[00:02] rick_h_: it's *always* mongodb [00:03] I'm hitting an error that I can't track down. Here is my wip: https://codereview.appspot.com/76670043 [00:03] error: launchpad.net/juju-core/testing/testbase.PatchEnvPathPrepend(0): not defined [00:09] wallyworld: launchpad.net/juju-core/testing/testbase.PatchEnvPathPrepend(0) [00:09] ^ still importing the lp version [00:28] https://bugs.launchpad.net/juju-core/+bug/1293269 [00:28] <_mup_> Bug #1293269: juju fails to destroy environment due to missing configuration key

[00:29] great, now destroy environment doesn't work [00:32] o/ davecheney [00:32] I have a branch for the local provider lxc settings [00:32] davecheney: ah, I know what you did there [00:32] thumper: thanks [00:32] davecheney: that was all me... [00:33] they used to be omit, then I changed the default to "", and now they are omit again [00:33] pretty sure they were only omit for releases [00:33] :( [00:33] dev envs are likely screwed [00:34] here is my environment.yaml [00:34] local: [00:34] type: local [00:34] < yeah... [00:34] you will need to do the manual thing [00:35] although, I have a bug on my plate to fix destroy environment [00:35] however due to the early dying [00:35] won't help this case [00:37] fudge [00:37] thumper: has this screwed 1.17.4 -> 1.17.5 upgrades ? [00:38] davecheney: no, those are fine [00:38] the setting was changed and changed back in beteween releases [00:39] right [00:39] cool [00:39] sorta [00:40] yeah... [00:40] kinda [00:40] * davecheney remains alarmed, but not alert [00:41] rick_h_, davecheney thanks :) I'm on the packaged Mongo for Ubuntu 14.04 -- tried building 2.4.9 from source but I got tons of build errors for some reason. bleh [00:42] bodie_: that is the worst of all choices [00:42] sudo apt-get install juju-mongodb [00:42] oy [00:42] ok [00:42] much appreciated [00:42] this should have already happened for you [00:42] this package is a dep of juju-local [00:42] is this a documentation bug ? [00:44] Maybe it didn't go in because I already had Mongo [00:44] but, I don't think Go would fetch a package, would it? Just source, I thought [00:45] I'm still just getting my head wrapped around the whole thing, I think I may have been told the packaged Mongo was suitable [00:45] not suitable [00:45] manditory :) [00:45] this smells like a bug [00:46] juju should detect that you had /usr/bin/mongdb [00:46] where we rerequire [00:46] /usr/lib/juju/bin/mongodb [sic] [00:46] if you uninstall mongo [00:46] hm [00:46] and run a juju command it *should* give you reliable advice [00:47] thumper: well fuck, how do I delete that environment then ? [00:47] just table flip it ? [00:47] yeah... [00:47] lxc-destroy all the machines [00:47] stop the services [00:47] and delete the upstart entries [00:47] delete the rsyslog entry [00:47] right-o [00:47] and delete everything from /var/lib/juju/containers [00:50] I think I have some kinda conflict with a mongo already sitting on my system, but I just now removed everything mongo using apt... maybe the one I'd built from source [00:50] yeah [00:51] okay, removed that and now it's suggesting I install mongodb-server [00:51] but you're saying I need to use juju-mongodb [00:51] bodie_: this is absolteyly a bug [00:54] yay, I accomplished something [00:54] davecheney: you recently updated the ec2 provider to add ppc and arm arches to tools lookup constraints, right? [00:56] wallyworld: only ppc [00:56] someone else is tpo blame for arm [00:56] arm64 [00:56] davecheney: because we have a list of hard coded aws instances types (m1.small, etc) which are recorded as supporting i386 and amd64 only [00:56] wallyworld: i know [00:57] hence image matching will fai; [00:57] fail [00:57] this was raised in review and I believe I removed those lines were removed [00:57] which lines? [00:57] the tools lookup arches? [00:57] the ones in the ec2 and azure provider from memory [00:58] so [00:58] return &simplestreams.MetadataLookupParams{ [00:58] Series: e.ecfg().DefaultSeries(), [00:58] Region: region, [00:58] Endpoint: ec2Region.EC2Endpoint, [00:58] Architectures: []string{"amd64", "i386", "arm", "arm64", "ppc64"}, [00:58] }, ni [00:58] guess I didn't get all of them [00:58] the Architectures above should just be i386, amd64 right? [00:58] not sure about arm [00:58] no problem [00:58] i'd say no [00:58] but I think arm via stgrabers' qemu