/srv/irclogs.ubuntu.com/2016/04/04/#juju-dev.txt

davecheneygrrr, something is leaking /tmp/gui* turds01:20
=== menn0 is now known as menno-afk
axwthumper: "Generally LGTM" = shipit? I've answered your question about newline01:25
axwthanks for reviewing btw01:26
thumperaxw: yes01:31
axwthumper: ta01:31
=== menno-afk is now known as menn0
davecheney        // If any post-MVP command suite enabled the flag, keep it.02:50
davecheney        hasFeatureFlag := featureflag.Enabled(feature.PostNetCLIMVP)02:50
davecheney        s.BaseSuite.SetUpTest(c)02:50
davecheney        s.FakeJujuHomeSuite.SetUpTest(c)02:50
davecheney        if hasFeatureFlag {02:50
davecheney                s.BaseSuite.SetFeatureFlags(feature.PostNetCLIMVP)02:50
davecheney        }02:50
davecheneywut02:50
davecheneyif the feature flag is enabled, then enable the feature flag02:50
davecheneytaht feel when you find the same bug copy pasted into several test suites02:57
natefinchheh02:57
thumperheh04:26
thumpergolint needs more smarts04:26
thumperdon't use underscores in Go names; type interface_ should be interface (golint)at line 13 col 604:27
thumperalso complains about type_04:27
menn0thumper: here's an important one: http://reviews.vapour.ws/r/4414/04:53
* thumper looks04:53
menn0that was fun04:54
* menn0 hasn't been in the zone like that for a long time04:54
thumperI'm sure will will like that one04:54
menn0thumper: yes, he doesn't like these custom watchers04:55
menn0thumper: and I now pretty much agree with him04:55
thumpershipit04:57
menn0thumper: cheers04:58
=== thumper is now known as thumper-afk
thumper-afkbbl to catch europeans05:00
davecheneyfixing cleanup suite failures in the ec2 tests are making me sad07:28
davecheneyall day07:28
davecheneyno fucking idea what is broken07:29
davecheneytouch one thing, and magic values in other parts of the program dont' get overwritten07:29
davecheneyi hate patch value07:29
davecheneyit's tumor07:29
rogpeppedavecheney: sorry about that07:39
rogpeppedavecheney: it seemed like a good idea at the time07:39
davecheneyFFFFFFFFFFFFFFFFUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUU07:51
davecheneyCleanup suite uses a package singleton07:51
davecheneyso you can have multiple CleanupSuites07:51
davecheneypooped all the way through your suite construction07:51
davecheneyand they all COMPETE FOR THE SAME FUCKING SingleTON!07:51
davecheneyoh07:51
davecheneyno07:51
davecheneyi was wrong07:51
davecheneyi take that rant back07:52
davecheneyhttps://github.com/juju/testing/pull/9407:58
dooferladfrobware, voidspace: hangout time09:00
voidspacedooferlad: frobware: be there in 2 minutes!09:01
voidspacedooferlad: omw09:06
voidspacebabbageclunk: I'll ping you shortly, just fending off some emails09:32
babbageclunkok cool09:32
voidspacebabbageclunk: I've pulled in the changes from use-boot-resources onto our maas2instance branch09:34
voidspacebabbageclunk: and am doing the rename09:34
babbageclunkvoidspace: the one Tim suggested? ok09:34
voidspacebabbageclunk: yeah, I think it's clearer09:34
* babbageclunk nods09:35
voidspacebabbageclunk: so the old maasInstance becomes maas1Instance and the interface becomes maasInstance09:35
voidspacebabbageclunk: we need to update to the latest version of gomaasapi and fix verifyCredentials to not hit Machines endpoint plus use the new gomaasapi error instead of the net/http one09:36
voidspacebabbageclunk: we can do that in a new branch though09:36
babbageclunkvoidspace: shall I make a start on the verifyCredentials change?09:45
=== menn0 is now known as menn0-afk
voidspacebabbageclunk: sorry, missed your message10:26
voidspacebabbageclunk: sure you can look at errors too10:26
