| wallyworld | menn0: if i have a txn-queue uuid from a db.lease.find(), what's the cmd to show the queue size? | 00:01 |
|---|---|---|
| menn0 | wallyworld: the queue size is literally the length of the txn-queue field on the document (it's an array) | 00:02 |
| wallyworld | ah right, the array is length 1 with a uuid | 00:02 |
| wallyworld | on a system with 5 units | 00:02 |
| menn0 | wallyworld: a single item in the array is totally normal | 00:02 |
| wallyworld | yep | 00:03 |
| menn0 | mgo/txn leaves the last txn that touched a document in the field | 00:03 |
| wallyworld | i guess i've leave it run for a bit nd see if the queue size explodes | 00:04 |
| menn0 | wallyworld: the system that experience that bug had a number of services deploy | 00:04 |
| menn0 | deployed | 00:04 |
| wallyworld | i have 2 | 00:04 |
| wallyworld | with 5 units each | 00:04 |
| wallyworld | i can add more | 00:04 |
| menn0 | wallyworld: this one had 8 or something, not sure if that makes a difference | 00:04 |
| menn0 | wallyworld: are you try to repro with master or checking your fix? | 00:04 |
| wallyworld | my fix | 00:05 |
| wallyworld | maybe a should have repoed first | 00:05 |
| wallyworld | i might try a bundle | 00:05 |
| menn0 | wallyworld: it would be good to know if you can see the problem with master first, otherwise you don't know whether you've actually helped | 00:05 |
| wallyworld | true, i just assumed it would show up with enough load | 00:05 |
| wallyworld | deploying landscape bundle anyhow, we'll see what happens | 00:07 |
| wallyworld | axw: anastasiamac: i have to head out for a blood test, been fasting for 12 hours. bbiab with food :-) | 00:23 |
| anastasiamac | wallyworld: good luck :D | 00:23 |
| wallyworld | i studied all night | 00:24 |
| wallyworld | hope i pass | 00:24 |
| menn0 | waigani: is there any reason why this PR hasn't been reviewed yet? http://reviews.vapour.ws/r/1748/ | 00:28 |
| menn0 | waigani: never mind | 00:28 |
| menn0 | waigani: I just saw that it's a forward port | 00:28 |
| menn0 | waigani: to avoid having these clutter up RB i've been manually marking them as submitted in RB even before they get merged | 00:29 |
| menn0 | waigani: it avoids some confusion | 00:29 |
| === kadams54 is now known as kadams54-away | ||
| waigani | menn0: okay, I'll do that in the future. | 00:54 |
| mup | Bug #1459057 was opened: multiple storage-add constraints for the same storage unsupported <juju-core:New> <https://launchpad.net/bugs/1459057> | 01:10 |
| mup | Bug #1459060 was opened: add support for "storage-add <name>" <juju-core:New> <https://launchpad.net/bugs/1459060> | 01:13 |
| axw | menn0: would you kindly take another look at http://reviews.vapour.ws/r/1781/ | 01:19 |
| mup | Bug #1459060 changed: add support for "storage-add <name>" <juju-core:New> <https://launchpad.net/bugs/1459060> | 01:25 |
| davecheney | % tree | grep logger | 01:31 |
| davecheney | │ ├── logger | 01:31 |
| davecheney | │ │ ├── logger.go | 01:31 |
| davecheney | │ │ ├── logger_test.go | 01:31 |
| davecheney | │ ├── logger | 01:31 |
| davecheney | │ │ ├── logger.go | 01:31 |
| davecheney | │ │ ├── logger_test.go | 01:31 |
| davecheney | │ ├── logger | 01:31 |
| davecheney | │ │ ├── logger.go | 01:32 |
| davecheney | │ │ ├── logger_test.go | 01:32 |
| davecheney | │ │ │ ├── logger.go | 01:32 |
| davecheney | juju has at least three packages called logger !>! | 01:32 |
| davecheney | how does this help anything ? | 01:32 |
| anastasiamac | davecheney: the more the merrier? :D | 01:33 |
| mup | Bug #1459060 was opened: add support for "storage-add <name>" <juju-core:New> <https://launchpad.net/bugs/1459060> | 01:40 |
| davecheney | you get a logger! and you get a logger! and you get a logger! /oprah | 01:40 |
| thumper | davecheney: pretty sure those are all mine :) | 01:48 |
| thumper | davecheney: at least I didn't call it "state" | 01:48 |
| anastasiamac | :D | 01:48 |
| menn0 | axw: ship it! looks great. | 01:49 |
| axw | menn0: cheers | 01:49 |
| menn0 | axw: sorry it took me a while, had to go to the post office | 01:50 |
| axw | no worries | 01:50 |
| davecheney | so, today i'm looking at races in the logger and leadership packages | 01:53 |
| davecheney | but I can't figure out which one i need to look at | 01:53 |
| axw | davecheney: leadership and lease are in flux a bit, FYI | 01:53 |
| axw | there's some critical bugs that need fixing | 01:53 |
| davecheney | eek | 01:54 |
