nurfethey guys, which branch I should create PR to for a new provider, develop?00:12
thumperanastasiamac: https://github.com/juju/juju/pull/680400:17
thumpernurfet: yes00:17
* anastasiamac looking00:20
anastasiamacthumper: +1...00:22
rick_hthumper: ping01:14
thumperrick_h: hey01:14
rick_hthumper: hey, did you want to catch up today at all?01:14
rick_hthumper: what's your schedule look like?01:15
rick_hthumper: k, https://hangouts.google.com/hangouts/_/canonical.com/rick?authuser=101:15
thumperoh ffs02:48
thumperfuckity fuck...02:56
* thumper considers this race02:56
thumperbabbageclunk: around?02:56
thumpergot this will be horrible, but necessary...02:59
anastasiamacthumper: sounds like fun \o/ what have u got?03:02
thumpera race condition03:02
thumperis used throughout the tests03:02
thumperto get watchers to get values03:02
anastasiamaci've seen these and added a few of my own03:03
anastasiamacsome tests do not run well without them03:03
thumperbut places assume that it is sufficient03:03
thumperin the 390 tests, there are calls to change, startsync, more change, start sync03:03
anastasiamaci think the bigger isse (and probably better approach) is to address our watchers03:03
thumpernow the assumption is that first change comes independently of the second03:03
thumperbut you can't ensure that03:04
thumperbecause the start sync just says "hey, please start a sync", it doesn't wait for it, or even be sure it has started03:04
thumperI'm beginning to think that the test is just bollocks03:04
thumperthis failure http://reports.vapour.ws/releases/4711/job/run-unit-tests-xenial-s390x/attempt/903#highlight03:05
anastasiamaci'd say we should seriously consider all intermittently failing tests and what values they bring03:05
anastasiamac,ost of them need to b re-designed03:05
thumperfunc (s *watcherSuite) TestWatchUnitsKeepsEvents(c *gc.C) {03:05
anastasiamacor actually made into unit test03:05
anastasiamacyeah, that's one... but there are a lot of these sprinkled thru03:06
anastasiamacjust look for StartSync in test files03:06
thumperoh I know exactly what is happening03:07
thumperbut it's all fake03:07
thumperand meaningless03:07
thumperthe test is asserting a false proposition03:07
* thumper fixes test03:16
thumperit is bollocks03:16
thumperthe assertion that the events are separate is needless03:17
thumperand impossible to ensure without hoop jumping03:17
thumperso why bother03:18
thumperanastasiamac: https://github.com/juju/juju/pull/680603:24
* anastasiamac looking03:24
thumperanastasiamac: the strings watcher test helper calls start sync before any assert03:31
thumperthe extra ones aren't necessary03:31
anastasiamacthumper: there was at least another test where it was necessary... let me dig it up just for ur perusal :D03:32
anastasiamacthumper: https://github.com/juju/juju/pull/660803:33
anastasiamacin that instance, adding additional sync eliminated the race :(03:33
anastasiamacsyncs even03:33
thumperanastasiamac: yeah, using watchertest.StringsWatcherC would have saved you all that03:34
thumperrather that statetest03:34
thumpersee watchertest/strings.go03:35
thumperwatchertest.NewStringsWatcherC(c, w, s.BackingState.StartSync)03:35
thumperpass a "pre-assert" function in03:36
thumperin this case, the StartSync on the BackingState03:36
thumperthat way you don't have to sprinkle the code with startSync calls before every assert03:36
thumperbecause it does it for you03:36
anastasiamacthumper: k. i'll circle back to it tomorrow... have a killer migraine atm :(03:39
gsamfira_Hello folks. Anyone have time to review a 5 line PR? :) https://github.com/juju/utils/pull/26010:54
=== gsamfira_ is now known as gsamfira
perrito666gsamfira: hi dude, long time no see11:20
gsamfiraperrito666: yeah, it's been a while :D11:20
perrito666looking at the patch11:20
perrito666gsamfira: I am keen to approve the patch even though you missed the QA steps :p but I would love an explanation of what are those two windowses?11:21
gsamfiraIt's been a while, I am not familiar with the new QA steps :D. I may need a crash course :P11:24
gsamfiraOne is Hyper-V server 2016. The free version of windows that just gives you the hypervisor and nothing more11:24
gsamfirathe other is the Windows storage server. That one does not really add a new series. Just enables detection for that particular version11:25
perrito666gsamfira: sure, we sent the new rules over a mail or some other not really good communication form so I cant easily point you to them (I know, we need to fix that)11:25
perrito666gsamfira: so the first step is to re-propose to develop instead of master :D11:25