voidspacebabbageclunk: just grabbing coffee then I'm ready10:28
voidspacebabbageclunk: use-boot-resources landed10:30
voidspacefrobware: dooferlad: gah, updating to use the latest version of gomaasapi from thumper causes 23 test failures :-/11:40
voidspacejust venting11:40
voidspacenow bisecting to find what caused it11:41
frobwarevoidspace: 23 test failures in juju or gomaasapi?12:03
perrito666marcoceppi: hello, you had mentioned the name of a charm that implements update-status what was it?12:23
perrito666also mgz ping12:23
babbageclunkfrobware: juju - it turned out to be an errors.Trace that was added in gomaasapi.12:49
babbageclunkfrobware: it meant adding errors.Cause in a few places.12:49
perrito666anyone knows the exact version of lxd that works with master?12:54
voidspacefrobware: that was in juju13:18
voidspacefrobware: we found the cause, and fixed it13:18
voidspaceah yes13:18
voidspacebabbageclunk: grabbing coffee13:18
voidspacefrobware: dooferlad: http://reviews.vapour.ws/r/4422/14:06
marcoceppiperrito666: any of the bigdata charms14:17
perrito666yup, already found tx a lot marcoceppi14:18
mupBug # changed: 1229903, 1292157, 1323446, 1365542, 153195414:18
mupBug #972515 changed: Charm store needs search functionality <store> <juju-core:Invalid by niemeyer> <https://launchpad.net/bugs/972515>14:27
mupBug #972515 opened: Charm store needs search functionality <store> <juju-core:Invalid by niemeyer> <https://launchpad.net/bugs/972515>14:42
frobwarevoidspace, sorry, sidetracked by a customer issue.14:46
frobwaredooferlad: would you mind obliging with voidspace's review ^^14:46
dooferladfrobware: sure14:46
mupBug #972515 changed: Charm store needs search functionality <store> <juju-core:Invalid by niemeyer> <https://launchpad.net/bugs/972515>14:48
mupBug #1565826 opened: Unable to build juju2 from master <juju-core:New> <https://launchpad.net/bugs/1565826>14:48
mupBug #1565827 opened: TestTimeoutRun fails <ci> <test-failure> <juju-core:Incomplete> <juju-core feature-juju-run-action:Triaged> <https://launchpad.net/bugs/1565827>14:48
dooferladvoidspace: got a +114:52
voidspacedooferlad: thanks!14:59
katconatefinch: ericsnow: standup time15:01
alexisbmorning all15:05
alexisbif anything needs attention from me today please ping me directly15:05
alexisbcrawling through email atm15:06
katcoalexisb: wb15:08
alexisbthanks katco !15:08
mupBug #1565831 opened: unable to create authenticated client with maas 1.9 <bootstrap> <ci> <maas-provider> <juju-core:New> <juju-core maas2:Triaged> <https://launchpad.net/bugs/1565831>15:09
perrito666returning from holidays my brain is all :"oh that english thing again, sure, lemme switch everything, in the meanwhile, talk like scooby doo"15:15
katcoperrito666: you did fine ;)15:17
perrito666katco: when I just connected my brain was in "what are these people talking?" mode15:17
natefinchperrito666: lol15:22
=== ericsnow is now known as ericsnow-afk
=== ericsnow-afk is now known as ericsnow
ericsnowfwereade: thanks for the review; processing now15:47
voidspacerick_h_: is the meeting on today?16:00
voidspacefrobware: dooferlad: is the meeting with rick_h_ on today?16:01
voidspaceI added babbageclunk as a guest16:01
natefinchericsnow: charmstore meeting16:02
mupBug #1565826 changed: Unable to build juju2 from master <juju-core:Invalid> <https://launchpad.net/bugs/1565826>16:09
rick_h_voidspace: frobware babbageclunk sorry!16:10
voidspacedooferlad: frobware: rick_h_ is here!16:10
katcoericsnow: natefinch: so when we all have a moment, we should point new work, and i can give you update on projected completion of project to see if we're in trouble16:12
