[07:54] mornin' all [07:54] urulama: hiya === rogpeppe2 is now known as rogpeppe [08:17] rogpeppe1: morning [08:17] urulama: how's it going? [08:17] urulama: i just responded to your comments on the bundles spec, BTW [08:18] great, thanks, had some time this morning to read the docs [08:19] agree to keep things simple === rogpeppe1 is now known as rogpeppe [11:09] morning [11:10] morning [12:10] hey rick_h__ (i got disconnect after my last ping) [12:14] bac: morning [12:55] * rogpeppe1 now has a keyboard with all keys in place [12:55] yay? [12:55] darn fiddly, replacing those hinges [12:55] yay! [12:55] hah, laptop keyboard? [12:56] only one hinge slightly broken on installation, seems ok so far [12:56] rick_h__: yeah [12:56] rick_h__: broken after i spilled cream on it the other day at the end of the hangout [12:56] oh :( [12:56] rick_h__: not broken by the cream, but broken when i took the keys off to clean off the cream... [12:57] I did some tea with honey once. Yea, those scissor switches are a pain to get off and back in place [12:57] it's the tiny little plastic protrusions which are so so so easy to break [12:58] and you actually have to exert a reasonable amount of force in *just the right* way to get it in [12:58] anyway, lunch! [13:11] jrwren: morning [13:11] jcsackett: can you and jrwren get together and go through the issues jrwren hit yesterday on getting cloud providers going and help find if there are existing bugs around the issues? [13:12] oh boy. [13:12] jrwren: well if you're going to find broken stuff we need to make sure they're going to be addressed :) [13:12] agreed. [13:12] this shall be a fun dive. [13:12] I think the azure one is likely to be the timeout issue that bac hit before [13:12] yup, that is what made me think its a bug. [13:13] and there's an email thread around apt-get upgrade timing out juju in azure [13:13] the joyent one might be the same or similar bug. [13:13] but bac said it worked the next time. that was not the case for me. [13:13] I think bac was using 1.18, I was using 1.19.[34] [13:13] jrwren: ok, haven't seen anything on joyent, but we don't use it actively [13:14] yes, because environment upgrades are only supported from stabel to stable releases [13:14] understood. [13:14] he moved back to 1.18 (stable is even) while you were on 1.19 (odd is dev) [13:14] so it's more important to figure out the bugs/issues in 1.19 to make sure they're addressed before 1.20 is released (1-2wks) [13:16] jrwren: but get wit jcsackett when he's around and see if he can help find bugs, intro you in #juju if there's questions, etc. [13:16] sounds great. [13:16] jrwren: and in between time move forward with the tasks using other cloud providers so that you can move forward [13:16] i'll jump in on my own too, so i'm not entirely blind. [13:17] ok [13:33] jrwren, welcome to Canonical! [13:34] jcastro_: thanks! [13:34] consider me your friendly chaos monkey. [13:35] I heard [13:35] <3 [13:37] jujugui: i have crap all for internet at home. TWC claims it will be fixed this afternoon. running to the coffee shop. [13:46] jujugui: http://ci.jujugui.org:8080/ may be happy again but i guess we won't know for sure until a controlled branch is proposed/approved. please let me know when you do so. [13:52] jcsackett: ping [13:59] marcoceppi: where do i file a bug against a charm? precise/jenkins specifically? [14:00] bac: https://bugs.launchpad.net/charms/+source/jenkins [14:00] marcoceppi: thanks. i poked around but couldn't find it. [14:01] bac: there's a branch from Makyo to alnd [14:02] bac: we can use to test the lander part at least [14:03] rick_h__: ok [14:04] rick_h__: this install may be suspect since i had to do so much stuff behind the scenes. juju may not be happy with me. [14:04] bac: ok, well let me know if we need to retry to restore? [14:05] on the one hand we need to get this up, but the goal is to get something up we can use going forward and not have to repeat this [14:08] rick_h__: agreed. i guess i'd say let's just keep an eye on things. it is up now so until we see something funky i'd leave it be. [14:08] bac: ok, trying to shipit the branch from Makyo [14:12] Makyo: around? my next meeting will overlap the