[00:01] davecheney: yo [00:08] m_3: wazzup ? [00:10] davecheney: just wanted to touch base with you on scaling [00:10] davecheney: make sure you weren't waiting on me for anything [00:14] m_3: nope [00:14] did you use the test harness on friday ? [00:14] i was going to bring it up to the latest changes and try the 300 node test again [00:14] HP cloud totally shat itself once I got to 300 node last week [00:16] davecheney: saw email on it, but didn't get a chance to play [00:16] took a real weekend :) [00:17] m_3: no probs [00:17] you should continue to take that weekend [00:17] haha [00:17] yeah, sprinting this week [00:17] conference last [00:17] conference next [00:17] then another sprint [00:17] whoohoooo! party [00:18] shit, you travel more than I do [00:18] you should sell your house, or rent it out and live in your basement [00:19] haha, yeah [00:19] it's utah, so it's cheap [00:20] houses in australia don't have basements [00:20] and if they did [00:20] we'd convert them into garages [00:21] we do rack up the hilton points though [00:21] unfortunately it's all in my wife's name.... so I'm always Mr Martin [00:21] not Mr Mims [00:21] and not even Dr Martin [00:22] shitter [00:48] * thumper has installed the ec2 command line tools to try to work out the difference between us-east-1 and ap-southeast-2 [00:48] yay [00:49] thumper: digging in boto is also a good way to figure out how this shit is supposed to work [00:49] boto? [00:51] python boto [00:51] _the_ python ec2 library [00:51] ah [00:52] * davecheney is really wishing he hadn't let the TA talk him into flying via LAX next week [00:52] heh [00:52] * thumper has AKL->SFO [00:52] but the only option for SYD -> SFO [00:52] was flying united [00:52] yep, that sucks [00:53] * thumper likes Air New Zealand [00:53] UA flying an original 1970 vintage 747-400 [00:53] fuck that [00:54] so now I have a 2 hour layover in LAX to clear customs (who are on strike) and land on time (flight controllers are also on strike) and transfer to a domestic flight [00:54] soooo not going to happen [01:01] why is everyone striking? [01:06] furlough [01:07] air traffic controllers are paid with federal money [01:07] and it's run out [01:08] http://www.13wham.com/news/local/story/FAA-Furloughs-Kick-In-But-Flights-Come-In-On-Time/81mfoekV-0SKgJ8HusTqMA.cspx [01:21] oh yay [01:21] * thumper has just realised a big problem... [01:21] well, bigish [01:21] FSVO big [01:21] do tell [01:21] * thumper tries to work this out [01:22] when running tests we have MakeFakeHome [01:22] which writes the environments.yaml into the juju home dir [01:22] however MakeFakeHome doesn't take JUJU_HOME into account, but the code to work out the juju home to write the config file into does [01:22] so if any dev has JUJU_HOME set [01:22] the tests will break their environment [01:23] oops [01:23] * thumper enfixorates [01:23] * thumper files bug first === wedgwood is now known as wedgwood_away [03:54] * thumper thinks about how to side step this panic [07:10] mornin' all [07:15] morning === ChanServ changed the topic of #juju-dev to: https://juju.ubuntu.com | On-call reviewer: TheMue | Bugs: 2 Critical, 64 High - https://bugs.launchpad.net/juju-core/ [09:57] I need a floating ip on canonistack, has anyone got one I can have? [11:24] mgz, wallyworld_, jam, danilos: sorry guys I might be 5m late for the standup [11:25] ok [11:25] np [11:25] jam, dimitern, wallyworld_: I am not sure I'll be able to get mumble to work here at all :/ I think I'll drop out to find a new PSU instead of trying to fix the existing one [11:26] what's wrong with it? [11:27] wallyworld_: he is on his laptop, and having trouble getting it working. The PSU apparently blew out some capacitors [11:34] jam, wallyworld_: I have mumble working here but with a high-pitched noise (like echo) coming in from the built-in microphone, and I am sure nobody