gsamfiraI didn't see a develop branch in juju/utils11:26
gsamfirawould be happy to do it11:26
perrito666ahhh utils11:26
perrito666I forgot we moved that11:26
* perrito666 makes a note about having develop o utils11:27
perrito666gsamfira: ship it11:27
=== junaidal1 is now known as junaidali
junaidaliHi guys, I have a failed machine (a controller instance in HA), I have run 'juju enable-ha' to ensure HA again but now I can't remove the failed machine13:06
junaidaliit is giving error 'ERROR no machines were destroyed: machine <machine number> is required by the model'13:06
junaidaliAny idea, what I'm missing here? Juju HA is already ensured.13:06
perrito666junaidali: hey, I assume this is juj 2.x?13:12
junaidaliperrito666: yes13:14
perrito666mm, could you post in a pastebin your "juju status -m controller --format=yaml" ?13:17
junaidaliperrito666: my bad, I removed two instances from my HA (out of three, need to restore from backup now), can't access controller right now13:23
perrito666junaidali: ah I see13:24
perrito666well ping me if you need ay help13:24
perrito666ill be on this channel all day13:24
junaidaliperrito666: Thanks, did you want to check the member-status for that machine?13:24
perrito666junaidali: yes, also I was curious what was that juju thought the statuses for those machies where13:25
junaidaliI remember, the value was 'no-vote'. I will let you know when I restore the controller and re-enable HA13:26
perrito666tx a  lot13:27
thumperI hate intermittent failures21:19
babbageclunkthumper: yeah they suck21:20
thumperhmm actually I think this is another case of slow race run with complex cert sometimes takes too long21:22
thumperbabbageclunk: https://github.com/juju/juju/pull/680921:35
babbageclunkthumper: looking21:37
thumperbabbageclunk: many of these timeout type failures are due to expectations that the server starts quickly21:37
thumperon the race build, on the CI machine, generating a CA cert can take 4s and I have seen the 2048bit server cert take 10s21:37
babbageclunkthumper: LGTM21:40
redirI sometimes hate intermittent failures21:50
redirMore often I hate intermittent success21:51
redirI am a bit stuck running into a bug https://bugzilla.redhat.com/show_bug.cgi?id=132508521:54
redirbut dannf is off since it is a holiday here today.21:55
redirso I can't ask if he got past that via a workaround.21:55
redirSo now might be a good time to have a run and a think on it.21:56
redirthe issue is theoretically fixed in libvirt 1.3.3, but xenial ships with 1.3.121:56
rediralso it seems that we can't expect kvm to work on arm64 on trusty21:57
thumperredir: if that is the case, then we should blacklist it in the code22:02
babbageclunkthumper: https://github.com/juju/juju/pull/6810 - bit of a weird one.22:07
* thumper looks22:07
babbageclunkthumper: Ok, I'll add the names. Do think it would be better as MoveInstancesToController(controllerUUID string, ids ...instance.Id)?22:10
babbageclunkthumper: Now that I've typed that out I don't think it's better.22:11
thumperbabbageclunk: I'd keep consistency with the other methods22:11
perrito666morning babbageclunk thumper ... redir? arent you in US?22:13
redirperrito666: yes, but planning to swap the holiday for later22:13
* redir really goes for a run now22:14
babbageclunksure buddy22:14
redirthumper: you mean bl arm64 or trusty on arm64?23:02
thumperredir: trusty on arm6423:02
redirlet me run it by dannf tomorrow23:03
anastasiamacthumper: babbageclunk: veebers: wallyworld: redir: perrito666: axw: i've added an HO to standup meeting invite... could we plz try it today instead of ...?23:07
redirthey both work for me23:07
anastasiamacredir: i'd take it as a 'yes' ;)23:09
babbageclunkanastasiamac: fine by me23:09
anastasiamacbabbageclunk: \o/23:10
babbageclunkthumper: could you take a look at https://github.com/juju/juju/pull/6813?23:14
thumperbabbageclunk: ack23:33
thumperbabbageclunk: were we tagging instances with controller uuid in gce before?23:34
* thumper needs to go kick everyone in the house off the internet for the team hangout23:37
babbageclunkthumper: yup - look in ControllerInstances for example.23:40
Miguel_UbuntuTring to install Juju on a MAAS server. The bootstrap fails. There is a similar report on Ubuntu Solutions Engineering o9n github, but I cannot use the sugestion. It is the top issue.23:47
thumperwallyworld, anastasiamac: standup hangout?23:47
thumperperrito666: hangout change?23:48
wallyworldwhat's the hangout name?23:49
perrito666thumper: sure, fire up a hangout23:49
anastasiamacHO is in standup invite incalendar23:49

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