standup and need you to run it if possible. [14:19] bac, rick_h__: my PR is also ready to land. Waiting to see how things go with Makyo's… [14:20] kadams54: ok, it's running keep an eye on it [14:20] I've got a call in a few so will not be able to track it down [14:20] what is the juju way of not repeating this? custom charms? [14:20] jrwren: fork a charm, push it up under your namespace [14:21] cool. got it. [14:21] jrwren: so bzr branch the original, and bzr push to ~jrwren/charms/jenkins [14:21] or hte like [14:21] need to match the exacty syntax, should be in the docs [14:22] as for my "nothing working", i've figured it out. I guess I'll file a bug, although its a pretty niche issue. [14:22] jrwren: ok, well still worth documenting for other's to search/find [14:22] jrwren: but good to hear it's something that sounds like you can move forward around [14:23] yes. I should have thought of it sooner. I was frazzled from breaking your CI yesterday. [14:23] jrwren: all good :) [14:28] jrwren: how are you going to fork the jenkins charm? [14:28] jrwren: if you're looking to make a change i have a suggestion! [14:28] bac: sorry. that was just my example [14:28] bac: current charm in my brain atm :) [14:28] oh, i'm not. I am more trying to understand some juju philosophies. [14:29] rick_h__: bug 1336273 [14:29] <_mup_> Bug #1336273: Add ability to specify PPAs [14:29] bac: oh cool [14:29] rick_h__: if we had that, we could go a long way to having a bundle that did most of our setup [14:29] jrwren: the branch philosophy is on the chopping block though. There's a spec doc I can share if you're interested [14:29] rick_h__: so if someone want experience in charm authoring... [14:30] s/want/wanted/ [14:33] rogpeppe1: for the standup, can we get a doc shared up for that work so we can start to look at things? [14:34] rick_h__: ok, will do [14:34] Looks like the merge build failed, but it seems like one of the random selenium failures [14:43] it was: https://bugs.launchpad.net/juju-core/+bug/1336313 [14:43] <_mup_> Bug #1336313: juju fails silently when .ssh/config has a ControlMaster set [14:49] jrwren: I've not hit that bug specifically, but I did appreciate that "juju ssh" seemed to use my own ControlMaster correctly. [14:50] jrwren: are you sure that you didn't have a cached connection to an existing machine with a matching %r/h/p that was hung or something?# [14:50] jujugui call in 10, kanban now [14:51] rbasak: i don't see how, since those hostnames are basically random from teh cloud provider each time. [14:52] rbasak: do you use control persist yes? or a timeout? maybe you have a small enough timeout? [14:52] jrwren: I'd be interested to know what the real problem is rather than disable control master use completely. I find it really useful - it speeds things up massively. [14:52] rbasak: ok, I'll experiment more. [14:52] I use ControlPresist 3600 [14:52] i'll be that is the difference. [14:53] hazmat: you approved my tls branch for juju-client. will you be merging it soon? i've got a card in kanban that i'd like to kill. [14:53] That's an hour, though. Surely the juju timeout isn't that long? [14:53] *i'll bet* [14:56] jrwren: I wonder if this is Azure-specific? I wasn't aware of any need to be able to ssh the bootstrap host during bootstrap. I thought Juju entirely used cloud-init? [14:57] * rbasak has commented on the bug [14:57] I hope that's useful and constructive. [14:58] already useful to me. I need to try with timed ControlPersist instead of just "yes" [15:00] jujugui call now [15:02] antdillon, urulama call now [15:03] urulama: is on call with me [15:03] ant as well [15:09] nothing quite like being hangouts with almost no bandwidth. :p [15:12] heh that's why I wasn't in today :) [15:23] bac, yeah.. sorry catching up on backlog.. i'll merge shortly [15:25] thanks hazmat [15:28] jcsackett: when you 'juju ssh' to azure instance do you find the ssh connection goes away frequently? [15:28] bac: i haven't done any juju stuff on azure; i only ssh directly into the jenkins units. [15:28] jcsackett: yeah, should be the same, though. they don't drop? [15:29] bac: i haven't noticed it, no. [15:32] jujugui: can someone confirm a bug for me? [15:32] sure [15:32] so... when I broke CI yesterday, why didn't I get a "environment is already bootstrapped" message? were there state storage changes? [15:32] maybe [15:32] jcsackett: yup, there's a bug [15:32] jcsackett: oh, you mean a *particular* bug? :-) [15:33] * jcsackett laughs [15:34] if someone can open comingsoon.jujucharms.com (or any other develop branch instance they have) and deploy a service, i'm seeing a gray bar instead of the charmbrowser or the service inspector, but i'm having lots of oddities today. [15:34] looking [15:34] this is w/o flags, so i believe the service inspector should open. [15:34] jrwren: i think you should have. try to bootstrap the same env twice on ec2 and see what happens [15:35] jcsackett correct, something is broken [15:35] hatch: cool, i'll file and set up a card. [15:35] it doesn't changeState to the new url any longer [15:35] jcsackett urgent imho [15:35] hatch: quick, steal that for your QA day bug! [15:35] agreed. [15:35] nope! [15:35] bac: exactly what I was thinking, and exactly what I just got on azure, but I didn't yesterday with the CI environment [15:35] it's mine, b/c i forgot to do QA the other day. :p [15:36] bac lol I think i've filed enough bugs this past week to get some wiggle room on qa day :) [15:36] * bac admits he sat on a bug for 12 hours last week [15:38] jcsackett I think this was because one of the branches that recently landed didn't get qa'd without the mv flag [15:38] * jcsackett nods [15:38] hatch: looks like. [15:39] card made in urgent lane, for whoever is looking for work. [15:40] hatch: is there any reason we can't rename the "UIState" thing back to State? [15:41] jcsackett none I can think of [15:42] jcsackett oh I left it like that because we have a 'state' attribute in app.js [15:42] which has NOTHING to do with state [15:42] lol [15:42] it's the environment or somethin [15:42] g [15:43] ok then. [15:43] we'll leave it be, but i'll remove the XXX saying we can rename it. [15:43] jcsackett well we CAN rename it, but we should probably also rename the other state thing in app to something that makes sense [15:44] else it'll just be even more confusing heh [15:44] there will likely be a ton of renaming required though [16:21] hatch: i have a growing feeling that to dispatch charm details and unit details on the inspector we need to get rid of slots. [16:22] why's that? [16:25] hatch: well, details are all rendered via the "showViewlet" stuff, which is slots based, isn't it? [16:25] hatch: i guess we don't have to get rid of it. [16:25] we're just dealing with more stuff to cleanup. [16:25] and inspectors now need to conditionally render viewlets in slots both on event and on init. [16:26] hmm lemme take a look [16:26] hatch: ok. [16:27] jcsackett https://github.com/juju/juju-gui/blob/develop/app%2Fviews%2Finspectors%2Fservice-inspector.js#L89 [16:27] here is where you would conditionally show the unit details [16:30] jcsackett here is where it shows the unit details https://github.com/juju/juju-gui/blob/develop/app%2Fviews%2Fviewlets%2Fservice-overview.js#L724 [16:30] currently [16:31] Don't forget charm details. That's also needing update. [16:31] So then the click fires change state, and I guess renderui gets called again... [16:32] Yeah, ok. I hadn't gotten back into state land; this isn't the problem I thought it was. [16:33] hmm [16:33] this might actually be a little more work [16:33] we can't re-render the inspector [16:33] so it has to know it's already rendered [16:33] maybe it already does that [16:33] heh [16:33] * hatch looks [16:34] jcsackett yeah you'll have to implement something for the inspector like there is for the charmbrowser [16:34] so that it doesn't 'aways' re-render the inspector, but insteads just sets new data in [16:35] instead [16:35] hatch: dig. [16:35] I *thought* that might be a wrinkle. [16:36] the proper inspector method is _inspector [16:36] I'm not sure where `inspector` method is used heh [16:37] it might need to be removed....the `inspector` method [16:51] I think it's leftover from the old system. [16:54] yeah === rogpeppe1 is now known as rogpeppe [17:20] hey guys, how goes? [17:20] Makyo: working on landing your branch, looks