| davecheney | http://paste.ubuntu.com/11381996/ | 01:55 |
| davecheney | axw: https://bugs.launchpad.net/juju-core/+bug/1459064 | 01:57 |
| mup | Bug #1459064: worker/leadership: data races in test and code <juju-core:New> <https://launchpad.net/bugs/1459064> | 01:57 |
| axw | joy | 01:57 |
| wallyworld | menn0: \o/ | 02:03 |
| menn0 | wallyworld: ? | 02:04 |
| wallyworld | trivially reproducable | 02:04 |
| wallyworld | "txn-queue" : [ "5565255b0c132d0fb0000150_978b92fe", "556525790c132d0fb0000198_5c4423d2", "556525970c132d0fb00001b8_741af2e9", "556525b50c132d0fb00001cb_36de73a9", "556525d30c132d0fb00001ef_156997b6" ] | 02:04 |
| wallyworld | and my fix works | 02:04 |
| menn0 | sweet! | 02:04 |
| menn0 | that's great news | 02:04 |
| wallyworld | yeah | 02:04 |
| * wallyworld goes to hit merge | 02:04 | |
| menn0 | now i'm curious as to what you changed | 02:04 |
| menn0 | i'll have to check out the diff | 02:04 |
| wallyworld | it's not a complete fix for all the issues | 02:05 |
| wallyworld | just this one specific thing | 02:05 |
| thumper | wallyworld: have you fixed leadership? | 02:05 |
| thumper | oh... | 02:05 |
| wallyworld | thumper: one teeny little bit | 02:06 |
| thumper | are ya gunna fix it all? | 02:06 |
| wallyworld | axw will :-) | 02:06 |
| wallyworld | we are both working on it | 02:06 |
| wallyworld | and illiam | 02:06 |
| wallyworld | w | 02:06 |
| mup | Bug #1459064 was opened: worker/leadership: data races in test and code <juju-core:New> <https://launchpad.net/bugs/1459064> | 02:07 |
| menn0 | wallyworld: one thing to check... do the txn-queue entries get automagically fixed when upgrading from a system with the bug to one with your fix? | 02:08 |
| wallyworld | menn0: that's a good point. i guess i should try that | 02:08 |
| wallyworld | since this started in 1.23 :-( | 02:08 |
| wallyworld | if it were 1.24 betas, i wouldn't bother | 02:09 |
| wallyworld | menn0: if it's not fixed, what would the answer be run the txn purge tool | 02:09 |
| wallyworld | ? | 02:09 |
| menn0 | wallyworld: I guess the new txnpruner worker should take care of them | 02:10 |
| menn0 | wallyworld: b/c they are completed txns | 02:10 |
| wallyworld | i'll try it out | 02:10 |
| menn0 | wallyworld: actually, it won't | 02:11 |
| menn0 | wallyworld: b/c the txnpruner doesn't modify the txn-queue field. it only removes docs from the txns collection for completed txns that are no longer referenced | 02:12 |
| menn0 | wallyworld: so it won't help | 02:12 |
| wallyworld | menn0: i'm worried this is a time bomb which will hit other deployments | 02:12 |
| menn0 | wallyworld: but I have a feeling that mgo/txn may tidy up the txn-queue fields for you | 02:12 |
| wallyworld | we'll see soon once my deployment bootstraps and i upgrade | 02:12 |
| menn0 | yep. fingers crossed. | 02:13 |
| wallyworld | menn0: my upgrade hung the state server - i think i may have hit bug 1457728. but in any case, i'm thinking maybe it might be best just to wipe the lease collection on upgrade | 02:39 |
| mup | Bug #1457728: `juju upgrade-juju --upload-tools` leaves local environment unusable <local-provider> <upgrade-juju> <vagrant> <juju-core:Triaged> <juju-core 1.24:In Progress by axwalk> <https://launchpad.net/bugs/1457728> | 02:39 |
| menn0 | wallyworld: is that safe? | 02:40 |
| wallyworld | i think so. i'll try upgrading again first | 02:41 |
| === kadams54 is now known as kadams54-away | ||
| menn0 | wallyworld: is one result of your PR that ClaimLeadership doesn't block forever? that's the issue that's causing me hassles at the moment | 03:17 |
| wallyworld | menn0: that aspect is not deliberately fixed | 03:17 |
| menn0 | wallyworld: ok. i'll have to figure out another way to do this | 03:18 |
| wallyworld | might be by accident, but there was no intent to change that behaviour | 03:18 |
| wallyworld | menn0: what s the scenario | 03:18 |
| menn0 | wallyworld: i'm creating feature tests to ensure that logging to mongodb works | 03:18 |
| menn0 | wallyworld: the one that spins up the machine agent works | 03:19 |
| menn0 | wallyworld: the unit agent variant hangs during teardown | 03:19 |
| wallyworld | ah, deadlocks on test teardown - william has found ways to work around that | 03:19 |
| menn0 | wallyworld: it's because in the unit agent the leadershiptracker hangs making it's initial request for leadership | 03:19 |
| wallyworld | menn0: i'm am fairly sure william has been through this pain and i though he had a solution | 03:20 |
