[00:03] perrito666: i made some more comments [00:05] how's LP, is it still crapped up ? [00:13] wallyworld: ack, re registering in the shim, I just cargoculted that from other use of that pattern, ill change it [00:16] wallyworld: I can do the per series dissallowance, that is free [00:16] wallyworld: I answered [00:21] perrito666: so we copy whatever init files are relevant to that internal directory? upstart or systemd? [00:23] thumper: when you skip a test you should use t.Skipf [00:26] thumper: https://github.com/juju/juju/pull/5496/files#diff-758d572a629db5b9982ba1841b64dda5R14 [00:31] thumper: but, two thumbs up for not running the state tests on windows [00:32] wallyworld: yup, its quite a clever thing that eric did, we create the init files there and are linked to the right path [00:33] perrito666: ok, so long as it all works :-) [00:55] mmpf, trying to bootstrap trusty with --config enable-os-upgrade=false --config enable-os-refresh-update=false never ends :( [00:56] davecheney: ah, ok [00:57] will remember for next time [01:02] Bug #1587689 opened: Can't upgrade controller with --upload-tools (due to rename) [01:02] thumper: s'ok, I had to rebase my branch anyway, so I fixed it for you [01:03] wallyworld anastasiamac: I'm back [01:03] free to chat now [01:03] axw: wallyworld: tanzanite? [01:04] ok, one sec [01:04] anastasiamac wallyworld: I'm there [01:09] thumper: can we do the same thing we did to the windows state tests timing out to here [01:09] https://bugs.launchpad.net/juju-core/+bug/1581157 [01:09] Bug #1581157: github.com/juju/juju/cmd/jujud test timeout on windows

