[00:16] davecheney: I just answered your question about why I didn't just use a 64-bit int === anthonyf is now known as Guest55250 [00:23] wallyworld: I've just marked https://bugs.launchpad.net/juju-core/+bug/1433566 fixed released [00:23] ci jobs passed [00:23] Bug #1433566: Precise unit test failure discoverySuite.TestDiscoverInitSystemScript

[00:23] thumper: awesome, just need to wait for manual CI job now [00:24] oh, look like tha's gone too [00:24] wallyworld: https://bugs.launchpad.net/juju-core/+bugs?field.tag=ci+regression&field.tags_combinator=ALL is clear of critical [00:24] so unblocked [00:24] \o/ [00:24] gentlemen, start your engines [00:24] * thumper tries to land his branch [00:25] while waiting for the bot, merge master and run all tests... [00:25] * thumper hopes no one added a test that'll break with his changes [00:25] oh FFS [00:26] build fails [00:27] thumper: I find linden tea helps during merges :D [00:27] axw: i'll move the go-amz to propose against v3, not sure if you started looking yet [00:28] does not help the actual merge, but hey...its good tea :P [00:28] menn0: that's an annoying issue with the mgo driver [00:29] davecheney: which annoying issue is that? [00:37] thumper: menn0 's [00:39] ah [00:39] the db size issue [00:43] anyone know how to check the diff between a branch and master for just the changes the branch adds? [00:44] master is 5824 revisions ahead [00:44] and I wanted to see what this old branch of mine did [00:44] effectively what are the changes that would be merged in [00:46] Bug #1433566 changed: Precise unit test failure discoverySuite.TestDiscoverInitSystemScript

[00:47] wallyworld: \o/ [00:47] yeah [00:49] wallyworld: gentlemen?.. :( [00:49] anastasiamac: it's a car racing term [00:49] wallyworld: doesnt fly [00:49] no, cars drive [00:50] wallyworld: m smiling in disappointment [00:50] :-) [00:50] heh [00:53] damnit, had to unblock while I was afk :) [01:05] wallyworld: you said there's a typo, I don't see one [01:05] the "if when" is intentional, if that's what you were thinking of [01:05] axw: from memory "if when" [01:05] oh ok [01:06] maybe not very common grammar, but I don't think it's incorrect [01:12] axw: when you are free, i've proposed against V3 https://github.com/go-amz/amz/pull/41 [01:12] wallyworld: thanks, will look now === kadams54 is now known as kadams54-away [01:35] wallyworld: done [01:37] axw: ty, will look === kadams54-away is now known as kadams54 [01:43] \o/ merge bot doing my branch now [01:44] axw: your one failed BTW [01:46] doh, thanks thumper [02:19] thumper: will be interested to hear how that goes, I've been thinking of trying a standing desk. my knees and back don't like all this sitting [02:19] I wish ubuntu wouldn't reset my laptop's brightness to "staring directly at the sun" every time I reboot. [02:19] axw: I've been waking with a sore back for the last few weeks [02:19] natefinch: heh yeah :( [02:19] axw: this morning, it was all good [02:19] and that was just after about using it four hours yesterday [02:19] ok, sleep, bye [02:19] o. perrito666 [02:19] thumper: nice [02:19] there's like a million studies out there that say that sitting all day is really bad for you, even if you work out regularly and are in shape. [02:20] natefinch: yeah, this is one driving force for me trying it [02:20] and I've been thinking about it for a while [02:20] natefinch: yeah I know, just need to get my arse into gear and do it [02:20] and I decided to hack something up to trial it before I spent any money on a 'custom' solution [02:21] so laptop is on a step-stool thingy [02:21] thumper: good plan, since most of those standing desks are crazy expensive [02:21] and keyboard is balanced on a few draws balanced on six books [02:21] heh [02:21] not the most robust or stable [02:21] but enough to try it out [02:21] * thumper goes to make a coffee [02:25] thumper: I assume you found your answer about your old branch but: 'git checkout ; git log master..' (assuming old branch was off of master) [02:26] or: git cherry -v master === kadams54 is now known as kadams54-away [02:42] anyone online any good with watchers? Wayne and I are trying to get one working for the HA --to feature working, but the state machine keeps giving an error: "unknown watcher id" (adding some logging shows the id is somehow an empty string) [02:44] natefinch, do you have some code to look at? [02:45] natefinch: sounds like the watcher wasn't registered ? [02:45] is this on a private branch? [02:46] dimitern: https://github.com/natefinch/juju/tree/nate-ww3-ha-to specifically https://github.com/natefinch/juju/blob/nate-ww3-ha-to/apiserver/converter/converter.go#L44 [02:47] natefinch, looking [02:47] jw4: quite possibly not registered. Not sure what I need to do to register one [02:49] natefinch, where are you getting the error from? [02:50] dimitern: apiserver/watcher.go newNotifyWatcher [02:51] axw: when did you see Attachment being remove from VolumeParams? we still doing that? [02:51] dimitern: at least, that's where the error is returned by machine-0... it gets logged on machine-1 (the one calling the watcher) [02:52] changes ready. now let's see if thumper's recent merge broke mine. [02:52] wallyworld: not sure yet. why do you ask? [02:52] natefinch, right, so I think I see the issue [02:52] axw: because i need to handle in in CreateVolume() to do the attaching there [02:52] if specified [02:53] dimitern: oh good [02:53] wallyworld: can you just ignore the VolumeParams.Attachment in your branch, and handle it in AttachVolumes please? [02:53] natefinch, hmm.. no sorry - there's nothing apparently wrong with the apiserver code [02:53] axw: sure, i thought that might be the case, hence asking :-) [02:53] wallyworld: nps. we'll need to change the existing providers before it gets removed [02:54] natefinch, except maybe you don't need to call watcher.EnsureErr(watch) on line 66 [02:54] wallyworld: i.e. so loop would create the file in CreateVolumes, and then losetup in AttachVolumes [02:54] yep, sounds good [02:55] natefinch, this is it I think - comparing to other cases (e.g. uniter.WatchConfigSettings) if there's any error, just set the Error part of the result [02:59] natefinch: are you running a specific test? [03:00] go test ./apiserver/converter/... ? [03:00] I guess not [03:01] jw4: sorry, no, the test is "juju bootstrap && juju add-machine" and then watch the log for error messages [03:01] kk [03:03] natefinch: fwiw I see the call to resources .Register(watcher) [03:04] jw4: right right, yeah I saw it too [03:13] axw: thanks for the review [03:14] menn0: nps [03:14] menn0: what're the headers for btw? [03:15] axw: the /log API (used by debuglog) and /logsink (new, used by agents to send logs) use HTTP basic auth [03:16] menn0: ah, right. still using HTTPS, right? [03:17] axw: they use TLS websockets but without the RPC layer that most of our API uses on top [03:17] natefinch, dimitern fwiw I'm pretty sure the NotifyWatcherId on the params.NotifyWatchResult is an empty string somehow [03:17] right, just wanted to know that the basic auth is encrypted [03:17] menn0: thanks [03:18] axw: definitely still encrypted [03:18] jw4, no, I think you're mixing up the Id parameter of the RPC Request [03:18] jw4, watchers are the only entities that actually use the Id [03:18] axw: my recent changes in the api package are all about sharing as much of the connection setup code between normal API connections and lower level websocket only API connections [03:19] axw: and that includes the TLS setup [03:19] sounds good [03:20] dimitern: yeah I think you're right [03:30] exit [03:31] * jw4 in wrong window [03:33] jw4, dimitern: I don't k ow what is going wrong, but this code was supposed to be in over a week ago, and this is one of the main problems that's been tripping us up. [03:34] natefinch: yeah - I'm still betting on either an RPC serialization issue or an unchecked param being sent down [03:35] natefinch, ok, let me try that branch locally, might think of something [03:35] natefinch, got a link? [03:35] dimitern: I appreciate it [03:35] dimitern: I can repro with the branch and steps natefinch gave in the scrollback [03:36] dimitern: https://github.com/natefinch/juju/tree/nate-ww3-ha-to or just git remote add nate https://github.com/natefinch/juju branch nate-ww3-ha-to [03:36] jw4, oh, yeah - I should've looked [03:36] * dimitern thinks it's still only 5:36 am [03:36] :) [03:36] natefinch, thanks, pulling it now [03:38] natefinch: bingo [03:38] api/converter/converter.go [03:38] line 41 [03:39] (actually 38 or so, I added logging) [03:39] result.NotifyWatcherId:"" [03:40] gah [03:40] ignore me [03:40] jw4: heh [03:42] natefinch, dimitern actually... [03:42] jw4, dimitern: maybe resources.Register is somehow returning an empty string [03:42] hmm.. that shouldn't be possible [03:43] natefinch yeah - it must be a serialization issue? [03:45] apiserver WatchForJobsChanges is certainly returning "" for the NotifyWatcherId [03:45] and that's wrong [03:46] whether it's an RPC issue or not we'll know in a minute when my new logging takes effect [03:47] boom [03:47] it's gotta be RPC serialization [03:47] on the apiserver side it's populated [03:48] on the api side its "" [03:48] jw4: interesting [03:48] but that doesn't make sense 'cause it's just a string [03:50] natefinch, dimitern oooh [03:51] api uses params.NotifyWatchResult, apiserver uses params.NotifyWatchResults [03:51] natefinch, I see you're returning params.NotifyWatchResults but in the client-side api you're passing a single NotifyWatchResult [03:51] the slice is getting lost [03:51] jw4, ah :) you're faster [03:51] OH FFS [03:52] dimitern: well I was working on it longer and it's 9pm here not 6 am [03:52] jw4, :) fair point [03:53] thanks a lot, static typing :/ [03:53] :D [03:55] natefinch, well that part of the code is decisively not all static typing [03:55] reflect magic [03:55] dimitern: I know... it's totally all reflection BS [03:56] dimitern: seems like we could be a little smarter about recognizing when we're deserializing into something completely wrong, though. Also, the whole "two types with the same name except for an S on the end" is pretty shit [03:56] that fixed it [03:56] thanks, guys, that's awesome [03:57] natefinch, \o/ [03:57] fun debugging [03:57] wayne and I have been pulling our hair out for a couple days now [03:57] fresh eyes usually help [03:57] natefinch, well, i'll just add it to the growing list of beers for nuremberg *wink* [03:58] dimitern: certainly :) [03:58] yep confirmed it works here too nice [03:58] jw4, indeed [04:02] * jw4 eod [04:02] jw4, have a good one! :) [04:03] whelp bedtime. in 5 hours I can get up and try to get the rest of this working. Thanks guys. [04:03] dimitern: thanks, you too - your day is off to a great start [04:03] jw4, oh I hope so hehe [04:03] lol [04:52] Bug #1433116 changed: 386 compilation error: dblogpruner/worker.go:32: constant 4294967296 overflows int

[05:11] axw: review of a 2-line fix please: http://reviews.vapour.ws/r/1208/ [05:11] looking [05:11] axw: I just noticed that the tests were still failing on i386 [05:12] menn0: was the state package already updated? [05:12] to accept MB [05:12] axw: yes [05:12] axw: that const was there from when it worked in MB [05:12] axw: it just need to be high enough that pruning by size doesn't come in to play for that test [05:13] axw: sorry [05:13] axw: I meant to say "when it worked in bytes" [05:13] that makes more sense :) [05:13] thanks [05:13] axw: sorry. it's EOD and I'm feeling frazzled. [05:14] menn0: nps :) LGTM === urulama__ is now known as urulama [06:15] wallyworld: when you have some time, http://reviews.vapour.ws/r/1209/ [06:15] wallyworld: some bits required for watching filesystem attachments [06:15] axw: sure, will do it after scholl pickup soon [06:15] cheers [06:16] axw: implementing DescribeVolumes() - we pass in provider ids, but there's no easy way to get the juju volume tags to populate the storage.Volume items with [06:17] wallyworld: ah yeah, we should pass both in I guess. [06:17] hmm [06:17] brain fart [06:17] axw: for now, i'm just finishing off tests and will propose the working ebs volume source, i'll add a todo [06:18] * axw looks at interface [06:18] wallyworld: SGTM [06:18] axw: these sorts of things always happen writing infrastructre first :-) [06:19] wallyworld: a couple of options; we could not require the Volume to have a VolumeTag, or we could pass it in with the volume ID [06:19] wallyworld: probably the former -- we should really separate identify from other details [06:20] axw: of hand, i think we should always generate fully populated storage.Volume items in case they are passed around, hence we shuld pass in tags ? agree ? [06:20] I've had in the back of my mind that we should have a storage.VolumeInfo, which describes properties of a volume without the tags and so on [06:20] VolumeInfo works too [06:20] wallyworld: we don't need to implement Describe yet anyway [06:20] we could embed it in Volume [06:20] it's easy to implement, already done, i'll just add a todo [06:21] ok [06:21] right, off to school pickup, bbiab [08:40] morning all [08:42] morning [08:43] morning [08:45] axw: i have to have dinner, i'll review your branch after, i've just finished this one http://reviews.vapour.ws/r/1210/ , i have a small tweak i need to propose for go-amz also, will do that after dinner [08:47] wallyworld: nps. I may have to rethink this stuff again :/ I think either the diskformatter needs to be folded into the storageprovisioner, or otherwise we need to do the mounting in the diskformatter after all [08:47] will take a look at yours [08:59] is there a link to online docs about what you can do with placement? [09:14] hey, anybody else getting bad gateway errors on the juju docs site? [09:14] urulama: ^^^ [09:15] mramm: for /docs yes [09:15] mramm: it's being fixed, deploy on a way [09:15] mramm: but https://jujucharms.com/docs/commands works [09:15] refreshing seems to help [09:15] mramm: getting started affected [09:16] mramm: getting started: aws, azure and testing your setup are broken currently [09:17] (and a bunch others) [09:22] mramm: sorry, didn't find any user oriented placement doc with examples :( [09:24] urulama: no problem thanks for looking [09:33] axw: did you still want that branch reviewed, or do you need to think about it? [09:35] wallyworld: still review please. it needs to be there either way, it's just a question of how filesystems-on-volumes are handled [09:35] ok [09:35] wallyworld: I'm thinking of chucking out the diskformatter, and creating a new filesystem provider. the storage provisioner will still need some special handling for that though, I think [09:36] ok [09:37] wallyworld: I realised I didn't add tests for api/storageprovisioner... doing that now, at the same time as adding the rest of the filesystem methods [09:47] wallyworld: I've updated the branch with the rest of the methods in api/storageprovisioner, and added tests to that package [09:47] ok, ta [09:47] server-side will come in another branch [09:58] axw: +1, was fairly easy to review because there's so much similar boilerplate, not just in this stuff but everywhere [09:59] wallyworld: which missing tests are you referring to in the message? [09:59] apiserver [09:59] unless i missed them [10:00] wallyworld: ah yeah, that code isn't really exercised yet. will do [10:00] thanks [10:00] sure [10:03] dimitern: hangout? [10:03] dimitern: stdup? [10:12] axw: here's that small go-amz fix [10:12] https://github.com/go-amz/amz/pull/42 [10:13] wallyworld: lgtm [10:13] ty === urulama is now known as urulama|lunch [11:12] wallyworld: if you're still around, another fairly straight forward one: http://reviews.vapour.ws/r/1211/ [11:13] sure [11:20] wallyworld: I'm going to now focus on getting tmpfs working, so we can test filesystems. getting filesystems in volumes is probably another couple of days work, due to some required changes I hadn't thought about [11:21] I mean, tmpfs invoked via dynamic storage provisioner [11:21] axw: sgtm [11:21] axw: lgtm on rb [11:22] just an import fix [11:22] wallyworld: thanks [12:16] wallyworld: last one for tonight http://reviews.vapour.ws/r/1212/ [12:16] ok [12:22] axw: done [12:22] wallyworld: tyvm. good point about todo, will fix now [12:23] ta [12:27] wallyworld: if you've got any ideas about how we can unify params and so on for volumes and filesystems, let me know... lots of very similar code in that package :/ [12:28] indeed, we need to step back and re-evalutate the model a bit [12:49] axw: last one if you're still around https://github.com/go-amz/amz/pull/43 [12:56] dimitern: if you had a few minutes, this is a small change to the amz test server ^^^^^ [12:58] wallyworld, sure, will have a look [12:58] ty [13:05] wallyworld, LGTM [13:05] ty :-) [13:24] Bug #1433254 changed: manual provider on trusty/precise syntax error near unexpected token `then'

[13:24] Bug #1434070 was opened: upgrades are broken in master 1.24-alpha1

[13:37] wwitzel3: how goes? === tvansteenburgh1 is now known as tvansteenburgh [13:53] fwereade: is this what you had in mind? http://reviews.vapour.ws/r/1206/ [13:53] fwereade: I waited to land it so that you could look it over first [13:54] Bug #1434092 was opened: updateSeriesVersions() gets called too late in the initialization process === urulama|lunch is now known as urulama_ [14:10] jam, fwereade, I'd appreciate a review on this http://reviews.vapour.ws/r/1118/ when you have time [14:14] * dimitern needs to go out for a while, might not come back until late [14:27] natefinch: ericsnow wwitzel3 that was intended to be unmute but I instead used hang [14:27] because my mouse does those things when running out of mem, sorry [14:28] lol [14:28] we just all hung up anyway [15:06] ericsnow: Could I have a very quick review? https://github.com/juju/testing/pull/56 [15:06] dooferlad: you bet [15:06] dooferlad: wrapping up another one first [15:06] ericsnow: thanks! === urulama_ is now known as urulama [15:50] natefinch: we could go for a more explicit OneNotifyWatcherResult, ManyNotifyWatcherResults :p [15:51] perrito666: I actually don't think that's a horrible idea. I was going to give suggestions in the email, but didn't really want to bike shed it [15:52] perrito666: it doesn't help that they have to match up across files [15:52] we could make a test that checks that two variables on a same namespace have a certain levenstein distance :p [15:53] perrito666: that's a very interesting linter rule... I kinda like it, actually. [15:53] uff, wfh is terrible at lunch time when you live in a neigborhood