[02:23] hey wallyworld, looks like there's a windows unit test failure on master from the new xdg home dir work: http://reports.vapour.ws/releases/3580/job/run-unit-tests-win2012-amd64/attempt/1891#highlight [02:34] cherylj: ok, will look. we really need to gate on windows unit tests. not sure why we don't yet [02:34] yeah, would be nice [02:34] we have been talking about it for months and months [02:35] just hasn't made it to the top of their list of priorities, I guess [02:36] yeah. seems like a small change for a big win though [02:56] axw: a small tweak to fix windows unit tests http://reviews.vapour.ws/r/3763/ [02:57] wallyworld: are we trying to use XDG things on Windows? :/ [02:58] axw: no, there's a separate func to get the right location, but the tests have always been linux specific [02:58] ok [02:59] axw: on windows, it's %APPDATA%/juju [02:59] wallyworld: done [02:59] ty [02:59] wallyworld: ok, thanks [04:10] oh god, so many test failures. My eyes! MY EYES!! [04:11] wallyworld, perrito666, so did the new juju home stuff break out JUJU_HOME into JUJU_DATA and something else? [04:11] because it looks like CI wasn't [04:29] cherylj: JUJU_DATA replaces JUJU_HOME, but it will default to something sensible. if CI uses JUJU_HOME, then i'll need to back out that change [04:31] or i might set JUJU_DATA to JUJU_HOME till CI is updated [04:31] wallyworld: that sounds like a good compromise. I'll bring it up to the QA guys in the morning [04:31] ok. i'll do that now [04:31] thanks, wallyworld! [04:32] sure, sorry about the breakage [04:44] axw: another small one http://reviews.vapour.ws/r/3764/ [04:45] wallyworld: done [04:45] ty [04:47] wallyworld: whenever you're ready: http://reviews.vapour.ws/r/3765/ [04:47] sure [04:57] axw: how much smaller is asn encoding vs json? [04:57] did it make much of a difference? [05:04] wallyworld: it was reasonably significant. I can't recall the overhead exactly. I can measure again if you like [05:04] no need, just curious [05:21] axw: a couple of small things [05:33] axw: when you are free, another trivial http://reviews.vapour.ws/r/3767/ [06:09] what's changed now ? https://bugs.launchpad.net/juju-core/+bug/1542985 [06:09] Bug #1542985: Juju won't switch environments [06:13] axw: wallyworld what has juju switch been renamed to ? [06:13] davecheney: that one stays the same [06:13] let me double check [06:13] see linked bug [06:14] yup, stays as switch [06:14] why can't I change to ap-southeast-2 anymore [06:14] hmmm, haven't see that bug till now [06:15] it's there in juju/environemnts.yam [06:15] it's there in juju/environemnts.yam [06:15] it's there in juju/environemnts.yaml [06:15] it's supposed to work [06:15] something broke :-( [06:16] lucky(~/src/github.com/juju/juju) % juju bootstrap -v --debug [06:16] 2016-02-08 06:15:41 INFO juju.cmd supercommand.go:59 running juju [2.0-alpha2 gc devel +fa5e547 Sun Feb 7 03:18:28 2016 +0000] [06:16] 2016-02-08 06:15:41 ERROR cmd supercommand.go:448 the name of the model must be specified [06:16] lucky(~/src/github.com/juju/juju) % juju bootstrap -v --debug ap-southeast-2 [06:16] we really hate our users [06:16] error: unrecognized args: ["ap-southeast-2"] [06:16] lucky(~/src/github.com/juju/juju) % juju bootstrap -v --debug -m dev [06:16] 2016-02-08 06:16:36 INFO juju.cmd supercommand.go:59 running juju [2.0-alpha2 gc devel +fa5e547 Sun Feb 7 03:18:28 2016 +0000] [06:16] 2016-02-08 06:16:36 ERROR cmd supercommand.go:448 open /home/dfc/.local/share/juju/environments.yaml: no such file or directory [06:17] juju home has changed [06:17] i'm in the process of updating release notes [06:17] it only landed on the weekend i think [06:17] davecheney: FYI, in the near future we will completely ignore environments.yaml [06:17] just copy ~/.juju to ~/.local/share/juju [06:18] more info later [06:18] lucky(~/.local/share) % ls -al . | grep juju [06:18] lrwxrwxrwx 1 dfc dfc 15 Feb 8 17:18 juju -> /home/dfc/.juju [06:18] problem solved [06:18] ~/.juju will be ignored in a day or so [06:19] as soon as CI is updated [06:19] then it will be ~/.local/share/juju [06:19] looks like it's being ignore now [06:19] i landed a change maybe 2 hours ago to temporarily use it [06:19] until CI is fixed [06:20] so just copy stuff to ~/.local/share/juju to be future proof [06:20] commit fd9cd4bc60d55571ce2db8562462703ed3baa228 [06:20] Author: Ian Booth [06:20] Date: Mon Feb 8 14:41:59 2016 +1000 [06:20] Fall back to JUJU_HOME if JUJU_DATA not set [06:20] yup, i have that rev [06:20] you need $JUJU_HOME set [06:20] that's what it falls back to [06:20] 2016-02-08 06:20:42 DEBUG juju.provider.common bootstrap.go:318 connection attempt for 54.79.164.245 failed: Warning: Permanently added '54.79.164.245' (ECDSA) to the list of known hosts. [06:21] /var/lib/juju/nonce.txt does not exist [06:21] 2016-02-08 06:20:48 DEBUG juju.provider.common bootstrap.go:318 connection attempt for 54.79.164.245 failed: /var/lib/juju/nonce.txt does not exist [06:21] ^ these won't exist [06:21] my user has no permission to write to those locations [06:21] right, but that's not what we're taling about [06:21] it's the XDG data home [06:22] ok [06:22] as opposed to XDG config home [06:22] or XDG runtime home etc [06:22] Bug #1542985 opened: Juju won't switch environments [06:22] Bug #1542988 opened: bootstrap tries to write to /var/lib/juju on local machine [06:22] oh yuk [06:22] really? [06:22] what provider [06:23] ec2 [06:24] the nonce.txt is written by cloud init i believe [06:24] the bootstrap client has never used that file [06:24] yes, written by cloud-init. bootstrap looks for it to check that cloud-init has finished [06:24] how the fark it is getting written locally i'm not sure [06:24] maybe that error is leaking through from the remote side [06:27] wallyworld: PTAL [06:27] sure [06:30] axw: lgtm. we need to double check that a newly added user on a controller only has access to the models they create or are shared with them. i can't recall whether in the initial poc done with the initial jes work all users had admin access on the controller or not [06:52] Bug #1542990 opened: tools version checker restarts every 3 seconds [06:52] Bug #1542992 opened: ec2: destroy-controller causes rate limiting failure [07:16] wallyworld: yep. I did test, but not very rigorously, and it appeared to be checking access [07:17] axw: ok, we'll need to ensure going forward it works as we hope [07:37] axw: if you get a chance at some point, could you take a another pass at http://reviews.vapour.ws/r/3737/ now that it's been updated? [07:37] wallyworld: yup, was about to, just fixing my branch [07:38] awesome ty [08:02] wallyworld: what would you call the combination of controllers.yaml, accounts.yaml, models.yaml (sic), etc.? currently called "Cache" in that branch, but it doesn't sit well with me [08:02] wallyworld: only the model info is really a cache [08:02] axw: agreed [08:02] um [08:04] LocalControllerState ? [08:05] it's local to that machine, it's related to controllers (and accounts for those), and it's state cached in memory or on disk [08:05] naming is hard :-) [08:06] wallyworld: a bit better, but I'm not completely sold on that either. I'll think about it for a bit. [08:06] yeah, i wasn't that happy with that suggestion [08:09] wallyworld: I'm thinking ControllerStore, NewFileControllerStore, NewMemControllerStore, DefaultControllerStore [08:09] hmmm, that works [08:09] wallyworld: along the lines of existing configstore [08:09] it's similar to what's there [08:09] yup [08:28] wallyworld: axw: my problem with "store" is that its prime meaning is not "storehouse, warehouse" but actually is closer related to "retail" [08:28] apart from that - i'm phased at all.. whatever [08:28] disagree :-) [08:28] in computer terms, store ahas a well defined meaning [08:30] m not* [08:31] wallyworld, hey there [08:31] hey [08:31] wallyworld, re that juju home 2.0 change [08:31] should i brace? [08:31] wallyworld, how is going to work with non-public clouds - i.e. maas ? [08:31] wallyworld, juju bootstrap mymaas maas/? [08:32] assuming the lack of environments.yaml [08:32] dimitern: there's a new clouds.yaml for private clouds [08:32] dimitern: hop onto #juju [08:32] omw [08:52] dimitern, morning! [08:53] dimitern, are merging master -> maas-spaces? I started but need to decide whether I should continue. [08:54] frobware, morning [08:54] frobware, yeah, I think it's good to do it often until we can merge [08:55] dimitern, I didn't finish resolving all the conflicts [08:55] frobware, the bigger merge landed on friday [08:55] frobware, so this morning it was much easier - only 1 test to fix [08:55] dimitern, ah - that's the one I'm talking about. I started on Friday, but didn't complete. [08:56] frobware, I've run make check after the merge on friday and did a quick live test to be sure [08:57] dimitern, and that's now landed in maas-spaces? [08:57] frobware, this morning's merge is landing now [08:57] frobware, and the friday's one is already in [08:58] frobware, I have a few tasks planned to unblock maas-spaces merging into master [08:58] frobware, so we can then concentrate on the remaining tasks [09:32] frobware: we had a CI run on Friday, on maas-spaces, that never completed [09:32] frobware: http://reports.vapour.ws/releases/3578 [09:32] frobware: but there was a xenial unit test failure which isn't good - featuretests [09:34] frobware: that maybe looks space discovery related, which is unexpected [09:34] voidspace, 2 of those attempts where due to issues with the cloud archive getting the wrong package [09:34] frobware: however they pass locally [09:35] dimitern: ah right, infrastructure problem then? [09:35] dimitern: the featuretests pass locally for me (wily), but I get a cmd/status failure [09:35] voidspace, and the last one is flaky due to the controller space not getting handled [09:35] yeah, we're still expecting that to fail [09:35] I believe [09:41] cmd/juju/status fails consistently for me [09:41] checking master [09:41] voidspace, how does it fail? [09:42] dimitern: six failures, unexpected number of machines in some of them [09:43] dimitern: same failure for me on master [09:43] dimitern: huuuuge amounts of logging so hard to summarise the failures, I'll pick one of the tests and investigate [09:43] dimitern: it could be an oddity with my machine [09:44] hah, latest run of master on CI failed to build [09:45] Wily unit tests passed on Friday on CI though [09:48] voidspace, I see both master and maas-spaces-merge-master-7* pass here, although I did see the LoginsDuringUpgrade test failing once [09:48] dimitern: how did it fail? [09:54] voidspace, something like can login as machine expected IsTrue, got false [09:54] ah [09:56] with status test, I'm getting five machines when only one was expected [09:56] digging into it [10:50] dimitern: http://pastebin.ubuntu.com/14992568/ [12:19] dimitern: here's a trivial one line fix for a failing windows unit test (i hope), can you take a quick look? http://reviews.vapour.ws/r/3771/ [12:20] wallyworld, looking [12:20] ty [12:21] wallyworld, LGTM [12:21] tyvm [12:21] wallyworld, and I assume a better fix will be done there once possible, right? [12:21] dimitern: that whole line will be deleted [12:22] since JUJU_HOME won't be supported and the fallback