wants to listen to that [11:34] danilos: you can maybe just listen in [11:35] danilos: yeah you can sit in and listen [11:39] so, I can't connect to the standard mumble port right now [11:39] it appears to be a day of mumble fail [11:39] if there's a public server on 443 I could get to that instead [11:43] smgz, have you tried setting up an ssh tunnel? [11:50] jam: forgot to mention - i had goose bot issues as well - it wouldn't land stuff and ssh to the goose bot machine failed [11:50] wallyworld_: yeah, I haven't been able to ssh into canonistack instances for a while, I was hoping mgz would have some insight, but he only just got back. [11:51] jam: so i justed landed my branch manually [11:51] wallyworld_: right, I've been doing that for the last few patches until we get goose-bot sorted out again. [11:51] ok, glad it's not just me then :-) [11:54] danilos: have one now, it's... not going to be much good for voice packets [11:55] udp over tcp, great latency and reliability at the same time [11:57] dimitern: thanks for the review. i copied that imagesFields() function from ec2 - i'll look for a reference to the file format and point to that [11:57] also, because of lack of floating ips, I'm going hpcloud->chinstrap->canonistack [11:57] were I just in the datacenter, it might be more practical [11:58] wallyworld_: cheers! [12:50] test [13:02] hi guys, I noticed that with go-juju the binaries like open-port, close-port, juju-log and others are in /var/lib/juju/tools/machine-1/ [13:02] I suppose you change PATH, so that all hooks can easily find them, right? [13:03] ahasenack: that's right [13:03] I just stumbled upon a charm/script that sanitizes PATH to the usual bin paths, and then gojuju broke [13:03] ahasenack: interesting [13:03] ahasenack: where are those executables in py juju? [13:04] rogpeppe1: /usr/bin [13:04] ahasenack: i'm not sure there's anything we can do about that unfortunately [13:04] ahasenack: we can't put binaries in /usr/bin [13:04] why? Ubuntu policy? [13:04] ahasenack: because we allow upgrades [13:04] we can fix the script... [13:04] ahasenack: and there can be more than one version of the tools binaries on the same machine [13:05] or symlink probably, but really fiddingly with path shouldn't be needed [13:05] one per deployed unit, basically? [13:36] ahasenack: one for each agent - that is one per deployed unit and one for the machine agent [13:40] rogpeppe1: hm, unrelated, but did bootstrap with raring break again? It was working yesterday [13:40] ahasenack: yes, it's just been deliberately broken again [13:40] ahasenack: pending a proper fix [13:40] ok [13:40] ahasenack: which is an outstanding bug in raring [13:41] isn't raring going to be released tomorrow? [13:41] no, 25t [13:41] 25th [13:41] hardy dies tomorrow [13:42] so, hm, I can't really test my PATH fix in this charm [13:43] ahasenack: try with a slightly older version of juju-core [13:43] the ppa gets rid of the older ones [13:43] I can postpone this, np [13:48] ahasenack: you can use the version currently in the ppa [13:49] smgz: it's what I had installed, bootstrap wasn't working === smgz_ is now known as smgz === wedgwood_away is now known as wedgwood [14:03] rogpeppe1: kanban? [14:03] dimitern: i'll see how the network works today... [14:08] darn it! [14:08] rogpeppe2: no chance [14:09] phone company says it might be fixed tomorrow. i'm not gonna hold my breath [14:11] dimitern, mramm: sorry, i'm giving up. my mobile connection (previously quite reliable) doesn't like me now either [14:12] rogpeppe2: that's fine [14:51] rogpeppe3, ping? [14:57] mattyw: pong [14:58] rogpeppe3, I'm trying to run a unit test using the MgoTestPackage stuff in juju-core, do we have a version of mongo that will run on a 32 bit machine? [14:58] mattyw: i'm not sure [14:59] rogpeppe3, I think we might be ok actually, we can get the enterpise edition of