| veebers | right, mgz babbageclunk sorry for the delay, that test has run through. Should I put the 2 controller logs somewhere or is there something obvious I can check for? | 00:57 |
|---|---|---|
| babbageclunk | hey veebers - yeah, could you look to see whether the migrationmaster says "successful, removing model from source controller" twice for some reason? | 00:59 |
| veebers | babbageclunk: is that the exact text to search? | 01:00 |
| babbageclunk | veebers: I'm chasing a big about that at the moment. | 01:00 |
| babbageclunk | veebers: yup | 01:00 |
| veebers | ack | 01:00 |
| babbageclunk | *bug | 01:00 |
| babbageclunk | veebers: are you running your test against develop? | 01:01 |
| babbageclunk | natefinch: ping? | 01:02 |
| veebers | babbageclunk: hah, i mess that up I don't have the logs for the destruction period. I don't see that string at all in either controller logs (machine-0.log) | 01:02 |
| veebers | babbageclunk: I'm running it against the latest 2.1 build | 01:02 |
| babbageclunk | veebers: so maybe the migration isn't successful? | 01:03 |
| babbageclunk | veebers: ah, ok - I think I've tracked down this bug to something that landed in develop about 4 hours ago, so it won't be related | 01:03 |
| babbageclunk | veebers: and what are you doing in the test - deploying a charm, migrating, migrating back? | 01:04 |
| veebers | babbageclunk: I see "setting migration phase to SUCCESS" once in the logs, but that'll be for the first migration not the migration back | 01:04 |
| veebers | babbageclunk: exactly that | 01:04 |
| veebers | babbageclunk: also after the migration adding a unit to that model and ensuring it's working | 01:05 |
| babbageclunk | oh, nice | 01:05 |
| babbageclunk | veebers: hmm. I think maybe I should fix this one and then just try to reproduce your one on 2.1. | 01:07 |
| veebers | babbageclunk: ok, it's easy enough to reproduce | 01:07 |
| natefinch | babbageclunk: what did I break? | 01:38 |
| natefinch | veebers: ^ | 01:41 |
| veebers | natefinch: I don't know if babbageclunks ping has something to to with the bug I was talking to him about sorry | 01:42 |
| natefinch | veebers: quite possible... I landed some stuff today that is triggered during migrations | 01:43 |
| veebers | natefinch: the issue I'm seeing was happening yesterday (and before) too. Perhaps it's to do with the other bug he was working on | 01:45 |
| natefinch | ahh hmm, then yeah, probably not my thing | 01:45 |
| veebers | natefinch: we'll have to wait with bated breath to find out :-) | 01:45 |
| veebers | babbageclunk: FYI I have a run of this test where I have the environment still in tact (i.e the controllers weren't killed) so we can poke around that if it's of interest | 02:05 |
| babbageclunk | natefinch: still around? It looks like your migration status change fails at the end of the migration - it tries to set the status for the model after it's been removed | 02:22 |
| babbageclunk | (sorry, was on a call) | 02:22 |
| natefinch | babbageclunk: hmm... crud | 02:22 |
| babbageclunk | natefinch: https://bugs.launchpad.net/juju/+bug/1650425 | 02:23 |
| mup | Bug #1650425: migration: migrating back gives "target prechecks failed: model is being migrated out of target controller" <juju:New> <https://launchpad.net/bugs/1650425> | 02:23 |
| natefinch | ahh, you know, I saw that error and brought it up at the sprint, but someone said it was a different bug that would get fixed later. guess it was the same effect but different cause | 02:25 |
| babbageclunk | natefinch: I've made a change to the unit test for migration.SetPhase so it fails - if I send that to you can you look at this tomorrow? | 02:25 |
| natefinch | babbageclunk: yeah, thanks | 02:26 |
| babbageclunk | natefinch: then I can chase some other migration bugs :( | 02:26 |
| babbageclunk | natefinch: awesome, thanks! | 02:26 |
| babbageclunk | (like veebers's one) | 02:26 |
| natefinch | babbageclunk: thanks for writing that test | 02:27 |
| babbageclunk | natefinch: it's a tweak to the existing one to make it match reality more (by removing the model before setting the phase to DONE). | 02:28 |