| menn0 | wallyworld: I suspect it's possibly also b/c i'm using JujuConnSuite so there isn't a full set of workers running on the server side | 03:20 |
| wallyworld | possibly | 03:20 |
| menn0 | wallyworld: which is possibly why it's getting stuck in the first place | 03:20 |
| menn0 | wallyworld: but i don't know how the lease/leadership stuff works all that well | 03:20 |
| menn0 | wallyworld: i'll bug will if I don't get anywhere by EOD | 03:21 |
| wallyworld | yeah, i'd ask william, i'll mention it as i'm meant to be talking to him later | 03:21 |
| wallyworld | menn0: until a proper worker with tombs is used, and the singleton goes away, it will remain broken | 03:21 |
| menn0 | wallyworld: ok | 03:22 |
| menn0 | wallyworld: i'm just wondering if I can get enough infrastructure in place so that leadership does work in the tests | 03:22 |
| wallyworld | menn0: ack, but sadly i can't give you a good answer | 03:23 |
| mup | Bug #1459082 was opened: flag for debug-hooks to skip non-existent hooks <debug-hooks> <juju-core:New> <https://launchpad.net/bugs/1459082> | 03:23 |
| menn0 | wallyworld: np, thanks | 03:24 |
| menn0 | wallyworld, thumper: I think I just cracked it. spinning up the lease manager loop allows the leadership functionality in the uniter to work and unblocks the test. | 03:37 |
| wallyworld | ah, that sounds plausible - i didn't realise the loo[ wasn't running | 03:37 |
| wallyworld | menn0: when/if you have a moment, i'd like to check something wth you about upgrades, maybe ping me when you've done writing tests | 03:38 |
| menn0 | wallyworld: under JujuConnSuite it's not | 03:38 |
| wallyworld | :-( | 03:38 |
| menn0 | wallyworld: i wouldn't have expected it to... JujuConnSuite gives you State and an API server but not necessarily all the workers that the state servers run | 03:39 |
| wallyworld | that's true | 03:39 |
| wallyworld | but 'tis a trap | 03:39 |
| menn0 | wallyworld: yep and it's a hard to diagnose trap | 03:47 |
| menn0 | wallyworld: i'll be free in a sec to discuss that upgrade issue | 03:47 |
| wallyworld | ok, ta, see you in onyx | 03:47 |
| davecheney | thumper: https://bugs.launchpad.net/juju-core/+bug/1459085 | 03:50 |
| mup | Bug #1459085: worker/logger: data race in tests <juju-core:New> <https://launchpad.net/bugs/1459085> | 03:50 |
| davecheney | this is a bug in loggo | 03:50 |
| davecheney | i have a fix to propse | 03:50 |
| davecheney | which branches of juju does this need to land on > | 03:50 |
| davecheney | which branches of juju does this need to land on ? | 03:50 |
| thumper | hmm | 03:51 |
| thumper | davecheney: what is the loggo data race? | 03:51 |
| thumper | davecheney: go test -race in loggo seems fine... perhaps we need more coverage? | 03:52 |
| thumper | davecheney: the races probably need to land on master | 03:52 |
| thumper | unless they are obviously fixing broken behavour in supported jujus | 03:52 |
| davecheney | i'm on the fence on that one | 03:53 |
| davecheney | if I can avoid spending a day nursing it back to 1.22 | 03:53 |
| davecheney | i'd prefer to avoid it | 03:53 |
| davecheney | thumper: btw, you going to land this ? | 03:54 |
| davecheney | https://github.com/juju/loggo/pull/6 | 03:54 |
| * thumper looks | 03:54 | |
| thumper | no | 03:55 |
| thumper | decided against it | 03:55 |
| davecheney | thumper: then close it | 03:56 |
| davecheney | thumper: https://github.com/juju/loggo/pull/11 | 03:56 |
| davecheney | will try to add a test | 03:56 |
| davecheney | it is likely that older versions of the race detector cna't spot this bug | 03:57 |
| thumper | davecheney: just did | 03:57 |
| thumper | close it that is | 03:57 |
| thumper | I wonder if the bot is doing this yet | 03:57 |
| * thumper pokes to see | 03:57 | |
| thumper | davecheney: ah... ok | 03:58 |
| thumper | davecheney: without this patch, do the current loggo tests exhibit a problem? | 03:58 |
| thumper | I'm guessing no, because I don't have goroutines doing things in the tests | 03:58 |
| davecheney | ok, the loggo tests dno't trigger this because they always use ConfigureLogger which uses a mutex which acts as an accidenal memory barrier | 03:58 |
| davecheney | to trigger it you need to do; logger.SetLevel(x); go logger.LoggerInfo() | 03:59 |
| davecheney | it has to be in a second goroutine otherwise by definition there is not race | 03:59 |
| davecheney | lemmie try to add a test as well | 04:03 |
| davecheney | i'll do that in another PR | 04:03 |