to be running tests now at least [17:20] kadams54: yay for yours landing [17:21] Yup [17:21] jcsackett: you're set then? [17:22] rick_h__, ah, thanks, was trying to figure that out [17:22] Makyo: yea, no idea why it's doing that and didn't for kadams but oh well. [17:27] rick_h__: i'm done for the day now. here's a link to the on-going API proposal doc: https://docs.google.com/a/canonical.com/document/d/1ILHRpOe-qDlmjxHBbLUea7InDpehx5_roJ1ynZmcZDc/edit [17:27] rick_h__: formatting is a bit scrappy i'm afraid :-) [17:28] rogpeppe: awesome thanks. Will take a look. Can you move ot to the 14.04 specs folder if it's not in there currently? [17:28] and make sure the team's got edit [17:28] rick_h__: everyone's got edit, i think [17:28] rick_h__: not quite sure how to do the folder thing in google docs [17:29] rick_h__: i didn't even know there *were* folders for google docs :-) [17:29] https://drive.google.com/a/canonical.com/?usp=docs_web#folders/0B95cjYKYHu9oMm50aVlLS0pYZWc [17:29] is the specs folder [17:29] somehow want to get it in there but yea not sure about it myself [17:30] jujugui heads up the two calls with Mark S today went well. He's excited we've got urulama and jrwren starting and loved hearing the osx stuff got out and is interesting in checkout out the usage of that [17:30] cool [17:30] rick_h__: i'll work out how to put something in a folder tomorrow... [17:30] rogpeppe: ok, will try to figure it out. Even if we have to create a new doc in that folder and copy/paste [17:30] \o/ [17:31] rogpeppe: have a good evening [17:31] * rick_h__ goes to find some food now that calls are done for now [17:31] he's usually pretty good about seeing stuff go by on G+ so i was surprised that went unnoticed [17:31] urulama: see ya tomorrow morning [17:31] rick_h__: have a good rest-of-day [17:31] g'night all [17:31] rogpeppe: bye [17:38] rick_h__: after looking at the other jenkins installation i think we may be better to manage charmstore on our own. their jenkins is 1.4, the plugins we need aren't installed properly, etc. [17:39] bac: ok, then let's do it [17:39] rick_h__: i'm trying to install a slave now. if it works i'll run with it. otherwise i'll talk to sinzui about their upgrade plans [17:39] bac: feel free to create cards and let me know if you need help [17:39] bac: rgr [17:41] ok, now really going to head out for lunch and move to the coffee shop for the afternoon. === urulama is now known as uru-away [17:47] rick_h__: i am set. [17:58] http://askubuntu.com/questions/457635/what-is-a-configuration-to-allow-access-to-juju-gui-in-local-provider [17:59] do we support the gui like this? [17:59] and if so, how? [18:00] jcastro: looking [18:01] jcastro: not yet, juju has to support the networks in containers and be able to expose the internal port80 stuff out. [18:01] jcastro: I'm not 100% sure on the status of that in juju-land. [18:01] ok want me to answer it or you got it? [18:01] jcastro: but the gui is hanging because it probably can't talk to juju from that container [18:01] I am going through our question queue [18:01] jcastro: I'll put something together. [18:02] jcastro: side note, do you know if brew has download/install stats? [18:02] jcastro: we're curious on measing quickstart install on osx [18:02] no clue, will investigate [18:02] jcastro: ok, wasn't sure since you guys have some brew stuff/juju itself. We can look into it [18:03] Alright, got a legit failure. At least it's something to go by. [18:04] rick_h__, doesn't appear that way [18:04] rick_h__, from looking at the formula [18:04] Makyo: :/ ok [18:05] jcastro: ok thanks [18:05] https://pypi.python.org/packages/source/j/juju-quickstart/juju-quickstart-1.4.0.tar.gz [18:05] it snags from there [18:05] since no one links to that except for brew, I would think that's where the stats go [18:05] jcastro: right, but wondered if brew had a popcorn or the like to track installs from itself [18:05] oh, does brew pull that down every install? [18:06] yeah [18:06] it's a homebrew, that's the point [18:07] heh, ok cool [18:07] I would do your next release at a URL you have GA on, and then update the formula to get the tarball from there [18:08] Makyo: hmm, that's one I've not seen