frobwarevoidspace, rick_h_: omw16:13
ericsnowkatco: k16:13
frobwarerick_h_: sorry was clicking as you were speaking...16:15
rick_h_frobware: all good16:16
mupBug #1565872 opened: As a juju user I would like to use docker on the local provider <adoption> <juju-core:New> <https://launchpad.net/bugs/1565872>16:18
alexisbkatco, ping16:19
katcoalexisb: pong16:19
mupBug #1565872 changed: Juju needs to support LXD profiles as a constraint <adoption> <juju-core:New> <https://launchpad.net/bugs/1565872>16:27
mupBug #1565872 opened: As a juju user I would like to use docker on the local provider <adoption> <juju-core:New> <https://launchpad.net/bugs/1565872>16:39
mupBug #1565880 opened: juju list-credentials --show-secrets does not do anything <docteam> <juju-core:New> <https://launchpad.net/bugs/1565880>16:39
natefinchkatco: done meeting with the charmstore guys, would like to get lunch if that's ok?16:40
katconatefinch: that's fine16:42
katcoericsnow: natefinch: we'll meet after16:42
ericsnowkatco: k16:42
perrito666natefinch: just as a reference, re lxd, the version required for juju to work is the one in the ppa no the one in wily16:44
natefinchperrito666: wait, I thought they reversed that16:45
perrito666natefinch: so did I but juju says otherwise16:46
natefinchperrito666: whatever, my lxd works, that's all I care about for now16:46
perrito666natefinch: heh, mine too16:47
natefinchperrito666: you're right, mines from the PPA16:47
* natefinch lunches16:47
=== natefinch is now known as natefinch-lunch
voidspacefrobware: dooferlad: review if you get a chance http://reviews.vapour.ws/r/4425/16:58
=== natefinch-lunch is now known as natefinch
natefinchkatco, ericsnow: ready when you guys are17:50
ericsnownatefinch: k17:51
katconatefinch: ericsnow: let's do it!17:52
marcoceppican I remove storage from a service?18:17
marcoceppiwhy do we have a detaching hook, but no way to detach storage?18:19
marcoceppirick_h_: ^18:19
katcourulama: rick_h_: hey, can you be authorized to a specific channel?18:39
urulamayes18:39
urulamaacls are different for each channel18:39
natefinchurulama: so a macaroon is channel-specific?18:39
mupBug #1503029 changed: juju plugins which exit > 0 report a subprocess ERROR <charmers> <juju-core:Fix Released by cmars> <https://launchpad.net/bugs/1503029>18:40
rick_h_katco: yes, the idea being you give your tester folks access to dev channel18:40
katcorick_h_: makes sense... suspicion confirmed ty :)18:40
urulamanatefinch: macaroon is not, but ACL is18:41
mupBug #1565943 opened: Can't bootstrap on VSphere <vsphere> <juju-core:Triaged> <https://launchpad.net/bugs/1565943>18:46
natefinchurulama: so that's kinda weird... there's no channel parameter sent up when you get a charm archive.. I guess if that charm is in a public channel it'll just work? and if it's not it'll return an access denied error of some sort?18:59
urulamanatefinch: sure it does. check out https://api.jujucharms.com/charmstore/v5/~jorge/bundle/wiki-simple/archive?channel=unpublished vs https://api.jujucharms.com/charmstore/v5/~jorge/bundle/wiki-simple/archive19:09
natefinchurulama: ahh, ok, I was looking at the docs, where it doesn't seem to indicate channels get passed to get the archives19:11
natefinchurulama: https://github.com/juju/charmstore/blob/v5-unstable/docs/API.md#get-idarchive19:12
urulamanatefinch: ha, ok, that needs some updating then19:12
natefinchurulama: oh, I guess the blurb about channels at the top covers that, for any endpoint that takes an id can take a channel19:12
urulamanatefinch: works also on /meta endpoint19:13
natefinchurulama: cool, thanks for clarifying19:13