| babbageclunk | natefinch: I'll send you the patch | 02:28 |
| natefinch | awesome | 02:28 |
| === frankban|afk is now known as frankban | ||
| === tinwood is now known as tinwood_holiday | ||
| Andrew_jedi | jamespage: Hello, I was wondering that Juju creates an "Admin" role but in policy.json, mapping is done for "admin" role. How does that work here? | 10:39 |
| Andrew_jedi | jamespage: Hello, I was wondering which service user owns the "users and projects" created by heat ? juju or heat-cfn_heat ? | 12:45 |
| natefinch | perrito666: https://pbs.twimg.com/media/CzzeFsSXUAA815I.jpg:large | 14:25 |
| natefinch | looks doable | 14:25 |
| perrito666 | natefinch: I dont see signs of component burn (the rightmost green one is the one you want to change) there should be some if youheard a buzzing noise | 14:27 |
| perrito666 | perhaps in the other side of the board? | 14:27 |
| perrito666 | the fuse might be at fault but that kind of fuse usually burns with a dry "click" noise and also, if the fuse blows it means something else is failing | 14:29 |
| natefinch | yeah, I'll look on the other side. It was definitely a hiss, like air rapidly leaving a small hole, not a click | 14:32 |
| perrito666 | capacitors could do that if leaking fluid, most other components make a sound like a match burning | 14:33 |
| perrito666 | also, next time use a better cam or more light | 14:33 |
| natefinch | heh | 14:34 |
| natefinch | perrito666: http://imgur.com/a/4rmIU | 14:45 |
| deanman | Any idea why juju 2.0.2 won't let me deploy my own charm and i get this error ? http://paste.ubuntu.com/23638330/. I can deploy other charms just fine | 14:52 |
| natefinch | deanman: seems like a proxy problem... we try to HTTP POST the local charm to your controller, but there's a proxy getting mad at you for some reason | 14:55 |
| deanman | natefinch, Yes im behind proxy and there is a bug reported and found a workaround but it seems like this is something new in 2.0.2. Does this look ok to you ? http://paste.ubuntu.com/23638370/ | 15:00 |
| deanman | ok found it, juju prefers NO_PROXY over no_proxy and in my case i had script setting NO_PROXY which i wasn't monitoring. | 15:26 |
| natefinch | deanman: glad you figured it out. Sorry I had to bug out for a while | 16:46 |
| === frankban is now known as frankban|afk | ||
| perrito666 | ill be relocating bbl | 18:02 |
| redir | maas instructions say to choose "multiple server install with maas" however I don't see that when I boot from the xenial server iso. Did something change? | 18:17 |
| deanman | natefinch, no worries, still learning my way so more than often im usually to blame | 18:27 |
| perrito666 | redir: mm, what options do you have? | 18:30 |
| redir | perrito666: region or rack controller | 18:33 |
| perrito666 | redir: rack controller | 18:35 |
| redir | perrito666: tx | 18:35 |
| perrito666 | hold | 18:35 |
| perrito666 | I got it wrong | 18:35 |
| perrito666 | mm, I did this in a much different way last time | 18:35 |
| perrito666 | I used the maas package | 18:36 |
| perrito666 | redir: here is a detail https://maas.ubuntu.com/docs/install.html | 18:36 |
| redir | perrito666: I saw that which also says to choose multiple server install.... or do it from the cli | 18:37 |
| redir | which maybe I should just do the latter | 18:37 |
| perrito666 | redir: I am not sure how to do this using the actual install disk | 18:38 |
| perrito666 | redir: have you tried #maas? | 18:38 |
| redir | perrito666: nope, me goes there | 18:38 |
| redir | asked in maas as well. | 18:40 |
| katco | alexisb: happy birthday :) | 19:09 |
| natefinch | sinzui: getting a weird represenative test failure... trusty unit tests pass, but the trusty job still fails... http://juju-ci.vapour.ws/job/github-check-merge-juju/462/artifact/artifacts/trusty-err.log | 19:10 |
| perrito666 | katco: I believe alexisb is k.o with a flu | 19:11 |
| perrito666 | alexisb: hb in any case :) | 19:11 |
| katco | perrito666: doh. trying to fight off a respiratory infection myself. travel is hard on the body | 19:11 |
| perrito666 | katco: It only affects my stomach, which is not that bad because I take gastritis meds anyway :) | 19:12 |