| thumper | davecheney: looks like there is a bot | 04:06 |
| mup | Bug #1459085 was opened: worker/logger: data race in tests <juju-core:New> <https://launchpad.net/bugs/1459085> | 04:11 |
| davecheney | thumper: hmm, this is going to be tricky to write a test for | 04:14 |
| davecheney | let me try a bit more, then give up if the worker/logger test's race is fixed | 04:14 |
| natefinch | son of a.... ERROR juju.worker runner.go:219 exited "upgrader": invalid series "wily" | 04:16 |
| natefinch | haven't we fixed this dumb bug by now? | 04:17 |
| natefinch | (the "Juju doesn;t know about future series" bug | 04:17 |
| natefinch | wallyworld, thumper ^ thoughts on how I can get around this? I just want to test the upgrade step I'm writing | 04:19 |
| wallyworld | natefinch: install the distro info package | 04:20 |
| wallyworld | juju can't guess what a new series is | 04:20 |
| natefinch | wallyworld: but it can install the distro info package itself :/ | 04:20 |
| wallyworld | i guess so, yeah | 04:20 |
| wallyworld | natefinch: any progress on bug 1370896 as we need to look to get 1.24 wrapped up | 04:21 |
| mup | Bug #1370896: juju has conf files in /var/log/juju on instances <canonical-bootstack> <logging> <rsyslog> <juju-core:Triaged by natefinch> <juju-core 1.24:In Progress by natefinch> <https://launchpad.net/bugs/1370896> | 04:21 |
| natefinch | wallyworld: exactly what I'm working on. We need an upgrade step to move the config files from the old location to the new location. | 04:22 |
| wallyworld | makes sense | 04:22 |
| wallyworld | you should be in bed anyway | 04:22 |
| davecheney | thumper: got a test | 04:22 |
| davecheney | it's going to come with a LARGE comment | 04:22 |
| natefinch | wallyworld: well, I work when I get a chance, sometimes having 3 kids under 4 years old means that chance is quite late at night | 04:23 |
| wallyworld | 3!! | 04:23 |
| wallyworld | you've been busy | 04:23 |
| wallyworld | s/quite/very | 04:24 |
| natefinch | wallyworld: heh, yeah, we had our third, a boy, on Christmas eve. So we have a 5 months, almost 2, and almost 4 year old | 04:24 |
| wallyworld | and an upcoming vasectomy :-) | 04:24 |
| davecheney | actally no | 04:24 |
| davecheney | can't add a test for this easily | 04:24 |
| natefinch | wallyworld: Ironically, I'm the one that wants one, and my wife wants another kid... but i think she's crazy :) | 04:25 |
| wallyworld | me too :-) | 04:25 |
| wallyworld | you could always get one secretly | 04:25 |
| mwhudson | natefinch: my brother had 4 kids at ~2 year intervals | 04:25 |
| mwhudson | i can't say i recommend it | 04:25 |
| natefinch | lol | 04:26 |
| natefinch | yeah, no | 04:26 |
| natefinch | that's what my wife wants... but I'm quite happy with 3. 4 presents a host of problems we can narrowly avoid with 3 | 04:26 |
| wallyworld | larger car, larger house | 04:27 |
| natefinch | wallyworld: exactly. House is 3 bedrooms... the girls are doubling up, and hopefully that'll be fine for a long time... but the other bedroom is pretty small for two kids. | 04:27 |
| wallyworld | long time = until they turn into teenagers | 04:27 |
| natefinch | right... at which point the one we trust the most can move into the finished basement | 04:28 |
| wallyworld | ha, and you will be outside the door with a shotgun | 04:28 |
| natefinch | so anyway, distro-info has not helped, even after restarting jujud | 04:28 |
| wallyworld | hmmm | 04:30 |
| natefinch | yeah, distro-info --all doesn't contain wily | 04:30 |
| wallyworld | wtf, well that sucks | 04:31 |
| natefinch | sounds like one of those "someone put up a stream that they shouldn't have" things | 04:31 |
| wallyworld | i would have thought it woud have | 04:31 |
| wallyworld | add it by hand to get by for now i guess | 04:31 |
| natefinch | is there a reason we can't just ignore series we don't understand? I mean, if we're not even using that series, who cares? | 04:32 |
| natefinch | remind me where that file is, again? | 04:32 |
| wallyworld | /usr/share something, i'll check | 04:36 |
| wallyworld | /usr/share/distroinfo | 04:36 |
| natefinch | wallyworld: got it, thanks | 04:36 |
| wallyworld | i think the idea is to stop typos and misconfigured systems | 04:36 |
| natefinch | it just seems to screw us like clockwork, every 6 months | 04:37 |
| wallyworld | eg accidentally running a centos series with ubuntu tools | 04:37 |
| wallyworld | yeah | 04:37 |
| natefinch | no matter how smart we try to get | 04:37 |