in the code will be gome [12:22] but CI scripts need to be updated [12:22] wallyworld, right [12:22] the test fails because of a temporary fallback in the code [12:29] wallyworld: hey, you know anything about status tests? [12:29] a littlw [12:30] i do know i hate them [12:30] wallyworld: there's a few that fail consistently on my machine (but not anywhere else) [12:30] wallyworld: here's one http://pastebin.ubuntu.com/14992568/ [12:30] looking [12:30] wallyworld: the test is trying to limit status to an exposed service but gets five machines instead of one [12:31] wallyworld: that's running just that test with the preceding tests in the loop removed (so line numbers won't match yours) to eradicate state leakage between tests as a cause [12:31] and it's intermittent? [12:31] not on your machine [12:32] wallyworld: consistent on my machine, not seen elsewhere yet I believe [12:32] wallyworld: but in means I can't do a full unit test run without failures on my machine - so I'd like to find out the cause [12:33] indeed [12:33] i get mongo failures quite regularly :-( [12:36] voidspace: i don't see anything obvious, how long has it been doing this? [12:39] wallyworld: new today [12:39] wallyworld: thanks for looking anyway [12:40] voidspace: i do notice what appears to be old agent-state values in the yaml output. those are gone from master. you could try rebasing [12:41] wallyworld: I've just had a thought - I copied environments.yaml from CI [12:41] I wonder if that is screwing with things [12:41] wallyworld: I'm on current master head [12:41] probably not, but yoy could try without, i'd have to see what's in it i guess [12:42] shouldn't affect the tests [12:44] nah, doesn't affect it - still 6 failures [12:56] hmmm, there must be something unique to your setup, but i can't think of what [13:12] updated version of go as well, no difference [13:12] voidspace, check if you have cache.yaml in ~/.juju/environments/ [13:12] voidspace, and delete it if it's there before retrying [13:12] dimitern: I blew away .juju and did a fresh init [13:13] voidspace, hmm - well, then check ~/.local/share/juju ? [13:13] dimitern: I blew that away before the init too :-) [13:13] I'll check though [13:13] just reinstalling lxc to fix (hopefuly) another test failure on my machine [13:14] dimitern: no cache.yaml there [13:17] voidspace, what do you have in ~/.local/share/juju ? I only have ssh/ with juju_id_rsa + juju_id_rsa.pub [13:17] dimitern: I have an environments.yaml too [13:17] dimitern: which was created by juju init [13:18] voidspace, in ~/.local/share/juju/ ? [13:18] voidspace, I don't have this fwiw [13:19] yep [13:19] this is a fresh build with go 1.5, no preexisting juju home directory, then juju init [13:19] which created environments.yaml [13:19] it's just the boilerplate one [13:21] voidspace, juju init shouldn't be necessary anymore AIUI [13:21] dimitern: it certainly shouldn't cause tests to fail though! [13:22] I'll get rid of the environments.yaml and see [13:22] voidspace, yeah [13:22] go rid of canonistack and ec2 credentials from my environment - that didn't fix it either [13:23] removing environments.yaml didn't help [13:25] voidspace, try rm -fr $GOPATH/pkg/ followed by go install github.com/juju/juju/... ? [13:25] dimitern: heh, ok [13:27] I just updated go version which I think did the same thing [13:27] (effectively) [13:28] nope, same problem [13:28] back to digging into the code [13:31] voidspace, hmm ok, I hope you find it [14:37] hey whats going on everyone??? Can someone give an example of provisioning storage using the openstack provider? Thanks! [15:16] ericsnow: ready when you are. [15:50] Bug #1543178 opened: TestBlankJujuXDGDataHomeEnvVar fails because widows isn't looking at xdg location