[01:09] ? [01:09] the underlying cause is the same [01:10] state was different because we don't run apiservers on windows [01:10] however jujud is needed [01:10] I'd be loathed to skip those on windwos [01:10] however [01:10] we could skip it for now until the npipe implementation is fixed [01:11] as long as we tag the skip with a bug number [01:11] and have the bug targetted to 2.0 [01:11] thoughts? [01:13] thumper: it ain't going to get fixed soon [01:13] davecheney: why? can't the windows npipe impl just use select? [01:34] * thumper taps his fingers waiting for tests [01:37] wallyworld: around? [01:37] yes [01:37] wallyworld: answer my priv then :p [01:37] perrito666: was getting coffee [01:37] perrito666: it's already committed [01:37] perrito666: but actually [01:37] it can't be used yet, i'll fix [01:38] after you land [01:38] so ignore me [01:40] lol, ok, because I tried merging master and did not bring the change [01:45] * perrito666 ponders getting coffee then remembers the tim [01:45] time [01:55] menn0: got a second? [01:55] thumper: yep [01:55] menn0: 1:1 hangout [01:58] thumper: 'cos it's windows and select doesn't exist on windows [02:19] wallyworld: are you touching core/description with the Service -> Application rename at all? [02:20] thumper: yeah [02:20] ok [02:20] if you make changes between now and landing, i'll do the merge work [02:20] I'm not [02:20] but I was going to if you weren't [02:21] all good, i'm covering everything [02:22] thumper: the metrics sender uses a global channel to send metrics ... [02:22] one channel [02:22] per process [02:22] no matter how many workers or manifolds are running [02:23] I reader many writers? [02:28] no, it's just another global that has to be patched everywhere [02:31] thumper: oh, it gets worse [02:31] now I've extracted that global [02:31] it is celar nothing is draining that channel [02:32] so shit is sending into it (it's not buffered of course) and nothing is reading from it [02:32] uh... wat? [02:32] it's only used in tests [02:32] Bug #1587701 opened: worker/metricworker: test timeout under race detector [02:36] thumper: up, when not patched by tests, metricsworker.notify is nil [02:38] um... [02:38] isn't that ok though [02:38] ? [02:40] if notify is nil is it still sending? [02:41] its only used as a hook during tests [02:41] a nil channel is never ready to send or receive it is ignored in select [02:43] looking into this a little further, almost everthing in that package is exported because tests [02:43] nothing is called outside juju [02:43] err, calling outside test scope [02:45] * thumper sighs [02:46] * thumper smacks himself in the face again with the lxd tests [02:46] FAIL github.com/juju/juju/worker/terminationworker 1200.065s [02:46] failed to stop [02:46] /home/tim/go/src/github.com/juju/juju/worker/terminationworker/worker_test.go:63 +0x1ee [02:46] bah humbug [02:57] thumper: -n [02:57] do not execute test binaries, compile only [02:58] ^ I call this nope mode [02:58] i just added to to gb to compile, but not run the tests [03:07] axw: wallyworld:thumper: one less invalid login ... https://github.com/juju/juju/pull/5506 [03:08] looking [03:19] anastasiamac: reviewed [03:19] anastasiamac: and I don't know why, but I reviewed on github [03:20] axw: i saw.. thank you :D m moving logic around based on suggestion... [03:20] axw: revew on github is my mistake - i've sent u github link instead of RB [03:20] anastasiamac: that explains it :p [03:35] thumper: menn0 https://github.com/juju/juju/pull/5507 [03:35] davecheney, thumper looking [03:37] wallyworld: in cmd/juju/application/unexpose.go there's still "A application" [03:37] davecheney: how does this fix the timeout? [03:38] axw: damn, i must have had the case checkbox ticked, i'll fix in the next one [03:38] thumper: 1, removes the use of the suite [03:38] 2, adds buffering to the channel [03:39] thumper: look at metricmanager_test, that has a buffer of 2 [03:39] davecheney: good stuff. ship it. [03:39] but when testing the sender and the cleanup, they didn't have a buffer [03:41] davecheney: shipit [03:41] thanks [03:41] i ways also pleased that I could unexport most of the code in that package [03:41] which made passing the notify channel for test less gross [03:48] thumper: simply alias help formatting change: http://reviews.vapour.ws/r/4953/ [03:49] shipit [03:52] thumper: there's just one blocker left on the race build https://bugs.launchpad.net/juju-core/+bug/1587716 [03:52] Bug #1587716: worker: test timeout during race build [03:52] i'll look a that after lunch [03:52] awesome [03:52] thanks davecheney [03:52] no worries [04:00] thumper: password check now passes for controller workers using HTTP endpoints for a hosted model [04:00] hazaah [04:00] thumper: still need to sort out the nonce issue though [04:00] thumper: i'll be pulling this out into a separate PR [04:00] sounds good [04:00] it's reasonably terrible [04:01] and problematic if I screw it up [04:06] axw: cleaned and all tests pass. here is RB link :D http://reviews.vapour.ws/r/4951/ [04:07] axw: which is to say.. thre is probably nothing that really tests this well.. i'll talk to ppl to get better ci coverage i think [04:08] Bug #1587716 opened: worker: test timeout during race build [04:11] * thumper sighs [04:11] where has --upload-tools gone? [04:11] bugger [04:11] pebkac [04:22] anastasiamac: couple more small things === redir is now known as redir_afk [06:09] Bug #1587734 opened: worker/machiner: test failure during race build [06:30] is there anyway to kill a juju run action? [06:30] got some queued that will never run === frankban|afk is now known as frankban [07:31] axw: FFS, why is it when you really need to land something, the bot hates you [07:31] wallyworld: :( [07:31] still trying to land branch #2 [07:32] wallyworld: I've paused my branch, started a new one to restructure instancecfg. I need to pass more bootstrap params through, and the ball of mud in there was driving me crazy [07:32] now finding bugs in providers, woo [07:32] yay [08:16] dooferlad: ping [08:20] dooferlad: didn't we port the fix that puts add-juju-bridge.py in /var/tmp vs /tmp from 1.25 to master ? precise is still broken because of this.. [08:30] axw: yay, branch #2 landed, will propose #3 [08:30] dimitern: yes, we did. Thought I had. I am on holiday until tomorrow so can you do it / file a bug / drop me an email to remind me? [08:31] dooferlad: ah, ok - sorry then, enjoy your holiday :) [08:37] axw: only 235 files this time :-/ https://github.com/juju/juju/pull/5509 [08:48] dimitern: not sure we have [08:48] dimitern: re: /var/tmp [08:50] frobware: we haven't [08:50] frobware: I've verified this earlier, but it only affects precise fortunately [08:51] dimitern: I don't see it in master or 1.25 :( [08:52] frobware: it's in 1.25 [08:53] frobware: https://github.com/juju/juju/blob/1.25/provider/maas/bridgescript.go#L11 [08:53] dimitern: right. sorry - my ref juju clone was out of date [08:54] dimitern: the scripts should be identical in 1.25 and master -- at least that's my expectation [08:56] dimitern: so they /var/tmp is missing. the other change I think dooferlad has in fligh, or in review [08:56] frobware: the script is the same AFAICS, onlt the path is different [08:58] dimitern: I see a difference [08:58] dimitern: let me propose the /var/tmp change [09:03] fwereade, jam: standup? [09:13] wallyworld: looking now [09:14] 10000 lines fewer than last time I think? :) [09:14] babbageclunk, dimitern: http://reviews.vapour.ws/r/4957/ [09:16] axw: yeah, on home run now. still lots of vars etc not done. but those can wait [09:20] frobware: LGTM, testing on precise now [09:21] menn0: I wonder if we should split the help text out into separate files, and compile them in afterwards. might make it easier for, say, the docs team to manage them separately? [09:21] menn0: also would be a sensible direction if we ever want to i18n [09:30] axw, menn0: +1, but, omg, i18n, /shudder [09:33] fwereade: heh, yeah. but won't somebody please think of the enterprises! [09:33] axw: yeah, not a bad idea [09:34] fwereade: I meant to say during the call, that a potential idea for the sprint would be to run a workshop where we run through code with issues and review them as a group [09:35] I wouldn't use specific code from Juju but come up with code that's heavily inspired by existing code, past or current [09:36] it would be helpful to work through a few examples, say a new worker, some apiserver changes, a client API change and a CLI change [09:36] and then try to hit as many of the issues you address in your doc as possible [09:42] menn0, I am writing bits of http://paste.ubuntu.com/16886293/ when I need a break from agent tests -- and it STM that, to be useful, it needs to cover juju-specific scenarios and will thus inevitably end up with content a bit like that [09:48] frobware: your fix works as far as the bridges get created on precise, however it seems lxc is still broken on precise due to: 2016/06/01 09:39:17 http: TLS handshake error from 10.12.19.103:43989: tls: client offered an unsupported, maximum protocol version of 302 [09:48] wallyworld: shipit [09:48] axw: you are awesome ty [09:50] dimitern: I'm ignoring LXC until advised not to [09:51] frobware: yeah, just saying :) [09:54] dimitern: so I do notice on precise that we end up with an additional eth0.cfg (not 50-cloud-init.cfg) :( [09:54] frobware: that's coming from the cloud image [09:55] dimitern: yep, it's just different from trust, xenial === admcleod- is now known as admcleod [12:02] Bug #1587689 changed: Can't upgrade controller with --upload-tools (due to rename) [12:02] Bug #1587739 opened: rackspace,vsphere: firewalling in HA setup is broken [12:02] Bug #1587788 opened: MAAS bridge script needs to reside in /var/tmp on precise