| wallyworld | waigani: you working on bug 1459047? can you assign yourself and mark in progress? | 04:38 |
| mup | Bug #1459047: juju upgrade-juju --upload-tools broken on ec2 <juju-core:New> <https://launchpad.net/bugs/1459047> | 04:38 |
| wallyworld | so we know it's being looked at | 04:38 |
| === natefinch is now known as natefinch-afk | ||
| waigani | wallyworld: done | 04:40 |
| wallyworld | tyvm | 04:40 |
| wallyworld | waigani: helps when we have release meetings each morning | 04:40 |
| waigani | wallyworld: of course, I'll make a note to keep all that updated | 04:41 |
| wallyworld | ty | 04:41 |
| mup | Bug #1459093 was opened: Upgrade fails if there's a series in streams Juju doesn't recognize <upgrade-juju> <juju-core:New> <https://launchpad.net/bugs/1459093> | 04:59 |
| dimitern | thumper, hey, still around? | 05:52 |
| dimitern | menn0, ^^ | 05:53 |
| dimitern | fwereade, hey | 06:34 |
| dimitern | fwereade, re http://reviews.vapour.ws/r/1777/ | 06:35 |
| dimitern | fwereade, I'd appreciate if you reply to my mail from yesterday and give some rationale around why are we bothering to migrate the remaining state-bound workers to use the API, as it seems it's not generally obvious for a few people | 06:36 |
| fwereade | dimitern, the resumer worker one? | 07:23 |
| dimitern | fwereade, yes | 07:36 |
| fwereade | dimitern, ok, I'm a little bit more concerned that anyone considers it ok for random workers to touch the database directly, but will do | 07:37 |
| dimitern | fwereade, :) cheers | 07:38 |
| TheMue | morning o/ | 08:27 |
| TheMue | dimitern: I see there are discussions about the API usage migration of the workers? | 08:27 |
| dimitern | TheMue, morning | 08:34 |
| dimitern | TheMue, yeah | 08:34 |
| thumper | dimitern: I have come back to do a little more | 08:35 |
| thumper | dimitern: whazzup? | 08:35 |
| dimitern | thumper, hey, I wanted to ask about your opinion about moving the resumer worker to use the API and run once per apiserver - http://reviews.vapour.ws/r/1777/ | 08:38 |
| thumper | once per apiserver or once per environment? | 08:39 |
| thumper | I don't see any real reason to have the resumer use the API | 08:39 |
| thumper | it only ever runs on state server machines | 08:39 |
| thumper | dimitern: to make it use the api server seems weird to add layers for the sake of layers when it doesn't need it | 08:40 |
| dimitern | once per apiserver - it applies to all hosted environments AIUI | 08:40 |
| thumper | well, we only really need one running | 08:40 |
| thumper | what do we have now? | 08:41 |
| dimitern | thumper, fwereade can explain better why it's bad to have *any* workers coupled to state.State directly | 08:41 |
| thumper | dimitern: well... for me it seems like busy work, and something that isn't really needed to be done right now | 08:42 |
| thumper | why now? | 08:42 |
| dimitern | thumper, before my PR, it was running in the StateWorker of the machine agent, now I moved it in the method starting per-env workers, but I'll move it to postUpgradeWorker instead | 08:42 |
| dimitern | thumper, now is as good time as any, and it was planned with lower priority for quite some time | 08:42 |
| thumper | dimitern: what problem is being fixed by the work? | 08:42 |
| thumper | dimitern: if there is an active remit to move every worker away from state directly, then I'm fine with it | 08:43 |
| dimitern | thumper, decoupling the worker from *state.State, improving tests (dropping JujuConnSuite across the board in favor of BaseSuite + mocking), faster and more deterministic tests, the ability to evolve state.State and what the worker needs separately | 08:44 |
| thumper | sure | 08:44 |
| mup | Bug #1459148 was opened: azure: juju can't create compute/network optimized instances <juju-core:New> <https://launchpad.net/bugs/1459148> | 08:44 |
| dimitern | thumper, so can you review that PR with your comments why it's a good thing ? :) | 08:46 |
| thumper | dimitern: not right now :) | 08:46 |
| dimitern | thumper, ok, sure | 08:49 |
| thumper | dimitern: but I will bring it up with menno tomorrow | 08:49 |
| dimitern | thumper, cheers! | 08:51 |
| fwereade | thumper, do you recall "api everywhere" as being a major goal about a year ago? having resumer there indicates that we didn't bother to actually finish that work, and at least some people were actively lying about it being complete | 08:54 |
