 minutes! [07:57] more like 10 [07:57] dooferlad, :) wow [07:57] well, I think something hung. Gave up and hit reset. [07:57] dooferlad, just a heads up about tomorrow and friday - approved [07:57] dooferlad, please update the calendar [07:57] I had 8 KVMs running, so I figured those may be an issue... [07:58] :D [08:32] thumper, final-leadership-switch-1.24 [08:50] eugh. bug #919437 sucks [08:50] Bug #919437: Lost window, cannot get it back.

[09:40] Bug #1479278 opened: service status derivation happens in the wrong place [09:48] axw, ping? [09:57] mattyw: hey, I'm here for just a moment. what's up? [10:01] mattyw: sorry gotta go again, leave a message I'll check back later [10:03] axw, sorry, missed you. no problem, I've added another question about that branch, it's basically lgtm to me though, I just need enlightening :) [10:22] Bug #1479289 opened: statushistory uses sequence, fails in multi-env state servers [11:18] mattyw: thanks for the review [11:20] axw, welcome, sorry I missed it [11:20] axw, reviewboard colours invert if I have my monitor at the wrong angle [11:20] mattyw: oh :/ [11:21] axw, one of my faviourite features [11:36] dooferlad, hey [11:36] dooferlad, are available now for about 1/2h ? [11:36] perhaps less [11:57] dimitern: just grabbing coffee and then will be back from lunch. [12:02] dimitern: back [12:05] dooferlad, ok [12:05] dooferlad, let's use our 1:1 g+ and the c9 workspace? [12:05] dimitern: sounds good [13:06] dimitern: ping [13:06] TheMue, pong [13:07] dimitern: wanted to have a quick chat about the final failing test [13:07] dimitern: it is at https://github.com/TheMue/juju/blob/addresser-worker-using-api/worker/addresser/worker_test.go#L237 [13:07] dimitern: here I've got a timeout when waiting for a release op by the dummy provider [13:07] TheMue, sure, looking [13:07] dimitern: the rest works fine [13:09] dimitern: I'll paste the debug log, mom [13:09] dimitern: so, http://paste.ubuntu.com/11959517/, could help [13:10] TheMue, could it be due to this: https://github.com/TheMue/juju/blob/addresser-worker-using-api/worker/addresser/worker_test.go#L90 [13:11] more than 10 ops queued? [13:12] dimitern: it's worth a try, due to the now happening bulk requests. [13:13] TheMue, try logging each ReleaseAddress dummy method call [13:14] dimitern: nope, no change. and the number of dummy ops should be the same, even with bulk request. they happen per address on the server [13:16] TheMue, is the apiserver facade actually calling the environ ReleaseAddress ? [13:16] TheMue, I see no logs there [13:17] s/there/about it/ [13:17] dimitern: in principle it does, all other tests waiting for releases work, and the op is sent by the dummy provider [13:18] dimitern: but SetUp adds four addresses, later a fifths is added in this test, and I only see four tries to release [13:19] dimitern: this 0.1.2.9 seems not to be released [13:20] dimitern: when the machine is removed the address is dead, but it looks like the addresser doesn't recognises it [13:22] dimitern: TestWorkerRemovesDeadAddress() in line 179 has no problem *sigh* [13:22] TheMue, yeah, but there the addresses start as dead [13:22] TheMue, so it feels like a watcher/sync issue [13:23] TheMue, why call do you StartSync again in TestMachineRemovalTriggersWorker (twice even?) [13:23] dimitern: ok, will add some debug statements on this path [13:23] TheMue, IIRC StartSync spawns a goroutine [13:24] dimitern: the tests are take as they have been before, I only had to change one combined error message so far [13:25] dimitern: my hope have been good blackbox tests don't caring if the addresses uses and api or state [13:25] ;) [13:26] dimitern: but ok, have now a direction where to investigate, thanks [13:26] TheMue, well, since the logic has changed, the tests might break - e.g. 2 ops per list of ips (release, remove) rather than 2xN [13:27] dimitern: aren't the ops by the dummy, so that only the calls to env are counted? those are only the release calls, remove then works on state again [13:28] dimitern: and the number of release calls depends on the number of ips in the test case [13:28] TheMue, true I think - env.ReleaseAddress is called once per IP [13:28] dimitern: it's now only that it isn't called by the worker but by the server api [13:29] dimitern: yep [13:29] dimitern: the server api gets a bulk to release and then calls the according env menthod [13:29] ... n times [13:30] TheMue, I'd suggest more logging [13:30] TheMue, I need to be in a call now, but let's revisit it tomorrow if it's still an issue? [13:32] dimitern: yes, will do so, thanks for support [13:35] TheMue, np [13:40] * TheMue rants about the weather. we've got July 29th and I sit here with thick socks, blanket, and cardigan. :( cold.dislike == true [14:09] dimitern: https://github.com/juju/juju/compare/net-cli...dooferlad:net-cli-spaces-state?expand=1 should be more like what you were after. I hope. [14:09] dimitern: I am going to grab some tea, but am happy to talk about the branch. I know it needs more tests! [14:16] dooferlad, ack, will look shortly [14:33] dooferlad, looks mostly OK, just a few passing notes: use string and []string in params structs, not names.*Tag - the params struct describes the wire format (and for some other good reason that currently escapes me) [14:35] dooferlad, also - doc types in state are not exported for a reason; no need to declare SetUp*/TearDown* for SpacesSuite - this is needed only if you embed multiple "suite" types etc. [14:38] dimitern: I thought we were trying to always call the SetUp* etc calls, even if they weren't needed, just because of good habits. [14:38] dimitern: I don't care either way [14:38] dooferlad, if we do now, I'm not aware of it :) [15:13] dooferlad: you don't need to redirene SetUp* if you're embedding exactly one suite [15:13] *redefine [15:13] because gocheck will just call that setup [15:14] what you have to be careful about is if you embed multiple suites, or write your own SetUp* - then you need to upcall [15:15] mgz: apparently I imagined a conversation about it. Will get rid. [15:15] dooferlad: the thread happened, the circumstances are just a little narrower than you were thinking [15:15] mgz: good to know :-) [22:33] anyone able to review these? [22:33] http://reviews.vapour.ws/r/2276/ [22:33] http://reviews.vapour.ws/r/2275/ [22:34] (easy ones) [22:34] menn0: ill give it a shot [22:35] perrito666: thank you [22:38] menn0: any idea why the tags where there for bson? [22:39] perrito666: I suspect it was started by someone who didn't understand what they were doing and then other people copied [22:39] cargo cult programming [22:41] menn0: indeed, I especially like the seccond one [22:41] I am a bit amazed that nothing broke there [22:42] almost too good to be true [22:42] :) [22:42] perrito666: i'm about to make some more signficant changes that will almost certainly break things :) [22:42] menn0: yeah, its the interface->string change that has me baffled [22:44] * perrito666 also sees a lots of panics and supposes this code was just ugly from scratch [22:44] perrito666: that's just a simple tidy up... all the ids were strings anyway, and localID() was being called on all of them, so I just changed where that was being done so it's just in one place [22:44] menn0: shipthem [22:44] perrito666: thanks [22:44] perrito666: yeah, this area is a bit special [22:45] menn0: I was just suspicious of exactly that, as if the code was writen for public facing stuff but all methods where private [22:45] perrito666: yep. the stuff I changed is all internal only. [22:46] leadership is currently broken in master? [22:47] perrito666: pretty much [22:48] perrito666: will is working on it [22:48] menn0: I was not sure if it was broken or just extremely unstable [22:48] * perrito666 cannot ftloc make status tests pass [22:49] perrito666: extremely unstable ~= broken [23:18] Bug #1479546 opened: Storage provisioner timeouts spawning extra volumes on AWS [23:55] waigani: review pls? http://reviews.vapour.ws/r/2277/ [23:57] menn0: how expensive is that call to st.EnvUUID() ? [23:58] davecheney: not very. it just returns a field of the struct [23:58] davecheney: and ... you're back! [23:59] i am [23:59] hello [23:59] in taht case [23:59] LGTM [23:59] (as I'm also OCR today)