| * sinzui looks | 19:12 | |
| * perrito666 wishes CI tests would print a summary at the ened | 19:12 | |
| perrito666 | end | 19:12 |
| katco | perrito666: since having a kiddo, i have discovered that my lungs are apparently my weakness | 19:13 |
| katco | perrito666: never got sick < kid < respiratory infections | 19:13 |
| perrito666 | lol | 19:13 |
| perrito666 | katco: happens a lot | 19:13 |
| perrito666 | I am son of a doctor so I got an epecially good inmune system just because of how many things my mom broght back form the hospital :p | 19:14 |
| sinzui | natefinch: The tests are *incomplete* the 255 exit code implies ssh was dropped during the test. The test s got as far as github.com/juju/juju/feature | 19:14 |
| sinzui | natefinch: do you want me to requeue? | 19:15 |
| natefinch | sinzui: oh weird, ok, yeah, I missed that they didn't actually finish (or at least, we couldnt' tell they finished) | 19:15 |
| natefinch | sinzui: yes please requeue | 19:15 |
| sinzui | I sussed macgreagoir's bug. Juju's default contstraints for privisioning a non-controller machine are invalid for the new regions. If I use --constraints mem=4G as I think the controller contrains do I get a machine | 19:16 |
| natefinch | katco: I think eventually your immune system adjusts... I don't get sick as often any more, but it took a few years. | 19:17 |
| sinzui | natefinch: No shame in missing that. It tooks me a few weaks of looking at those abominalble ppc63el logs to learn they often dropped connection | 19:17 |
| katco | natefinch: i am definitely getting sick less. but seems as though travel still gets me with lack of sleep/dry air | 19:17 |
| mgz | sinzui: hm, invalid in what what? | 19:19 |
| mgz | *way | 19:19 |
| mgz | there are no possible instance types matching the constraint? or they match an instance type that doesn't really exist in the region? | 19:19 |
| sinzui | mgz --constraints mem=4G cs:trusty/ubuntu tu works. the default constrains for a hosted machine are invalif | 19:19 |
| rick_h | katco: yea, my son's second year is the 'year we don't talk about' | 19:21 |
| rick_h | katco: for 11 months at any given time at least one of us was sick | 19:21 |
| katco | rick_h: it is really the only thing about having a kid that took me by surprise. the rest i expected to be surprised by | 19:22 |
| rick_h | lol | 19:22 |
| natefinch | review anyone? Pretty simple: https://github.com/juju/juju/pull/6730 | 19:45 |
| natefinch | katco: if you have time, this adds ability to have defaults on open ended queries from Pollster | 19:45 |
| katco | natefinch: sure will tal | 19:45 |
| mgz | hm, so it would be nicer to not prompt, but if I'm reading this right the prompt does include the url that will be set if nothing is entered? | 19:47 |
| mgz | ah, it includes "use cloud api url" | 19:48 |
| mgz | which is fine | 19:48 |
| deanman | Is it possible to provide something like a local `simplestreams` service so you could deploy a centos7 lxd image served from your local lxd daemon ? | 20:02 |
| natefinch | deanman: you can definitely do custom simplestreams urls | 20:17 |
| natefinch | but obviously you can serve them from your laptop to an AWS machine (unless you have some wacky networking set up). | 20:17 |
| natefinch | s/can/can't | 20:17 |
| deanman | natefinch, What's the status with centos7 support? Any idea when will there be support other than private clouds? | 20:18 |
| natefinch | deanman: I'm not sure of the official status... but it's one of those things were you have to get 3 different very large companies to agree on something | 20:19 |
| natefinch | FWIW, custom image streams is bootstrap -config image-metadata-url=http://xxx.xxx.xxx.xxx/simplestream/images/ | 20:20 |
| deanman | will that work for LXD images? | 20:23 |
| natefinch | deanman: lxd provider, or lxd containers on another provider? | 20:27 |
| deanman | hmm that is a good question | 20:27 |
| deanman | dare to say any? Simply want to see that i can re-purpose my workload from centos7 to jujufied centos7 charm. | 20:30 |