| fwereade | thumper, and afaics left the door open so that everybody just thought, ehh, direct db access? not so bad, nobody really cares, ehh | 08:55 |
| thumper | fwereade: IIRC we stopped when the non-state server machine and unit agents were able to not use state | 08:55 |
| thumper | we agreed that doing all was a time sink at that stage | 08:55 |
| thumper | I'm not saying it is a worthwhile goal | 08:55 |
| thumper | just questioning timing | 08:55 |
| thumper | effort vs. priorities | 08:56 |
| fwereade | thumper, if that was the case why did we do, eg, firewaller? | 08:58 |
| * thumper shrugs | 08:58 | |
| thumper | wasn't me :) | 08:58 |
| fwereade | thumper, because we don't want all our workers running across the db like a gang of drunken monkeys | 08:58 |
| thumper | heh | 08:58 |
| thumper | now that is an amusing image | 08:58 |
| fwereade | thumper, we *will* fuck things up :) | 08:58 |
| thumper | dude, we screw up all the time | 08:59 |
| * thumper points fwereade at our bug list | 08:59 | |
| fwereade | thumper, I would like the bulk of our potential fuckups to at least pass through an authorization layer and not to have the power to do literally anything | 08:59 |
| thumper | sure | 08:59 |
| thumper | I'm not questioning the work, just the order of work | 09:00 |
| fwereade | thumper, ok, but we've got people writing new workers that use state now | 09:00 |
| thumper | oh? | 09:00 |
| fwereade | thumper, dblogpruner is new, right? | 09:00 |
| thumper | yeah... | 09:00 |
| fwereade | thumper, why would we leak the fact that the logs are stored in the db? | 09:00 |
| fwereade | thumper, the api server can have a "prune logs now please" method | 09:01 |
| thumper | point taken, I'll bring it up tomorrow | 09:01 |
| fwereade | thumper, cheers | 09:01 |
| thumper | actually, I can do better than that | 09:02 |
| * thumper puts a card on the board | 09:02 | |
| fwereade | thumper, <3 | 09:02 |
| voidspace | dimitern: it seems that with kvm we can supply a template to "uvt-kvm" | 10:08 |
| voidspace | dimitern: the template is "libvirt domain xml" format | 10:08 |
| voidspace | dimitern: and that can have a network interfaces section specifying MAC address | 10:08 |
| voidspace | dimitern: by default, we get the default template (of course)... | 10:08 |
| dimitern | voidspace, great! have you actually tried if it works ? :) | 10:08 |
| voidspace | dimitern: (i.e. we're not using the template argument currently) | 10:09 |
| voidspace | dimitern: no... | 10:09 |
| voidspace | dimitern: will do | 10:09 |
| voidspace | dimitern: we'll have to clone the default template and add a network interface definition | 10:09 |
| voidspace | dimitern: https://libvirt.org/formatdomain.html#elementsNICS | 10:10 |
| perrito666 | good morning | 10:13 |
| dimitern | voidspace, can't we generate the template always, like with lxc.conf? (based on the default template XML + some text/template tags) | 10:14 |
| voidspace | dimitern: well, yes | 10:15 |
| voidspace | dimitern: I meant that initially we'll have to clone the default one | 10:15 |
| voidspace | dimitern: expressed badly | 10:15 |
| dimitern | voidspace, sounds good | 10:19 |
| perrito666 | aghh I think the lease singleton is making my tests die :( | 10:33 |
| fwereade | perrito666, are you running a dummy provider, and are you trying to run your own lease worker? | 10:35 |
| perrito666 | fwereade: none, but I think that the fact of importing lease might be making my test go for 10m until they are killed | 10:36 |
| fwereade | perrito666, that is unlikely, I think | 10:37 |
| perrito666 | fwereade: well Ill know more in about 10m :p | 10:37 |
| fwereade | perrito666, -test.timeout 20s | 10:37 |
| fwereade | the mere fact of importing lease will not cause that | 10:38 |
| fwereade | perrito666, what are you trying to do with it? | 10:38 |
| perrito666 | fwereade: I stand corrected its something else | 10:44 |
| perrito666 | every outpug log looks different after coffee | 10:44 |
| perrito666 | (and putting my glasses on( | 10:45 |
| fwereade | perrito666, lol | 11:05 |
| === urulama__ is now known as urulama | ||
| * TheMue takes a bike-ride to the new potential co-location office now. we'll see how fast I'll be back online ;) | 13:04 | |
| mup | Bug #1459250 was opened: centos 7 is ambiguous in streams <centos> <simplestreams> <juju-core:Triaged> <juju-core 1.24:Triaged> <https://launchpad.net/bugs/1459250> | 14:00 |
| wwitzel3 | ericsnow, natefinch: ping | 14:03 |
| natefinch | sinzui: https://bugs.launchpad.net/juju-core/+bug/1459093 | 14:19 |