mongo free for development [15:00] mattyw: cool [15:24] rogpeppe3, is the juju-core api exposed or is it only available internal to the env? [15:24] in a security group sense [15:24] hazmat: it's exposed [15:25] rogpeppe3, cool thanks [15:25] hazmat: are you planning on using it? === rogpeppe3 is now known as rogpeppe [15:27] rogpeppe, yes.. i've got rewriting launchpad.net/juju-deployer on my plate for this week [15:27] its mostly cli forks.. but seems like a nice opportunity to use the api. [15:28] hazmat: staying in python, or rewriting in Go? [15:29] rogpeppe, not sure yet [15:30] might be better in go, if there's a desire to move into core [15:30] hazmat: it's a small enough rewrite that rewriting in Go is probably a net win [15:30] hazmat: 'cos you can use the api package directly [15:31] hazmat: you might want to consider putting it under https://launchpad.net/juju-utils i suppose [15:32] rogpeppe, thanks for the link [15:32] hazmat: speaking of which, here's the current way to open an API client from Go. i plan on moving some form of this function into core: http://bazaar.launchpad.net/~juju/juju-utils/trunk/view/head:/cmd/juju-wait/main.go#L134 [15:33] hazmat: PS spot the deliberate mistake :-) [15:33] rogpeppe, envname ;-) [15:34] hazmat: indeed [15:35] rogpeppe, i've had problems using any juju-core env.. any subsequent command post bootstrap does.. error: no CA certificate in environment configuration [15:35] rogpeppe, do you know what might cause that ? [15:35] hazmat: hmm, sounds like the CA certificate isn't being saved or loaded correctly [15:36] i am using JUJU_HOME to separate pyjuju/gojuju envs [15:36] hazmat: are you using the env from different machines? [15:36] rogpeppe, nope [15:36] rogpeppe, same machine, multiple trunk builds w and wo upload-tools, multiple regions.. etc. [15:36] hazmat: there has been quite a bit of churn in that area since i touched it. i'll just have a look. [15:36] might just be an issue with JUJU_HOME [15:37] hazmat: that's what i think [15:37] hazmat: it's very new [15:38] hazmat: yes, that's the bug [15:38] * hazmat reports a bug [15:38] hazmat: it always saves it to $HOME/.juju [15:38] and reads from $JUJU_HOME [15:39] hazmat: yup [15:39] hazmat: the bug is in environs.WriteCertAndKeyToHome [15:40] bug 1171910 [15:40] <_mup_> Bug #1171910: JUJU_HOME is broken, writes cert to wrong location [15:41] * rogpeppe welcomes _mup_ back [16:10] is there an lp2kanban instance running on the juju core board? [16:15] rogpeppe, fix in review [16:15] hazmat: ha, i've done one too [16:16] doh.. bad coordination [16:16] hazmat: https://codereview.appspot.com/8839047 [16:16] hazmat: very very slow upload speed here [16:17] rogpeppe, i went minimal.. https://code.launchpad.net/~hazmat/juju-core/fix-juju-home/+merge/160423 [16:19] hazmat: if poss, we generally go through codereview [16:19] rogpeppe, understood, having a credential issue with google atm [16:21] hazmat: i think that fix isn't quite right - the authoritative source for juju home is config.JujuHome() [16:21] * rogpeppe is still uploading [16:22] 56Kbaud FTW [16:22] rogpeppe, ack.. one minor with your branch is that it looses the cert write test if JUJU_HOME isn't set. [16:23] hazmat: it will panic, actually [16:23] hazmat: if config.SetJujuHome hasn't been called [16:24] hazmat: but MakeEmptyFakeHome calls it [16:24] hazmat: hmm, sorry ignore me [16:24] hazmat: speaking without looking [16:24] hazmat: i'm not sure what you mean then [16:25] hazmat: you mean it doesn't test what happens if $JUJU_HOME isn't set? [16:25] rogpeppe, yes [16:26] rogpeppe, but its not clear if that's auto setup via config to HOME and tested elsewhere [16:26] hazmat: the right place to test whether config.SetJujuHome is called is in cmd/juju [16:27] hazmat: the JUJU_HOME env var isn't part of the contract of the environs package [16:27] ack [16:28] hazmat: there is no test, but i'm disinclined to grumble. the call to InitJujuHome is at the start of Main. [16:29] hazmat: and juju.InitJujuHome itself is well tested === deryck is now known as deryck[lunch] [16:29] hazmat: and that's where the $JUJU_HOME / $HOME logic is [17:12] i'm done for the day [17:12] see y'all tomorrow [17:12] rogpeppe: cu, and much luck with your phone line === deryck[lunch] is now known as deryck [18:06] hi. I want to announce the GUI Juju core support today. What is the end user expected to do to start up juju core? I'd like to try it from that direction and make sure it works [18:06] ppa? [18:07] gary_poster: that sounds awesome [18:07] :-) [18:07] yep, ppa [18:09] https://launchpad.net/~juju/+archive/devel is the latest [18:09] mramm, cool, thanks that's what I was hunting for. [18:39] gary_poster: is bootstrap working on raring yet with juju-core? [18:39] it wasn't earlier today [18:40] ahasenack, my raring machine is still updating :-P . are you getting "error: no matching tools available"? [18:40] gary_poster: yes [18:41] gary_poster: they said it was broken on purpose today until the right fix lands [18:41] gary_poster: so maybe you should check before announcing [18:42] ahasenack, oh! ok. Yeah id does that on quantal too. I was wondering if they expected people to use --upload-tools until release or something [18:42] s/id does/it does/ [18:43] mramm do you know of a timeline when that will be resolved? I agree with ahasenack that it would be nicer to announce when the story is smoother [19:02] hola! heard rumours that the go version has reached feature-parity with the python version. is this what the announcement is about? [19:04] sidnei, my announcement would be that the GUI works fully with juju core, with feature parity. [19:04] ah, so pretty close [19:05] sidnei, the other announcement may be waiting in the wings, dunno :-) [19:05] does juju-core work with canonistack yet? or does it still need public ips? [19:05] I have not tried that. Still using pyjuju for that [19:21] gary_poster: I think mgz was working on it [19:21] gary_poster: I'll check in with him first thing in the morning (his time) [19:21] so I'll know before you get in to work ;) [19:22] sidnei: go juju works with canonistack [19:22] mramm: no more public ips issues? [19:22] sidnei: but it does not have constraint enforcement for openstack yet (code in a branch, will be merged in the next release) [19:23] thanks mramm [19:23] sidnei: well, there are public IP issues on canonistack, but there are ways to get around it built in [19:23] now [19:23] mramm: not even instance-type for constraints? [19:24] there will be mem, cpu/arch constraints (cross platform constraints) as soon as john's branch lands [19:25] but there was a lot of arguing about provider specific (instance-type) constraints, so those are not yet implemented at all [19:25] mramm: what about those workarounds to get around public ip issues, is it written down somewhere? [19:29] mramm, probably already known, but fwiw, when working with the PPA, juju bootstrap --upload-tools works fine in raring and quantal, but juju deploy on raring is giving me "error:cannot log in to admin database: auth fails". Works fine in quantal. Maybe I am doing something wrong; will investigate later [19:29] I couldn't get it to work with canonistack the other day, some error when trying to contact swift, I think 401 [19:29] but then the bootstrap errors started, and I didn't try again with canonistack [19:30] sidnei: I believe it is written down, but I can't find it at the moment [19:30] mgz and jam would know [19:30] gary_poster: that seems like something I am not yet aware of -- will look into it [19:30] thanks [22:53] hm.. if you have multiple relations, and one is in error, you can't tell which one has the error. === wedgwood is now known as wedgwood_away [23:30] pip install jujuclient