| deanman | natefinch, open to any suggestions. At the end probably publish the workload as an lxd image, and then write a charm that pulls that on top of a xenial charm ?? | 20:34 |
| natefinch | deanman: that sounds like it's outside the scope of what juju provides. The charm can, of course, tell lxd to download images from wherever | 21:02 |
| sinzui | deanman: There is some monkey business involving lxd images | 21:02 |
| natefinch | deanman: but then it'll be opaque to juju | 21:03 |
| sinzui | deanman: lxd images always come from cloud-images.ubuntu.com and ubuntu names them in such a way to use the alias as a cache | 21:03 |
| sinzui | deanman: I have created my own lxd images, then gave them an alias like ubuntu-xenial to convice juju to use it | 21:04 |
| rick_h | deanman: it's something we were looking at as part of some work to cache lxd images and such, but it's not implemented at the moment so you have to go hacky like sinzui mentions | 21:04 |
| sinzui | deanman: I don't know if juju+lxd can be tricked to serve a centos image, bug the alias might be ubuntu-centos7 | 21:05 |
| sinzui | deanman: I can point you to testing images for aws that has centos7. I would love to get some feed back about their usefulness | 21:06 |
| sinzui | deanman: I mean testing stream with centos7 images registered for aws | 21:06 |
| deanman | natefinch, agreed, but need to demonstrate the power of juju, working on local and then just deploy on a cloud for staging/production. | 21:07 |
| deanman | sinzui, Im happy to assist, give feedback with any workaround till official fix lands on 2.1.0 (at least from what i see on launchpad) | 21:08 |
| sinzui | deanman: bootstrap with image-metadata-url: https://s3.amazonaws.com/temp-streams/aws-image-streams/ | 21:09 |
| sinzui | deanman: or use model-config to set it | 21:10 |
| deanman | sinzui, i was looking into that https://streams.canonical.com/juju/images/releases/streams/v1/index.json | 21:10 |
| sinzui | deanman: That is the official local that augement ubuntu streams with windows and centos image data. the testing stream successes are promoted to that location. I believe the problem with centos images streams is that they break juju 1.25 | 21:11 |
| deanman | sinzui, should i select a specific aws region ? | 21:22 |
| natefinch | katco: how much do you care about those test names? My thought was that Test<FunctionName> is a reasonable convention to indicate "here's a normal happy-path style test for FunctionName" | 21:22 |
| katco | natefinch: well, the problem is "which happy path"? | 21:23 |
| katco | natefinch: i approved the PR, but i would prefer to see folks begin encoding some expectation into the name of the test | 21:23 |
| natefinch | katco: I know, but at the same time, I can't really see calling it TestEnterVerifyWhereVerifyPassesOnTheFirstTry | 21:23 |
| katco | natefinch: gentle probing into why not? | 21:24 |
| sinzui | deanman: oh, you might need to. us-east-1 defintely works | 21:24 |
| deanman | sinzui, ok yes, eu-central allready failed, retrying with us-east1 | 21:24 |
| sinzui | deanman: I see. Indeed the regions are only a subset. I see eight regions https://s3.amazonaws.com/temp-streams/aws-image-streams/streams/v1/com.ubuntu.cloud.released-aws.json | 21:26 |
| natefinch | Because it makes the test output harder to read, and because CamelCase can never fully encode the expectations of multiple lines of code, so why not give something easier to read? | 21:26 |
| natefinch | no matter what, when you see test Foo failed at line 87, expected "foo" got "bar".... you need to go look at line 87 to understand what exactly failed. | 21:27 |
| deanman | sinzui, eu-west-1 seems like a better fit for me Greek | 21:27 |
| katco | natefinch: i am not really convinced that it makes test output harder to read. i also use underscores in my tests, e.g. Test<precondition>_<expectation> | 21:27 |
| deanman | :-) | 21:27 |
| natefinch | katco: I am generally for being more verbose for tests, but I think you rapidly reach the point of diminishing returns with test names. In addition, I think that long names may obscure minor differences in the name and thus cause confusion. | 21:33 |