| mup | Bug #1459093: Upgrade fails if there's a series in streams Juju doesn't recognize <upgrade-juju> <juju-core:New> <https://launchpad.net/bugs/1459093> | 14:19 |
| natefinch | sinzui: wily isn't in distro-info (at least on trusty)... but I'm guessing someone published a stream for wily | 14:20 |
| sinzui | :/ How many times does that bug need to be fixed | 14:20 |
| natefinch | sinzui: just once... we need to stop caring if we see series we don't understand, and just ignore them | 14:20 |
| natefinch | sinzui: my environment shouldn't care if someone publishes a stream for the series "TOTALLY_INVALID" | 14:20 |
| sinzui | natefinch, we haven't build willy | 14:21 |
| sinzui | natefinch, unit tests are run on a machine with a fake series registered in distro-info-data to catch this error | 14:21 |
| natefinch | sinzui: it's not a problem with something extra in distro-info... it's a problem with someone somewhere mentioning a series that *isn't* in distro-info | 14:22 |
| sinzui | natefinch, I think wallyworld reported a similar bug/solution. | 14:22 |
| natefinch | ls | 14:22 |
| sinzui | natefinch, yep, it is the same long standing issue that juju thinks it has to know the world | 14:23 |
| mup | Bug #1459288 was opened: TestWriteTokenReplaceExisting fails <ci> <unit-tests> <juju-core:Triaged> <https://launchpad.net/bugs/1459288> | 14:54 |
| * fwereade has just had guests arrive, stopping for now | 15:00 | |
| sinzui | mgz, abentley to either of you have a moment to review https://code.launchpad.net/~sinzui/juju-release-tools/agent-archive/+merge/260321 | 15:02 |
| abentley | sinzui: sure. | 15:03 |
| abentley | r=me | 15:05 |
| abentley | sinzui: So I have this bug:https://bugs.launchpad.net/juju-core/+bug/1449208 | 15:12 |
| abentley | sinzui: I know it exists in jes-cli. I don't have evidence that it exists in master. | 15:12 |
| sinzui | :/ | 15:12 |
| abentley | Should I create a jes-cli series and target to that? | 15:13 |
| sinzui | abentley, I don't want to, but I think Lp requires it | 15:13 |
| abentley | sinzui: I will be the bad guy, then :-) | 15:13 |
| * sinzui really hates lp bug targeting | 15:14 | |
| * natefinch ^^^^ +1000 | 15:16 | |
| mgz | I think we need to sort-of ignore bugs on feature branches | 15:17 |
| mgz | eg, windows tests are currently borked on jes-cli - but that's kinda their problem | 15:18 |
| natefinch | or just target all bugs to master and use tags to track branches. Then we'd actually be able to search all the bugs at once, too. | 15:19 |
| mup | Bug #1459298 was opened: TestMissingServerFile fails <ci> <test-failure> <juju-core:Incomplete> <juju-core jes-cli:Triaged> <https://launchpad.net/bugs/1459298> | 15:24 |
| === kadams54 is now known as kadams54-away | ||
| mup | Bug #1278831 changed: debugging first run of install hook is not straight forward <debug-hooks> <juju-core:Triaged> <https://launchpad.net/bugs/1278831> | 16:18 |
| mup | Bug #1458693 changed: juju-deployer fills up ~/.ssh/known_hosts <juju-core:New> <https://launchpad.net/bugs/1458693> | 16:18 |
| mup | Bug #1458754 changed: $REMOTE_UNIT not found in relation-list during -joined hook <hooks> <relations> <juju-core:New> <https://launchpad.net/bugs/1458754> | 16:18 |
| mup | Bug #1459327 was opened: Juju MAAS netwoking with custom bridge inside service <juju-core:New> <https://launchpad.net/bugs/1459327> | 16:18 |
| mup | Bug #1459337 was opened: UniterSuite setup fails <ci> <intermittent-failure> <unit-tests> <juju-core:Triaged> <https://launchpad.net/bugs/1459337> | 16:18 |
| natefinch | abentley: is there a python equivalent of gofmt that'll fix the linting rules enforced by make lint? like trailing whitespace, visual indenting, etc? | 17:05 |
| natefinch | sinzui, mgz: ^^ | 17:08 |
| sinzui | natefinch, no. there are some tools that will do some that that, but nothing that reformats to pep8 | 17:09 |
| natefinch | boo.... manually wrapping lines by hand is uh... not a hugely valuable use of my time | 17:10 |
| natefinch | especially when it then complains that I don't indent things perfectly. | 17:12 |
| mgz | natefinch: what editor do you use? | 17:13 |
| sinzui | natefinch, maybe this will suffice https://pypi.python.org/pypi/autopep8 | 17:13 |
| natefinch | mgz: sublime | 17:13 |
| mgz | natefinch: top google result is a pep8 autoformat plugin | 17:14 |
| mgz | for "sublime python format" | 17:14 |
| natefinch | yeah, just found and installed one... so much better | 17:15 |