with too many old ladies [15:53] ......explain? [15:54] natefinch: young people eat at work [15:54] old ladies cook from 11AM and a really tempting smell comes trough the window [15:55] natefinch: lunch here is quite a heavy meal since our breakfast is mostly tea/coffee and crackers [15:55] I dont even have breakfast,I only eat a couple of fruits [15:56] so I have to work for two hs thinking on my relatively poor lunch while smelling stew, roasted meat, pizza, etc [15:57] dooferlad: FYI, while we don't have hooks set up (yet) for the testing repo, you can still use rbt to post a review request [15:59] ericsnow: ack [15:59] omg, someone is doing a wine reduction... how cruel [16:07] perrito666: lol [16:16] wwitzel3: you around? [16:26] anyone know if we already have something in machine agent that watches for changes to the machine? Seems like we would, but it's a little hard to figure out through all the levels of abstraction in this code === kadams54 is now known as kadams54-away === natefinch is now known as natefinch-afk [17:54] Bug #1434092 changed: updateSeriesVersions() gets called too late in the initialization process [18:00] Bug #1434092 was opened: updateSeriesVersions() gets called too late in the initialization process [18:06] Bug #1434092 changed: updateSeriesVersions() gets called too late in the initialization process === natefinch-afk is now known as natefinch [18:29] oh my god, the watcher code... jesus [18:29] there's like four levels of indirection that could get reduced to one or two lines of code in most cases :/ [18:33] natefinch, I've just had my fingers in the watcher code - if there's anything I've done you don't like I love to learn :) [18:33] natefinch, also, I'm pushing more changes to the meter status branch - I've marked it WIP in rb because there are still things to be cleaned up before its reviewed again... [18:34] natefinch, but how does this test grab you? https://github.com/mattyw/juju/blob/metricsmanager-meterstatus/state/meterstatus_test.go#L266 [18:40] mattyw: I don't actually know what the combinations mean, but it is nice and readable [18:41] mattyw: you don't really need the var block, though, they can each just be := instead [18:41] natefinch, if I did that gofmt made sure they were all nicely lines up for me :) [18:42] mattyw: ahh, interesting point [18:42] mattyw: that never would have occurred to me, but yes, I think that helps readability, so, cool, nice work. [18:43] natefinch, I'm going to call it a day [18:43] natefinch, but if you're around tomorrow I will pester you if you don't mind :) [18:43] mattyw: have a good evening [18:43] mattyw: I'll be around :) === kadams54 is now known as kadams54-away [19:25] Bug #1434246 was opened: Destroying environment takes down others with similar name === tvansteenburgh is now known as tvan-afk [19:40] hatch: on that bug about destroy-environment, how exactly did you invoke destroy environment? i.e. did you use -e or rely on the "current" environment? [19:41] natefinch: umm it was scripted, lemme take a peek [19:41] natefinch: I reproduced it once but I will do it again so I can capture the exact steps for the bug [19:45] hatch: also please note if you used --force [19:45] natefinch: juju_env = lambda: os.getenv('JUJU_ENV') is the line it uses to get the environment [19:45] but I'll try manually as soon as this test run is done [19:45] probably 10m now or so [19:45] hatch: I knew that if I asked if you did it way X or Y that you'd come back and say Z ;) [19:46] Bug #1331505 changed: destroy-environment shuts down joyent machines

[19:46] natefinch: isn't that the way it goes? :D [19:46] hatch: but, still... how does it use juju_env... to do -e? === kadams54-away is now known as kadams54 [19:50] natefinch: the command executor adds -e to the commands using the value returned from the above line [19:52] hatch: *shrug* works correctly for me [19:53] I'm on master, of course, but I doubt we've changed that any time recently [19:53] yeah ok I'll try again one this run is done - these tests take a while :) [19:54] hatch: no problem :) [19:56] sinzui: ping [20:02] hi thumper [20:02] o/ [20:03] just talking with abentley about ci tests [20:03] in standup now [20:03] ping after === kadams54 is now known as kadams54-away === kadams54-away is now known as kadams54 [20:16] sinzui: do you have a few minutes to chat? [20:18] I do [20:18] awesome [20:18] * thumper makes a hangout [20:19] sinzui, waigani: https://plus.google.com/hangouts/_/canonical.com/core-ci [20:29] thumper: you seen blocker bug 1434070 which is suggested in a comment may be jes related? [20:29] Bug #1434070: upgrades are broken in master 1.24-alpha1