| katco | natefinch: how do you propose differentiating happy path tests? | 21:34 |
| natefinch | katco: generally, by the code and/or comments. I mean, obviously they also need separate function names (unless you' | 21:35 |
| natefinch | you're using subtests) | 21:35 |
| katco | natefinch: yes, what will those function names be? | 21:35 |
| deanman | sinzui, http://paste.ubuntu.com/23639866/ | 21:35 |
| natefinch | katco: I guess I see it as mostly subjective and not worth worrying that much about. I wouldn't ding someone for making names I consider too long (unless they're truly egregious, like unable to fit on the screen) and I would hope others wouldn't ding me for making mine too short (again, except egregious outliers, like TestX TestY etc) | 21:37 |
| natefinch | as long as they're relatively descriptive... there's really only so much you can encode in a single line of text. You can't put everything in there, so where do you stop? | 21:38 |
| katco | natefinch: sorry, trying to get at the heart of your objection here. so you're retracting your earlier criticism? | 21:38 |
| katco | natefinch: my objection is that Test<noun> only tells me the thing i'm testing (maybe), but not the behavior, and i care more about the behavior | 21:40 |
| katco | natefinch: it is not about number of characters at all | 21:40 |
| natefinch | katco: my point is that nothing except the code of the test can tell you the behavior. The name of the test can give you a hint. And I thought that Test<Noun> was a reasonable "basic usage test" name. | 21:41 |
| deanman | sinzui, guessing that my juju aws account somehow needs to be authorized to spin centos7 images? | 21:41 |
| katco | natefinch: how do you name your test functions when you want to test 2 behaviors of <noun>? | 21:41 |
| sinzui | :/ | 21:41 |
| natefinch | katco: *shrug* something else. TestFooOdd TestFooEven. The problem comes with more complicated arguments. When you take a struct with 8 fields, it breaks down. | 21:43 |
| sinzui | deanman: I don't think that should be the case. the stream was made with public images. There were selected from a query for centos7 built by some party I don't recall | 21:43 |
| katco | natefinch: not sure what you mean re. # fields | 21:44 |
| * sinzui looks into images | 21:44 | |
| sinzui | deanman: oh, yes, I do recall that message from the pastebin. I think you just need to accept that you will use them and that the owner is not responsible for what might happen | 21:45 |
| natefinch | How do I name a test that tests SetUser with the value User{ Name: "Bill" , Age: 30, SSN: ###, Balance: 50, Maiden: nil, State: MA, Email: "foo@example.com", Active: false } | 21:46 |
| katco | natefinch: what behavior are we hypothetically testing? | 21:47 |
| natefinch | If you are over 25 and your balance is under $100 and you live in HasMinBalanceFee(st State) and you're active, we send an email with your name telling you that your balance is low. | 21:48 |
| natefinch | to be fair, that's only 6 fields that we actually look at | 21:49 |
| deanman | sinzui, unfortunately i can't log into AWS account to opt-in, i think no-one can from the AWS charmer program. | 21:49 |
| katco | natefinch: what label encapsulates the behavior of that situation? | 21:49 |
| natefinch | I would call it TestSendLowBalanceEmail | 21:50 |
| natefinch | But obviously the name doesn't encompass everything in the preconditions necessary. But that's what comments are for. | 21:50 |
| katco | natefinch: what is the label that describes those criteria? a low balance user? | 21:50 |
| sinzui | deanman: well I think that needs to be fixed. the charmer program should allow peple to develop with for centos and windows. I do recall that a member of my team had to accept the use of the centos images. I don't see it in billing though | 21:51 |
| katco | natefinch: TestSetUser_EmailSentWhenLowBalance, or whatever it is that describes that situation | 21:52 |
| natefinch | katco: my point is... it's bikeshedding. If you want a convention of TestFunctionName_TestDetails, I think that's fine as a way to easily find the tests for a specific function. Of course, some tests may test more general cases which aren' | 21:54 |