[16:20] Bug #1543189 opened: testing_test.go: fakeHome declared and not used

[16:20] Bug #1543193 opened: TestConfig fails on windows because of wrong dir

[16:32] Bug #1542992 changed: ec2: destroy-controller causes rate limiting failure [16:56] voidspace, any luck with those test failures btw? [17:02] voidspace, dooferlad: sync with rick [17:03] Bug #1542990 changed: tools version checker restarts every 3 seconds [17:04] fwereade: did you see waigani's email about moving the state workers to use the dependency engine? [17:07] frobware: omw [17:07] dimitern: buried in predicate matchers in the apiserver at the moment :-) [17:08] frobware: can't join [17:08] voidspace, right, np [17:08] worked [17:12] Bug #1543216 opened: EOF uploading charm

[17:24] Bug #1543216 changed: EOF uploading charm

[17:25] cherylj, dammit, I did, it looked broadly sensible/correct but I haven't done it properly yet [17:33] Bug #1543216 opened: EOF uploading charm

[17:33] Bug #1543223 opened: kill-controller fails on missing volume

[17:39] Bug #1543223 changed: kill-controller fails on missing volume

[17:42] Bug #1543223 opened: kill-controller fails on missing volume

[17:51] unitMatchSubnet returns true for match when it gets an error! [17:51] which seems very odd [17:52] however, if there's an error - match shouldn't be checked [17:52] I think we've got a double-nil error here - the error is != nil but errors.Trace is still returning nil [17:52] something like this [17:53] no, not the problem [18:01] cherylj: looks like i finally got #4314 to work :) [18:02] dimitern: frobware: ah, it's matching on the subnet - now to work out why [18:03] Bug #1543235 opened: UniterSuite.TestUniterRelations again fails

[18:04] dimitern: frobware: it's fucking BT dns [18:04] voidspace, been there. [18:04] dimitern: frobware: BT causes net.ResolveIPAddr("started") to work... [18:04] fsking bollox [18:04] ok, at least I found it [18:04] voidspace, this is the problem I had with "testing.invalid" [18:04] now to see if I can fix it [18:05] frobware: yeah, I will have that problem too... [18:05] so matchSubnet returns true for the "started" status filter [18:05] hah! [18:11] voidspace, wow! that really sucks for BT :) [18:14] tych0: yay!! [18:15] tych0: CI should pick up your branch to run now since you've updated it. [18:17] tych0: I see unit test failures, though, now that it's gotten through the build. [18:18] tych0: http://juju-ci.vapour.ws:8080/job/github-merge-juju/6271/console [18:18] cherylj: i think that's the last build [18:18] er [18:18] the n-1th build [18:19] i fixed those and the jujubot actually merged it [18:19] ah! [18:19] ok :) [18:20] cherylj: what i don't remember is how to go to the fancy web page you sent me to [18:20] cherylj: and if jujubot merged that, are the additonal tests that are going to be run now that i need to check that fancy web page for, or not? [18:22] tych0: http://reports.vapour.ws/releases#lxd-container-type [18:22] tych0: there are also reports emailed out from CI with summaries. [18:22] ah, reports.vapour.ws :) [18:23] tych0: I can keep an eye out for yours and let you know if there are any failures seen there that are unique to your branch. [18:23] cherylj: oh, that would be swell, thanks! [18:23] np! [18:23] cherylj: is there anything else you need from me right now? (modulo any fixes, of course) [18:24] tych0: nope, not until your branch runs through CI [18:24] cherylj: ok, cool. thanks! [18:33] Bug #1542985 changed: Juju won't switch environments [18:36] ericsnow: trivial review? [18:36] http://reviews.vapour.ws/r/3772/ [18:38] voidspace, dooferlad, frobware, http://reviews.vapour.ws/r/3773/ please have a look when you can [18:40] ericsnow: should I just remove the existing show-charm-resources command (currently called "resources")? It conflicts with the card I'm working on that wants list-resources to have an alias of "resources" [18:40] natefinch: sure [18:40] natefinch: I'll be removing it in my patch as well [18:40] ericsnow: figured. [18:40] ericsnow: cool [18:50] rick_h__ or rick_h___: we have a task to change push-resource to only accept a single resource at a time, in anticipation of adding a --comment flag... in that case, does it make sense to keep the format push-resource service name=file, or should we just make it push-resource service name file (i.e. drop the equals and make the name and file separate arguments)? [19:06] Bug #1538241 opened: 2.0-alpha2 stabilization