much. Is that legit then? [18:08] Bah, can't reproduce locally. Not sure it is, rick_h__ [18:08] Makyo: ok, feel free to retry [18:08] Makyo: and let's keep an eye. Now that we know the CI stuff is running again. Sorry for the pain in landing [18:08] No worries! I'm watching. [18:08] lol [18:11] wow 534 downloads in the last week [18:11] jcastro: ^ [18:32] nice [18:34] jcsackett so I'm going to take off for the day, any q's about that inspector stuff before I go? [18:36] hatch, http://www.buzzfeed.com/tanyachen/americans-fail-canada-again [18:37] Makyo haha I was going to paste that tomorrow :P [18:38] I accept not knowing all of the provinces but damn some were just horrible! [18:38] lol [18:39] that is awesome Makyo [18:46] Makyo: what are you up to? Can you look at the card in urgent nexte? [18:46] next that is [18:47] rick_h__, sure thing [18:47] Makyo: ty, I'm hoping to stick hatch with a release this week and that one seems a blocker [18:47] Alright, sounds good. [18:47] jrwren: can you make sure to file holiday time for the 4th please? [18:47] jrwren: and just how goes overall? [18:51] i'll figure out how to file it. [18:51] jrwren: sure thing, it's in the hr.canonical.com system [18:51] it goes... i have so many questions at this point, and mostly they are somehow influenced by yesterdays disruption. [18:52] jrwren: once you login there's a list of your "all other absenses" and you can create a "New" [18:52] jrwren: cool, want to chat? [18:52] sure. [18:53] jrwren: k, let's meet in the standup hangout from today [19:00] marcoceppi: you have any experience using the jenkins-slave charm? [19:00] bac: not for a long time, no [19:01] marcoceppi: do you know if after adding the relation the slave should be launched and marked as online? i've formed a relation but the master sees it as offline and i can't get it online. [19:02] bac: are you using the charm store charm? [19:04] Well, that was easy. [19:04] Lemme test it. [19:06] marcoceppi: i took the cs charm and pushed it to LP for trusty [19:07] bac: I used more updated branches that were in the canonical-qa or some other users branches [19:09] marcoceppi: ok, cool [19:09] bac: these charms are on our list for the audit [19:09] so I'll be pushing to update them [19:14] jujugui quick PR for ghost inspector https://github.com/juju/juju-gui/pull/414 [19:15] Makyo: looking, can you rekick your other pr for land please while I do? [19:15] Makyo: looks like it got hung up in sauce land this time :( [19:15] Yeah, boo. Will kick. [19:19] Makyo: sorry, can't QA. I need to setup an env on this new laptop and am heading out shortly [19:20] jujugui going afk for a bit and will be back later. Have some later calls tonight. [19:44] jujugui: what am i doing wrong here when trying to checkout a specific revision: http://paste.ubuntu.com/7732968/ [19:47] Fiiinally landed. Yeesh. [19:52] bac: git fetch first? [19:52] bac: where is that revision located? [19:53] bac: if it's in a pull request, you need to be watching the pr branches in the branch spec [20:00] rick_h__: this is in the jenkins slave [20:00] rick_h__: so do i need to do some manual config before jenkins can checkout the specific branch for the PR? [20:01] bac: ok, so if you look at the .gitconfig on the gui CI server there's some stuff to make git look at the special places github puts pr branches [20:01] bac: once that is in place, a git fetch should update and load the pr locations and then you can git checkout the hash [20:03] rick_h__: ok, i just copied the .gitconfig to the slave [20:04] and the checkout worked [20:04] hurrah [20:04] bac: cool [20:05] rick_h__: can you look at http://ci.jujugui.org:8080/job/charmstore/9/console [20:05] bac: looking [20:06] heh, java and git don't mix [20:09] bac: looking into it, can you give me access to the slave machine as well please? [20:11] rick_h__: alreadydid [20:12] bac: ok, everything looked ok, but with the gitconfig change I wiped the workspace and restarted a build using current master rev [20:12] bac: and it looks like it's running [20:13] i'm beginning to think the jenkins charms are dangerous. just give a false sense that they are doing something worthwhile when you still need to do so much