urulamanp19:13
mbruzekcherylj: Have you seen issues with deploying LXD issues with deploying xenial with the juju agent not finishing installation?19:57
katcocan anyone tell me why we aren't using logging here? https://github.com/juju/juju/blob/master/provider/common/bootstrap.go#L13319:57
mbruzekping anyone else working with LXD and xenial19:59
natefinchkatco: weird19:59
cheryljmbruzek: I haven't tried it in a couple days20:00
cheryljmbruzek: are you using lxd provider on a daily xenial image?20:00
mbruzekcherylj: yes.20:00
mbruzekMarco and I looked at the logs, we saw a modprobe fail.20:01
cheryljmbruzek: k, let me provision a new xenial instance20:01
natefinchkatco: that's the output for the bootstrap command, to the CLI20:01
katconatefinch: was just typing that20:01
katconatefinch: got lost in a lot of --debug output so incorrectly seemed odd looking at backtraces20:01
natefinchkatco: it would still make sense to wrap it in a logger so you don't risk interleaved writes, but it's less likely on the client, I guess20:01
mbruzekcherylj: I assert the agent will be stuck "Waiting for agent initialization to finish"20:03
mbruzekafter deploying a charm in xenial series20:03
mbruzekOr at least that is what I am seeing20:04
* thumper dives into email20:04
cheryljmbruzek: so it's also a xenial container you're deploying too?20:05
cheryljmbruzek: is it a trusty controller?20:05
mbruzekI was following this page: https://jujucharms.com/docs/devel/config-LXD20:05
mbruzekThat bootstrap command looks like I am requesting a xenial controller.20:06
mbruzekcherylj: Once you get bootstrapped, deploy a xenial charm.20:13
mbruzekjuju deploy ubuntu --series xenial --force20:13
mbruzekcherylj: ^20:13
cheryljmbruzek: are you using beta3, or tip of master?20:14
mbruzekbeta320:14
cheryljomg, I'm deploying an openstack bundle on my vmaas and my laptop is screaming!20:14
cheryljmuch lag, such slow20:15
mbruzekcherylj: Oh20:15
mbruzekcherylj: Yikes!20:15
mbruzekcherylj: I just deployed the ubuntu charm on both trusty and xenial20:16
cheryljk, my instance in aws is coming up :)20:16
mbruzekLXD inside aws?20:16
cheryljmbruzek: yeah, I like to spin up a clean xenial install when testing the lxd provider20:17
mbruzekcherylj: so I got both of the charms to deploy completely but I see the error in the machine log that will never work on lxd20:19
mbruzekThe juju agent checks for kvm:  WARNING juju.cmd.jujud machine.go:760 determining kvm support: INFO: /dev/kvm does not exist20:20
mbruzekNot there in the base cloud image20:20
mbruzekHINT:   sudo modprobe kvm_intel20:20
mbruzekcherylj: Then I see an error when the juju agent tries to do modprobe.20:21
mbruzekmodprobe: ERROR: ../libkmod/libkmod.c:556 kmod_search_moddep() could not open moddep file '/lib/modules/4.4.0-15-generic/modules.dep.bin'20:21
mbruzek: exit status 120:21
mbruzekno kvm containers possible20:21
cheryljmbruzek: and that's causing the machine agent to barf?20:21
cheryljmbruzek: that should be fine20:21
cheryljbootstrapping now...20:22
mbruzekcherylj: This time it does not appear that they barfed, both trusty and xenial seem OK now.20:22
cheryljmbruzek: did you do anything differently?  or did just doing it a second time work?20:23
mbruzekcherylj: why do you think that is OK? doing a modprobe within a LXD container is never going to work.20:23
cheryljmbruzek: it's just a check to see if we have the capability of hosting KVMs20:23
cheryljhappens on all providers20:24
mbruzekcherylj: but since LXD is sharing the kernel, a modprobe will not work in KVM. So the error is handled approprately?20:26
mbruzekcherylj: s/KVM/LXD20:26
mbruzekBecause they share the kernel20:26