ami is possible [00:59] being realistic, just 386 and amd64 [00:59] ok. so long as we are consistent across the board [00:59] that's what i'm aiming for [00:59] then we can add in the extra arhes [00:59] arches [01:00] i just got a little confused when reading the code [01:00] do I need to submit a bug report or can I simply lbox propose my fix? [01:00] bodie_: if you can fix it yourself, go for it [01:03] so it HAS to be juju-mongodb? I thought it just had to be a version with SSL support [01:03] bodie_: they are one in the same [01:03] especially if you are on precise [01:04] hmm [01:04] not in my case I guess since I'm using trusty tahr [01:04] it installed 2.7.0-pre [01:05] bodie_: oh bodie_ the problem is so many turtles deeper [01:05] I like turtles... but not this many turtles [01:06] was really confusing trying to make my tests pass -- I knew it had something to do with the Mongo version but beyond that it gets really murky [01:06] and then having to wrangle build issues on top of that was making me insane [01:07] I think GCC 4.8.2 treats a certain thing as a warning that isn't supposed to be treated as a warning [01:07] do you know if the Vagrant box is suitable for dev? [01:07] bodie_: is the README in the project incorrect ? [01:10] ah, the make install-dependencies bit? I hadn't tried that, I think I stopped reading when I saw the directions for installing the binaries and assumed the README was about that [01:11] the juju build was failing due to an incompatibility with the latest gwacl, but someone told me about the "known working revisions" doc [01:21] anyway, thanks for the assist :) [01:26] the "PatchEnvPathPrepend(0): not defined" error I was hitting was resolved by rebuilding the pkg dir [01:27] which means the branch is now ready for review: https://codereview.appspot.com/76670043 [02:22] thumper: good news [02:22] davecheney: yes? [02:22] building static tools means we don' tneed install moew pkgs into the environment [02:30] axw: how far away is this? [02:30] // TODO(axw) 2014-02-11 #pending-review [02:30] // Embed state.Prechecker, and introduce an EnvironBase [02:30] // that embeds a no-op prechecker implementation. [02:30] wallyworld__: oops, that was reviewed and should've had an issue added... [02:31] but the EnvironBase thing may not happen, as the implementation ended up changing [02:31] wallyworld__: why? [02:31] axw: i want to introduce some more common, shared functionality across all providers [02:31] in this case SupportedArchitectures() [02:32] wallyworld__: okay. I have no immediate plan to add it, so feel free to do it [02:32] i'll just do a new interface i think [02:32] Go loves single method interfaces :-/ [02:34] just gotta think of a name, and no I don't want to use Archer cause there's no arrows to be seen anywhere [02:34] why do you need another interface? [02:35] adding it to Environ seems appropriate [02:35] guess so [02:36] i think though having an EnvironCapability interface would be good [02:36] we can add more methods to it as needed [02:36] eg is block storage supported [02:37] and easier to stub out for testing [02:38] sounds reasonable [02:45] wallyworld__: https://codereview.appspot.com/74660044 [02:46] thumper: https://codereview.appspot.com/76710043/ [02:46] :-) [02:46] np [02:50] wallyworld__: I have a parent teacher interview, but will review ASAP on return [02:50] ok, no hurry [02:50] have fun [02:51] wallyworld__: why not utils.arch ? [02:51] then arch.AMD64 [02:51] or something [02:51] thought about that, yeah can do that [02:51] I'll keep looking first [02:51] just didn't want to introduce such a small package [02:51] just wondering out loud at the first thing I saw [02:51] ok, let me keep looking first [02:52] sure, but i think you may have a valid poitn cause i almost did it that way [02:56] 2014-03-17 02:54:32 INFO juju.worker.uniter uniter.go:474 running "config-changed" hook [02:56] 2014-03-17 02:54:32 ERROR juju.worker.uniter uniter.go:480 hook failed: fork/exec /var/lib/juju/agents/unit-u0-0/charm/hooks/config-changed [02:56] just hit this with 1.17.5.1 [03:01] latest test output.... [03:01] http://paste.ubuntu.com/7105935/ [03:01] going to bed, will look over this in the morning. any input welcome [03:01] o/ [03:03] bodie_: sorry mate, still the wrong mongodb [03:04] or more speciically [03:04] the test suite tries to bring