by hand. this is not idle carping. [20:13] bac: yea, the charms need so much work and jenkins, being so UI driven (vs a darn text file to config) makes it hard to setup and restore [20:14] rick_h__: well it certainly is getting much further [20:14] but then died [20:14] bac: ok that failed [20:14] needs hg installed? [20:14] uuid not installed. my bad. hg too [20:15] imports code.google.com/p/go.crypto/pbkdf2: exec: "hg": executable file not found in $PATH [20:15] ok cool, that looks like progress though [20:15] thanks [20:15] installing via the charm config so it'll be present if i download a bundle [20:18] rick_h__: i'm having a lot of trouble with the juju gooey being non-responsive. 'save changes' button goes dark but never returns, for instance. [20:18] same with 'deploy'. haven't seen this on real(tm) providers [20:18] :/ [20:33] geez, mercurial is a pig, in terms of dependencies [20:36] the move to github didn't include history? [20:41] jujugui anyone have a moment to qa https://github.com/juju/juju-gui/pull/414 ? [20:43] jrwren: it did for the gui, what are you looking at? [21:17] rick_h__: i'm stuck and unsure how to proceed [21:17] rick_h__: you still around or we can talk about it tomorrow [21:17] bac: yea, just got off call #1 [21:18] what's up? [21:18] rick_h__: here's the deal: jenkins checks out juju/charmstore into a workspace. but to be usable, it has to be in a properly set up go directory structure. [21:18] rick_h__: so i copy the workspace checkout to a freshly minted, GOPATH [21:18] ruh roh [21:18] but then 'go get' in that directory doesn't work due to git being unhappy [21:18] can we make the workspace a GOPATH? [21:19] dunno [21:20] bac: yea, so most things around jenkins means turning the workspace into the root of the world for that code/work [21:20] bac: so I'd assume we'd need to do something so that the workspace is the root/GOPATH source for all things that go on in there [21:20] bac: and that it's setup in the jenkins config via env vars or the like [21:20] rick_h__: well not the root but somewhere down the tree [21:20] bac: there's stuff in the job to define env vars and the like for the job/process [21:21] bac: ok, yea I'm not up on how GOPATH expects to work out [21:21] but in python world I'd treat it like a virtualenv root [21:22] if GOPATH is /tmp/foo then this branch needs to be checked out to /tmp/foo/src/github.com/juju/charmstore [21:22] and then it needs to fetch all of the supporting stuff [21:22] that's why i hoped just copying the freshly checked out directory to a well constructed GO directory tree would work. but it doesn't [21:23] rick_h__: anyway, that's where i'm at. need to run. talk tomorrow. [21:23] bac: ok let's chat tomorrow. Thanks for the great work on it [21:30] rick_h__: i was looking at juju/juju :( [21:30] jrwren: oh, it should have history? [21:30] maybe it does, but not branches? [21:30] they talked enough about how to do it [21:31] which, now that I think about it, makes sense given how bzr branches [21:31] right, they're all collapsed into a root history [21:31] i was hoping to git diff 1.18 with trunk [21:32] oh hmm, no tags [21:32] yea, probably have to go check out the 1.18 release in LP and check last comment or something and fine the rev in the git history to diff [21:32] pita [21:33] if that rev ever made it to git [21:33] since it was in branch, it may not have [21:33] it should have, I think all were kept. [21:33] I think the branch commits were inlined into the main one [21:33] at least that's how our juju-gui move went [21:33] but the 1.18 should have been a mainline commit/tag [21:34] in that case, I can find it in the reflog [21:59] jrwren: any luck on the reproducing? [21:59] jrwren: and if you're still around go away, day's over :P [22:02] imma keep working on it, at least until the family gets home. [22:02] and, yes, I think I did repro. [22:02] its as simple as we made it sound [22:02] but... azure: *sigh* [22:02] Failed to fetch http://azure.archive.ubuntu.com/ubuntu/pool/main/libe/liberror-perl/liberror-perl_0.17-1_all.deb 503 Service Temporarily Unavailable [22:04] i should try to repro on AWS too. [23:12] Morning [23:44] morning huwshimi [23:45] rick_h__: Hey