mbruzekyou can not insert any more mods20:26
mbruzekcherylj: OK I was able to bootstrap, deploy and clean up correctly this time.20:30
cheryljmbruzek: is there a better way to detect whether or not we can host KVMs?  (I'm not familiar with how we would do that)20:32
cheryljmbruzek: and on this lxd instance issue - when the unit was waiting for agent initialization, did it have an instance ID yet?20:33
mbruzekcherylj: Not that I am aware of, but perhaps we should check for LXD / containers first and if not in a container then check for kvm.20:33
mbruzekcherylj: let me check my status20:33
mbruzekcherylj: yes, I was even able to ssh to it, but the juju agent was dead20:34
cheryljah, I'm running into a different issue then20:34
cheryljmbruzek: were both unit and machine agents dead?20:34
mbruzekcherylj: I think it was the machine unit, and I was not able to destroy the controller20:34
cheryljmbruzek: when you did destroy-controller, what was the error?  or did it just hang?20:35
mbruzekcherylj: it went into an infinite loop20:35
cheryljmbruzek: ah, in the "waiting for x service" blurb?20:36
mbruzekI hit control-c and re ran it with --debug on20:36
mbruzekWaiting on 2 models, 1 machine, 2 services20:36
mbruzekcherylj: with debug:20:37
mbruzek2016-04-04 19:32:20 INFO cmd cmd.go:129 Waiting on 2 models, 1 machine, 2 services20:37
mbruzek2016-04-04 19:32:20 INFO cmd cmd.go:141 admin@local/default (alive), 1 service20:37
mbruzekfor ever, I could not --force it20:37
cheryljmbruzek: this sounds familiar.  I think I've run into this before, but couldn't recreate20:37
mbruzekcherylj: well kill the machine agent, and try to destroy the controller20:38
cheryljmbruzek: you can manually terminate the controller machine, then run kill-controller20:38
mbruzekBut this time everything came down OK20:38
cheryljok20:39
cheryljlet me poke more, see if I can recreate20:39
mbruzekcherylj: Yes I figured out the lxc commands to stop and delete the images20:39
mbruzekBut all the Juju stuff was left in the .local/share/juju20:40
cheryljmbruzek: yeah, if you just kill the controller machine and run kill-controller, it will give up talking to the controller and then go to the provider to clean up20:40
cheryljand then it will clean up your local cache information20:40
mbruzekcherylj: OK I was not able to reproduce that, but I didn't know that worked20:41
cheryljmbruzek: I wrote a lot of the destroy / kill controller code so I know ALL ITS SECRETS20:41
mbruzekcherylj: OK well thank you for the information20:41
cheryljmbruzek: there's also a bug open to provide a way to clean up stale controller information20:42
cheryljIt'd be super awesome if we could get that done for 2.020:42
mbruzekcherylj: OK so you were able to get a xenial ubuntu and everything looks good?20:43
cheryljmbruzek: no, I'm running into a different bug20:43
cheryljmbruzek: what I see is that on a new xenial install, deploying in lxd doesn't even get an instance associated with the service20:43
mbruzekcherylj: so let me understand, you are using amazon to run a local LXD provider?20:43
cheryljmbruzek: yes, I manually deploy a xenial machine, install juju2 on it, and bootstrap lxd20:44
mbruzekcherylj: OK, I just wanted to understand the workflow20:44
cheryljmbruzek: :)  I do it to make sure it's clean and I haven't messed things up some how20:44
mbruzekthat is a good idea!20:45
cheryljmbruzek: if you haven't seen it before, I use the cloud image finder to quickly launch a particular instance in a particular region:  https://cloud-images.ubuntu.com/locator/daily/20:45