up a mongodb, but fails [03:04] most of those tests don't consider the case that mongo couldn't fail to start [03:04] and so panic during tear down [03:09] well fuck [03:09] i need to log a bug about charm-helpers-sh being missing [03:09] but i can't confirm if it works on precise because I've hit another bug [03:40] davecheney: is that hook failure one where the hook doesn't exist? [03:41] davecheney: the code paths should catch that, and I was horribly confused... [03:41] thumper: that is correct [03:41] it shouldn't happen [03:41] I don't understand how we get that error [03:42] axw: was looking into it [03:42] i think [03:42] we explicitly catch that error [03:42] and I see that explicit catch happening locally [03:42] and on ec2 [03:42] so I'm dumb struck [03:44] I was looking at what? [03:44] nope.. don't think so [03:46] oh yeah I remember this... [03:55] /insert jarring chord [03:57] morning thumper, I'm in the hangout whenever you're around [03:58] thumper: so can we now drop JUJU_TESTING_LXC_FORCE_SLOW? [03:59] wallyworld__: yes [03:59] jam: ok [04:00] thumper: wanna do that then as part of your work? [04:00] wallyworld__: yeah [04:00] ok, i'll re-review then [04:15] davecheney: so the uniter code normally works because exec.Command internally uses LookPath [04:15] LookPath is the one that returns an *exec.Error [04:15] is it not working with gccgo? [04:17] axw: this was one ec2 [04:17] with gc [04:17] huh [04:17] this is not a ppc64el specific issue [04:31] wallyworld__: https://codereview.appspot.com/74660044/ [04:31] ok [04:41] davecheney: you're on Go tip aren't you? (or > 1.2 at any rate) [04:41] davecheney: behaviour of os/exec.Command has changed in 1.3 [04:41] axw: is that a question or a statement ? [04:42] first is a question, second is a statement [04:42] then yes, and ok [04:42] the statement leads to my question, I'm backwards like that [04:42] ok [04:42] but others have hit this problem using released tools [04:42] mmk, nfi why it would fail otherwise === vladk|offline is now known as vladk === vladk is now known as vladk|away === Ursinha is now known as Ursinha-afk === Ursinha-afk is now known as Ursinha === vladk|away is now known as vladk === vladk is now known as vladk|away === vladk|away is now known as vladk === vladk is now known as vladk|away === vladk|away is now known as vladk === vladk is now known as vladk|away === vladk|away is now known as vladk === vladk is now known as vladk|away === vladk|away is now known as vladk [06:05] axw: LGTM on https://code.launchpad.net/~axwalk/juju-core/lp1293310-non-existent-hooks/+merge/211247 [06:06] jam: thanks === vladk is now known as vladk|offline [08:11] mornin' all [08:13] morning rogpeppe3 [08:13] axw: hiya [08:14] rogpeppe3: I made the changes to https://codereview.appspot.com/75990043/, but thought I'd wait until you had a glance in case you weren't happy about the new package === vladk|offline is now known as vladk [08:14] axw: looking [08:19] axw: tiny thing: i wonder if the test should use /bin/sh rather than /bin/bash, as i think that might be what cloud-init is using. [08:19] axw: (i may be wrong there though) [08:20] rogpeppe3: the synchronous bootstrap bit is always bash [08:21] axw: ok, and the DumpFileOnError code is only ever run in that context? [08:21] rogpeppe3: and add-machine ssh, but that's also bash [08:22] axw: LGTM === vladk is now known as vladk|offline [08:22] thanks === vladk|offline is now known as vladk [08:39] * jam => luncdh [08:56] wallyworld_, are you around? [09:02] morning dimitern [09:02] jam, morning [09:02] dimitern: I missed you earlier [09:02] rogpeppe3: 1:1? [09:03] jam: ah yes! [09:03] jam, i'm sorry i trusted my nexus ubuntu touch to sound the alarm but it didn't and i read later that it's not implemented yet [09:27] dimitern: hi [09:27] dimitern: hi [09:36] hey vladk [09:36] wallyworld_, I have a question about upgrades [09:36] sur [09:36] e [09:36] wallyworld_, we need to create some dirs with root permissions [09:36] ok. where? [09:36] wallyworld_, specifically, when upgrading rsyslog config and logdir [09:37] wallyworld_, what's the best way to do that? i've seen provider/local calling exec.Command("sudo", "/bin/bash", "-s") and attaching stdout and stderr [09:38] dimitern: the machine agent runs as root, right? [09:38] wallyworld_, or maybe generate a script and pass it on stdin (i.e. mkdir -p ) [09:38] wallyworld_, so you're saying we should be able to create /var/log/juju- in an upgrade step? [09:38] i think so [09:38] ok, i'll try that [09:39] it's just an educated guess [09:39] but would be the simplest i think [09:39] cause an upgrade step is easy enough to write, and is only executed the once when needed to get to 1.18 [09:40] wallyworld_, yeah, i was testing my fix for bug 1291400 and found out some errors due to the logdir in var/log not getting created and hence rsyslog cannot create its certs there [09:40] <_mup_> Bug #1291400: migrate 1.16 agent config to 1.18 properly (DataDir, Jobs, LogDir)