[12:47] jam, incidentally, a picture of what's happening in an agent: http://paste.ubuntu.com/16889596/ [13:16] fwereade: do you have ~15m for a quick chat? [13:18] dimitern, sure [13:19] fwereade: I'm in today's standup HO [13:49] natefinch-afk, please ping when you are in === natefinch-afk is now known as natefinch [14:12] alexisb: ping [14:15] Bug #1587701 changed: worker/metricworker: test timeout under race detector

=== redir_afk is now known as redir === akhavr1 is now known as akhavr [15:59] dimitern: I'm tempted to have a goroutine sleep in a test. Am I a bad person? [16:00] babbageclunk: :) you mean os.Sleep? [16:00] Bug #1553292 changed: TestGoroutineProfile dial unix : no such file or directory

[16:00] dimitern: yup [16:00] babbageclunk: that actually is a known trick to yield back to the scheduler (with os.Sleep(0) for example) [16:00] babbageclunk: but otherwise it's rarely a good idea [16:00] babbageclunk: sleeping is really only bad if you wait until the thing wakes up. [16:01] indeed, natefinch [16:01] dimitern: Well, it's that or the goroutine goes into a tight spin until its OplogTailer gets stopped (which should happen very shortly after). [16:01] dimitern: you shouldn't ever need to yield to the scheduler manually... every function call is a yield point these days. Unless you're doing odd math stuff in a tight loop, you probably never need to worry about yielding [16:02] dimitern: But yeah, it seemed like it was probably a bad idea. [16:02] babbageclunk: well, ideally you'd have a channel that gets closed when the tailer stops [16:02] natefinch: personally, I haven't found a good use for os.Sleep(0) except "let's see what will happen" :) [16:02] natefinch: yeah, I do [16:03] natefinch: but between it getting running out of canned data and getting stopped by the test, it goes into a spin. [16:04] natefinch: When running for real it's querying mongo with a timeout, but that's faked out, essentially. [16:05] natefinch: I think I'm worrying about nothing. Or at least, putting a sleep in to mimic the timeout would be actively worse. [16:07] * babbageclunk deliberately adds imports out of order for the joy of seeing them pop into place. [16:07] lol [16:07] I copy and paste code and intentionally don't even try to make my edited code look nice, because why bother when gofmt will just fix it up? [16:19] Bug #1553292 changed: TestGoroutineProfile dial unix : no such file or directory

[16:19] Bug #1568179 changed: filestorageSuite.TestRelativePath fails because s390x host /tmp is different