mupBug #1564622 changed: Suggest juju1 upon first use of juju2 if there is an existing JUJU_HOME dir <juju-release-support> <juju-core:Triaged> <https://launchpad.net/bugs/1564622>20:46
mupBug #1565991 opened: juju commands don't detect a fresh juju 1.X user and helpfully tell them where to find juju 1.X <juju-core:Triaged> <https://launchpad.net/bugs/1565991>20:46
mbruzekcherylj: Thanks for the link20:46
mbruzekcherylj: which type do you pick, I imagine the ssd ones are more expensive20:47
cheryljmbruzek: I just go with the ssd ones20:47
voidspacethumper: ping20:53
thumpervoidspace: hey20:53
voidspacethumper: morning20:53
voidspacethumper: I just wanted to say thanks20:54
thumperfor?20:54
voidspacethumper: your changes to gomaasapi broke 24 tests and it took us nearly two hours to work out why20:54
voidspace;-)20:54
voidspacethumper: we fixed it, but I do wonder if we ought to change it (which is why I'm really pinging)20:54
thumperwont make that mistake again then will you?20:54
voidspacehehe20:54
voidspacethumper: you added an errors.Trace to the error the client returns20:54
thumpergomaasapi.GetServerError(err error) (ServerError, bool)20:55
voidspacethumper: which means that anything that attempts to cast to a ServerError fails20:55
voidspacesure20:55
thumperI added that for that exact reason20:55
voidspacebut it's backwards incompatible20:55
thumpereh...20:55
thumperyeah20:55
thumperish20:55
thumpera bit20:55
thumpernot heaps20:55
voidspacewell, it broke 23 tests on master...20:55
voidspaceso yeah - ish20:55
thumperyou're welcome20:55
voidspace:-D20:55
voidspaceit's *better*20:55
voidspacemaybe we just see if anyone else complains20:56
voidspacegomaasapi does have users20:56
voidspacenot many probably but some :-)20:56
voidspace(I mean the change is better)20:56
voidspaceanyway, I'm off to bed20:57
voidspacewell - watch TV anyway20:57
voidspacethumper: just thought as you were around I'd hassle you :-)20:57
voidspacethumper: have a good day20:57
thumper:)20:57
mupBug # opened: 1564157, 1564163, 1564165, 1566011, 156601421:34
mupBug # changed: 1564157, 1564163, 1564165, 1566011, 156601421:37
mupBug #1565880 changed: juju list-credentials --show-secrets does not do anything <docteam> <juju-release-support> <juju-core:Triaged> <https://launchpad.net/bugs/1565880>21:46
mupBug # opened: 1564157, 1564163, 1564165, 1564168, 1564670, 1566011, 1566014, 1566023, 156602421:46
menn0-afkthumper: PR #20 LGTM. It's a bit hard to read the #21 b/c it builds on it so let me know when you've merged #20.22:19
=== menn0-afk is now known as menn0
thumpermenn0: ack, will land, and update pr22:20
mupBug #1566044 opened: list-models and show-model should represent similar keys for controller model <juju-core:New> <https://launchpad.net/bugs/1566044>22:37
mupBug #1566044 changed: list-models and show-model should represent similar keys for controller model <juju-core:New> <https://launchpad.net/bugs/1566044>22:49
=== redir is now known as redir_afk
mupBug #1566044 opened: list-models and show-model should represent similar keys for controller model <juju-core:New> <https://launchpad.net/bugs/1566044>22:52
perrito666axw: anastasiamac is standup still happening?23:15
axwperrito666: yes, omw23:15
anastasiamacm here23:15
davecheney# github.com/juju/juju/provider/ec2_test23:38
davecheneylocal_test.go:93: ambiguous selector t.AddCleanup23:38
davecheneyFATAL: command "test" failed: exit status 223:38
davecheneypungent smell that CleanupSuite is present more than once in this suite23:39
perrito666davecheney: odd, although I found someone had refactored something upstream last time I was nearby23:44
=== redir_afk is now known as redir

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!