| === kadams54-away is now known as kadams54 | ||
| hatch | using 1.24-beta5.1-trusty-amd64 the deploy command simply hangs is this a known bug? | 17:23 |
| natefinch | hatch: pretty sure it works fine for me | 17:29 |
| hatch | natefinch: ok I'm going to restart this machine, see if I can get it working or reproduce reliably | 17:34 |
| sinzui | abentley, mgz: do either of you have time to review https://code.launchpad.net/~sinzui/juju-ci-tools/more-agents/+merge/260353 | 18:05 |
| abentley | sinzui: looking | 18:28 |
| abentley | sinzui: r=me | 18:31 |
| sinzui | thank you abentley | 18:41 |
| === kadams54 is now known as kadams54-away | ||
| natefinch | ahh yes, the old "you have to update 100 fragile mocks in order for your tests to pass" | 20:41 |
| natefinch | s/fragile/fragile and incomprehensible/ | 20:41 |
| perrito666 | natefinch: ah, welcome to my world | 20:42 |
| natefinch | perrito666: heh | 20:42 |
| natefinch | perrito666: I've had central A/C for 5 years... if the house is above 75°F/24°C, everyone's miserable. Actually, I'm not too bad, but my wife is really uncomfortable, which means I'm uncomfortable | 20:44 |
| perrito666 | oh, so 27 C is a bad temp for you | 20:44 |
| natefinch | yes | 20:44 |
| perrito666 | now I get it | 20:44 |
| perrito666 | I keep forgetting that you people have blood instead of coolant | 20:45 |
| natefinch | yep | 20:45 |
| perrito666 | you need more south americanness | 20:46 |
| natefinch | I agree, but my wife would probably be mad at me if I imported some | 20:47 |
| natefinch | mental note: upgrade steps get run as-is on your local machine during tests... which means if they delete files off disk... guess what?? | 20:47 |
| perrito666 | lol | 20:48 |
| natefinch | luckily I was just deleting stuff from /var/log/juju ... but still | 20:48 |
| === natefinch is now known as natefinch-afk | ||
| === _thumper_ is now known as thumper | ||
| wallyworld | sinzui: so looking at osversions, i can see why it was done - we just have 13.10 etc for juju series | 21:41 |
| wallyworld | but we use win8 etc for windows | 21:41 |
| wallyworld | so i guess for !ubuntu, we need to be more explicit | 21:41 |
| wallyworld | ericsnow: can i have a trivial review pretty please? http://reviews.vapour.ws/r/1800/ | 21:52 |
| ericsnow | wallyworld: sure | 21:52 |
| ericsnow | wallyworld: ship it! | 21:52 |
| wallyworld | ty | 21:53 |
| wallyworld | ty | 21:53 |
| sinzui | wallyworld: yeah, that was my expectation. | 21:58 |
| wallyworld | sinzui: fix landing for 1.24, soon master | 21:59 |
| === kadams54 is now known as kadams54-away | ||
| thumper | cherylj: regarding the forward port of the vivid/systemd issues, if the branch hasn't significantly changed from the version that landed on the earlier branch, you don't need another review | 22:35 |
| thumper | just get the bot to merge | 22:35 |
| thumper | if the branch does have significant changes to the previous, it is useful to say what those where in the description | 22:35 |
| thumper | this is to smooth the way for fixes to be forward ported | 22:35 |
| thumper | given that *most* of the time, it is exactly the same fix | 22:36 |
| thumper | in the newer versions | 22:36 |
| === perrito666 is now known as perrito | ||
| === perrito is now known as perrito666 | ||
| === perrito666 is now known as perrito | ||
| mup | Bug #1453634 changed: juju upgrage-juju --upload-tools hangs with jes flag <ec2-provider> <upload-tools> <juju-core:Invalid by waigani> <juju-core trunk:Invalid by waigani> <https://launchpad.net/bugs/1453634> | 23:04 |
| mup | Bug #1459047 changed: juju upgrade-juju --upload-tools broken on ec2 <juju-core:Invalid by waigani> <https://launchpad.net/bugs/1459047> | 23:04 |
| wallyworld | waigani: i have a bug for your todo list for beta 6 - bug 1441478 | 23:40 |
| mup | Bug #1441478: state: availability zone upgrade fails if containers are present <upgrade-juju> <juju-core:Triaged> <juju-core 1.24:Triaged> <https://launchpad.net/bugs/1441478> | 23:40 |
| waigani | wallyworld: yay :) | 23:40 |
| wallyworld | hopefully won't bee too bad | 23:40 |
| waigani | wallyworld: :), just finishing this one off, then I'll get onto it | 23:41 |
| wallyworld | ty | 23:41 |
| === perrito is now known as perrito666 | ||
| menn0_ | thumper: http://reviews.vapour.ws/r/1802/ pls | 23:44 |
| === menn0_ is now known as menn0 | ||
| menn0 | thumper: easy one :) | 23:44 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!