| deanman | sinzui, Is it Marco that should look into this ? Jorge? | 21:54 |
| natefinch | aren't merely a test of a single function | 21:54 |
| sinzui | deanman: I think jcastro is our point of contact | 21:55 |
| katco | natefinch: if we're testing > 1 thing in a test, this is a problem; in my view, another +1 for this approach | 21:55 |
| katco | natefinch: i'm sorry you see it as bikeshedding; i still can't glean what your objections are | 21:56 |
| sinzui | deanman: I honestly cannot see how my test of centos worked two hours ago. Some one my team must have accepted the terms at the start of this year, but I cannot see any billing regarding our use | 21:56 |
| sinzui | deanman: this is my last day of the year. We might not have an easy solution this month for this, but I am certain we can find a way to ensure developers can get access to our testing images. | 21:58 |
| deanman | sinzui, did try the centos hack for lxd with `lxc image copy images:centos/7 local: --alias ubuntu-centos7` but didn't work. Any hints for me ? | 21:59 |
| deanman | sinzui, thank you for your time, i might send an email to Marco or Jorge | 21:59 |
| natefinch | katco: my objections are that I feel like we're making a rule to fix something that hasn't been shown to be a problem. I do kinda like the ease of searching for TestFunctionName_stuff but I would kind of hope that a search for FunctionName itself in the test file would be just as useful. | 22:00 |
| sinzui | deanman: try an alias like centos-centos7. I think the naming conventions is OS-release | 22:01 |
| natefinch | katco: I have to run and make enchiladas.... but don't get me wrong. I'm not burning down the house to have my way here. I only care like 20%. | 22:01 |
| katco | natefinch: cheers, enjoy your enchiladas | 22:02 |
| natefinch | :D | 22:02 |
| deanman | sinzui, that didn't work either, it looks for ubuntu-centos7 http://paste.ubuntu.com/23639956/ | 22:04 |
| sinzui | oh, it is ignoring the alias-as-cache | 22:04 |
| deanman | hmmm, should that image be on host running juju command or on lxd controller ? | 22:08 |
| sinzui | deanman: I think I know what is failing us, but don't know how to hack around it. Juju once used the matching alias. Then Juju got smart and now checks that the alias's image is not stale. | 22:08 |
| sinzui | deanman: I don't see any setting to convicne juju or lxd to not check | 22:09 |
| sinzui | deanman: I had some misdventures with ubuntu devel images too. would manually copy the image from daily streams. that was good for a few weeks, then I get the image not foudn message you see, so I manually copied a new one in | 22:10 |
| deanman | sinzui, could it be the image autoupdate feature of lxd ? | 22:12 |
| sinzui | deanman: maybe | 22:13 |
| sinzui | deanman: I see "lxc image import" lets you set --expires-at. maybe set that to a decade from now | 22:14 |
| deanman | sinzui, but which lxd daemon is making this check? that of the host or of the controller ? | 22:15 |
| deanman | the logs saying that the image cannot be found come from the controller | 22:16 |
| sinzui | deanman: It is the host of each machine you deploy. the controller does not cache yet. | 22:16 |
| deanman | ok | 22:16 |
| sinzui | deanman: I need to leave. sorry for nit finding a solution today | 22:17 |
| jcastro | deanman: back, what's up? | 22:17 |
| deanman | sinzui, once again, thanks for the tips, have a nice vacation time :-D | 22:17 |
| jcastro | deanman: ok so Marco runs those accounts and he's on a work trip, when he calls me tonight I'll ask him to investigate | 22:18 |
| deanman | jcastro, thank you | 22:19 |
| sinzui | jcastro: Juju will use centos images from http://aws.amazon.com/marketplace/pp?sku=aw0evgkw8e5c1q413zgy5pjce which requires the user to agree to the cost. If the charmer accoutn accepts this, they anyone in the charmers program to test their centos charrms | 22:19 |
| jcastro | ok so what the ask is for us to accept that so that charmers can test on centos then? | 22:20 |
| sinzui | jcastro: yes | 22:20 |
| jcastro | ack | 22:21 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!