[16:28] alexisb: ping? [16:28] babbageclunk, pong [16:29] whats up? [16:29] o you acan merge you bug [16:29] just tag it [16:29] Ok, that was it - thanks! [16:29] bbl lunch [16:37] dimitern: I've tagged my bug with blocker and ci - should it be showing up on juju.fail? [16:38] dimitern: It isn [16:38] gah [16:38] babbageclunk: only if critical [16:38] babbageclunk: you can just run the script locally [16:38] mgz: Ah, that'll be it. [16:39] mgz: Where do I get it? [16:39] lp:juju-ci-tools [16:39] mgz: great, thanks [16:39] mgz: Welcome back by the way! [16:40] ./check_blockers.py check master [16:40] and wow, big list [16:41] want python-launchpadlib - can install via apt or pip [16:41] babbageclunk: thanks :) [16:45] mgz - ok, so as you said, needs to be critical as well. === frankban is now known as frankban|afk [16:51] babbageclunk, you sould be good now [16:51] alexisb: Thanks! [19:12] man, just used the 1.25 local provider.. forgot how fast it bootstraps. [19:32] fast to bootstrap, fast to wipe your env [19:32] and your machine [19:32] and your network [19:38] Bug #1588041 opened: juju bootstrap with vsphere provider hangs with xenial

[19:46] perrito666: I know I know... I just have a need for speed. [20:16] * thumper dashing to Jessie's school to drop off homework [20:32] shame on you, you do your daughter's homework? [20:32] homework delivered [20:35] there was a way to bootstrap (as in bootstrap.Bootstrap) that prevents upgrade after bootstrap, is that still possible? [20:45] perrito666: probably passing config in with bootstrap [21:31] wallyworld, thumper will be a few minutes late [21:31] ok [22:19] anastasiamac: should bug 1514874 be Fix Committed now? (for 1.25 and master) [22:19] Bug #1514874: "Invalid entity name or password" error with valid credentials.

[22:20] menn0: m waiting for stakeholder to verify.. [22:21] menn0: i'll have more info soon... m not too keen to "fix commit" atm.. it's a bit of a handful bug \o/ [22:21] Bug #1559299 opened: cannot obtain provisioning script

[22:21] menn0: but fwiw i *think* it's fixed ;-P [22:30] anastasiamac: ok np... just thought the bug status update had been missed :) [22:32] menn0: no :-D it's intentional. thnx for checking [22:51] Bug #1588084 opened: "upgrade-juju --upload-tools" is hard coded to the model name "admin" [23:03] thumper and anastasiamac : trivial change to change command alias formatting: http://reviews.vapour.ws/r/4961/ [23:04] shipit [23:04] menn0: u have 3 shipits \o/ [23:05] anastasiamac: that should do it :) [23:05] yes \o/ highly desired, i guess :-P [23:09] anastasiamac: do we still need to store image metadata in simplestreams format in the blob store? [23:09] wallyworld: ^^ [23:09] axw: for 2,0? [23:09] anastasiamac: yes [23:10] axw: i don't think so [23:10] axw: i don;t think we need to read from it anymore... [23:10] anastasiamac: I don't see why it should be there either, so I'll take it out and see what breaks [23:10] thanks [23:11] axw: nps :D [23:12] axw: yeah, simplestream is external to juju, we model the image/tools metadata in the blobstore [23:12] it's nice when putting things into a good structure shakes out a lot of old and broken things [23:12] wallyworld: yep, just wasn't sure if anything was still reading it. sounds like not, which is what I expected/hoped [23:13] yup [23:13] axw: r u resolving an issue around it or just a cleanup while u r in the area? [23:14] anastasiamac: I'm changing the way we transfer args to the bootstrap agent, and part of that means changing how we transfer custom image metadata [23:14] axw: awesome \o/ [23:14] anastasiamac: I'm transferring it in the original ImageMetadata struct format, so there's no need to save to disk then read back off [23:14] anastasiamac: and while I was there I saw that we were still writing to blobstorte [23:14] axw: brilliant - thnx :D [23:14] store* [23:14] it'll be much more straight forward to parameterise the bootstrap agent now [23:17] redir: standup? [23:21] Bug #1588084 changed: "upgrade-juju --upload-tools" is hard coded to the model name "admin" [23:22] sorry [23:22] brt [23:25] davecheney: ping [23:26] davecheney: unping [23:30] thumper, davecheney see my comments on LP:#1581157 [23:30] Bug #1581157: github.com/juju/juju/cmd/jujud test timeout on windows

[23:38] davecheney: actually if you are around, I'd like to chat quickly [23:58] Bug #1588092 opened: juju-2.0 has no way to cancel an action

[23:58] Bug #1588095 opened: help for juju run-action refers to commands that don't exist

[23:59] menn0, I will be a few minutes late