[09:42] dimitern: good luck, let me know if it works. my only question is if the owner of the dir would be root but i think so since machine agent runs as root [09:42] wallyworld_, it has to be a specific user/group i think, i'll check [09:43] maybe, not sure ottomh [09:43] * dimitern really hates how we don't chown local provider logs to the user so I don't have to do sudo less ... [09:43] even so, the process will have the priviliges to do it [09:43] dimitern: i agree. file a bug :-) [09:44] will do :) [09:44] hola [09:48] hi all [09:48] voidspace: ping [09:48] hey rogpeppe3 [09:49] wwitzel3: yo! [09:49] rogpeppe3: you have a good weekend? [09:49] wwitzel3: an excellent weekend, thanks [09:49] wwitzel3: we actually had some sun yesterday, and even better, managed to go out and take advantage of it [09:50] wwitzel3: you? [09:50] rogpeppe3: that's great :) [09:51] rogpeppe3: yeah, it was good, nothing exciting, but nice and relazing [09:51] relaxing [09:53] rogpeppe3, morning [09:53] voidspace: hiya [09:53] rogpeppe3, hi :-) [09:53] rogpeppe3, sorry about Friday [09:54] voidspace: how's your configuration working now? [09:54] rogpeppe3, well, mostly good [09:54] rogpeppe3, :-) [09:54] rogpeppe3, I can't get the USB monitor working with Ubuntu and I think I'm stuck on a maximum of three monitors [09:54] rogpeppe3, but other than that, good [09:55] rogpeppe3, I still need to move the drives to the new machine [09:55] voidspace: only three monitors, how can you survive? :-) [09:55] heh [09:55] it is hard [09:55] anyway [09:55] rogpeppe3, shall I pair with you today? [09:55] voidspace: i was just about to suggest that, yes [09:55] great [09:56] voidspace: whenever you're ready [09:56] rogpeppe3, I'm pretty much ready [09:56] rogpeppe3, hangout for audio? === rogpeppe3 is now known as rogpeppe [09:57] voidspace: let's start with that [09:58] * rogpeppe tries to work out how to create a hangout [09:58] rogpeppe: shall we use the juju-core one, or start our own? [09:59] voidspace: let's just use the juju-core one [09:59] voidspace: https://plus.google.com/hangouts/_/canonical.com/juju-core [10:08] jam: are you around? [10:09] mgz: I am, I just need to do a quick "get some coffee/tea" run, and then I'll be happy to chat [10:09] mumble or G+ ? [10:09] sure thing [10:09] lets try mumble [10:16] mgz: no love.... [10:16] I'll try again [10:17] mgz: so I'm guessing this is Dubai's anti-VOIP software going into effect. [10:17] we'll have to make a new hangout [10:17] cunning [10:17] At least, it is what it sounds like when I try to use Skype-out [10:17] mgz: there is a hangout associated with our 1:1 calendar event [10:17] can you get to that? [10:17] is the canonical mumble server not over ssl? [10:18] lets see [10:19] mgz: have you got the link to the instructions for reconfiguring the gobot, please? [10:20] rogpeppe: https://lists.ubuntu.com/archives/juju-dev/2014-March/002182.html [10:21] jam: thanks! [10:21] rogpeppe: he's in G+ with me, and its easier to paste via me, then copy it to IRC :) [10:27] mgz: is there a reason that the verify_command in the tarmac configuration doesn't do a godeps -u ? [10:28] because it would also need go get -u [10:28] rogpeppe: because half the time that would just break because you have to fetch first? [10:28] and go get -u can overwrite the current branch that we are trying to test? [10:28] jam: in that case, it *should* break, i think [10:29] jam: otherwise we're testing with the wrong deps [10:30] jam: currently i'm trying to help michael get his Instances aggregration branch in, and there's no easy way to do it [11:06] rogpeppe, https://pastebin.canonical.com/106567/ [11:06] voidspace: yeah, just fixed it, i think [11:08] rogpeppe, cool, thanks [11:08] rogpeppe, nope :-) [11:08] voidspace: indeed [11:10] now the hook borked [11:16] rogpeppe, I'm grabbing coffee [11:16] voidspace: k [11:32] hi everyone === vladk is now known as vladk|away [11:40] fwereade: just encountered an interesting error with juju resolved [11:41] * fwereade peers nervously at rogpeppe [11:41] hi again perrito666 :) [11:41] fwereade: it's not too bad actually [11:42] fwereade: i did "juju resolved", and the hook was still in error state [11:42] fwereade: i did it again, and it said "ERROR cannot set resolved mode for unit "tarmac/0": already resolved" [11:43] fwereade: but that was because a hook was still running [11:44] fwereade: i wonder if the uniter should set the unit status out of error state the moment it starts to try rerunning the hook [11:45] rogpeppe, it *was* designed that way but I'm trying to remember why [11:45] fwereade: BTW the first time i did "juju resolved --retry" [11:46] fwereade: it certainly felt a bit weird when it happen: "why can't i resolve this hook that's in error state?") [11:46] rogpeppe, IIRC the idea is that we make sure we actually try to handle the resolved before we mark it ready to try again [11:46] happened [11:46] fwereade: i think that's the bit that's not so intuitive [11:46] fwereade, I've been exchanging emails with Erik Naslund :-) [11:47] rogpeppe, the failure mode of setting resolved early is that we could not actually respond, but claim we did [11:47] fwereade: because the user has no visibility into when hooks are actually running [11:47] rogpeppe, I *think* that's worse [11:47] voidspace, ah cool -- how's he doing? [11:47] voidspace, say hi :) [11:48] fwereade, I will do [11:48] fwereade, he is using mock at his startup and sent an email saying thanks [11:48] voidspace, excellent, I'm pretty sure I introduced him to it :) [12:04] ng [12:05] natefinch: ping [12:07] wwitzel3: howdy [12:08] natefinch: quick hangout? [12:08] sure [12:10] x [12:13] is there a standard for which version of ubuntu this should work best on? davecheney says I'm still using the wrongodb, but this is definitely the one make install-dependencies set me up with [12:14] bodie_: most testing is done on Precise, we tried to enable Trusty correctly, but then ended up regressing support elsewhere, so we have a patch in progress to restore proper support for Trusty's local provider [12:15] makes sense [12:15] is there anything I can do to help there? [12:15] maybe I'll try the Vagrant setup in the meantime [12:16] fwereade: heya. did you want to review https://codereview.appspot.com/70190050/ before Jesse lands it? [12:16] it's a big 'un [12:17] looks (nearly) good to me, but just checking if I should provision a LGTM on your review [12:19] rogpeppe: yeah, I found that (resolved not changing error state immediately) a bit unintuitive too. some sort of feedback would be nice [12:25] axw, if I haven't looked at it by eod I think we can go ahead,it [12:25] just downloaded the 639MB vagrant image in almost exactly a minute... love this fios [12:25] fwereade: okey dokey [12:25] 's had several rounds and you were making smart comments last time I saw [12:26] axw, you can represent me :) [12:26] I'll do my best [12:26] axw, btw I have been trying to assimilate all the awesome azure changes, couple of quick questions [12:27] axw, 1 (possibly trivially working, probably deserves minor reflection) subordinates in fancy-azure mode [12:27] axw, I suspect they will all work fine, but if you haven't considered them explicitly please do some thinking [12:28] I haven't, and I think they'll work fine, but yes I will have a think about it [12:28] axw, 2 (maybe more of a sticking point) can we change the startinstance params so that it's expressed in terms of instances to avoid, rather than leaking the concept of "principals" into environs? [12:29] axw, I haven't figured out whether that screws everything up wrt AddMachine though [12:29] axw, I suspect they will all work fine, but if you haven't considered them explicitly please do some thinking [12:29] I haven't, and I think they'll work fine, but yes I will have a think about it [12:30] sorry, net cut out [12:30] axw, 2 (maybe more of a sticking point) can we change the startinstance params so that it's expressed in terms of instances to avoid, rather than leaking the concept of "principals" into environs? [12:30]