[00:32] sinzui: which version of Go does the win2k12 builder use ? [00:37] davecheney: It does build. The canonistack-slave builds windows and centos with go 1.6 and has done since the announcement on April 10 [01:00] sinzui: which version of Go does the win 2k12 builder use ? [01:00] menn0: https://github.com/juju/juju/pull/5388/files [01:01] davecheney: what ships with trusty [01:01] ^ quick review to help debug 1581157 [01:01] Bug #1581157: github.com/juju/juju/cmd/jujud test timeout on widnows

[01:01] sinzui: i'm sorry I don't understand, i'm askig about the windwso builder [01:01] sinzui: http://reports.vapour.ws/releases/3964/job/run-unit-tests-win2012-amd64/attempt/2384#highlight [01:01] which version of Go was used on this CI run ? [01:03] davecheney: IS will not support us using a windows builder so we dont. we cross-compile from a trusty host. the ALL CI builders use the golang-1.6 package. No other is accepted. Trusty currently has 1.6-0ubuntu1~14.04 [01:04] thanks, i undrstand now [01:05] can someone point me to the code that launches mongo during a bootstrap? [01:05] davecheney: PS, the cross-compile setup is your own script. No one has matched your script for ease of use [01:05] (manual provider, if that matters) [01:19] a script ? you don't need a script for go 1.6 [01:19] GOOS=windows GOARCH=amd64 go ... [01:19] is all you need [01:21] mwhudson: cmd/jujud/bootstrap.go is what initially calls out to start mongo. [01:22] cherylj: i think i've found it now, thanks [01:22] ubuntu@autopkgtest:~$ journalctl -u juju-db | head [01:22] -- Logs begin at Fri 2016-05-13 12:51:41 NZST, end at Fri 2016-05-13 13:20:58 NZST. -- [01:22] May 13 12:57:46 autopkgtest systemd[1]: Started juju state database. [01:22] May 13 12:57:46 autopkgtest mongod[2960]: 2016-05-13T12:57:46.932+1200 W CONTROL [main] No SSL certificate validation can be performed since no CA file has been provided; please specify an sslCAFile parameter [01:22] May 13 12:57:47 autopkgtest mongod[2960]: Invalid command: ignored [01:22] ^ looks bad [01:22] :/ [01:23] and yeah, hacking at the service file makes it go [01:24] which is a bit eeeeeeeerrrrrrrrrm [01:31] ooo kkaaay [01:31] so mongo is running with this command line: [01:31] /usr/lib/juju/mongo3.2/bin/mongod --dbpath /var/lib/juju/db --sslOnNormalPorts --sslPEMKeyFile /var/lib/juju/server.pem --port 37017 --syslog --journal --replSet juju --ipv6 --quiet --oplogSize 1009 --auth --keyFile /var/lib/juju/shared-secret --storageEngine wiredTiger [01:31] but if i run that same command it says [01:31] 2016-05-13T13:31:41.359+1200 W CONTROL [main] No SSL certificate validation can be performed since no CA file has been provided; please specify an sslCAFile parameter [01:32] help? [01:34] oh i see it always says that [01:34] davecheney: looking now. was lunching [01:36] well wtf [01:36] davecheney: ship it [01:36] apparently you have to say --sslPEMKeyPassword=ignored now, and --sslPEMKeyPassword ignored doesn't work [02:00] axw: be right there [02:05] axw: wallyworld: here's http://reviews.vapour.ws/r/4822 FWIW. I'll look back in later, but don't think there' much more I can do tonight. [02:05] redir_: np, ty [02:05] have a good evening [02:08] and https://github.com/fatih/vim-go/issues/814 is what happened to my editor. Syntastic stopped working... === rodlogic is now known as Guest75713 [02:17] whisky tango foxtrot: https://bugs.launchpad.net/ubuntu/+source/juju-mongodb3.2/+bug/1581284 [02:17] Bug #1581284: --sslPEMKeyPassword parameter must be passed with = on yakkety [02:18] mwhudson: welcome to the joys of working with mongo === rodlogic is now known as Guest21818 [02:50] davecheney: i'm not sure about the approach used here http://reviews.vapour.ws/r/4822, if you get a chance, you you take a look to givr your opinion? [02:50] o/ wallyworld I just realized charmsvg was uploaded to the store before resources were required/allowed. I'm about to patcha few things in the app and will upload again with resources [02:51] marcoceppi: ok, thank you [02:51] marcoceppi: and you'll load the resources to the store also? [02:51] wallyworld: I have no choice ;) [02:52] duh [02:52] wallyworld: if you want a really fun demo, you could have an old copy of python-jujusvg on your disk and "upgrade" the resource. I'll start rendering bundles with squircles [02:52] ooooh, that sounds kinda cool actually [02:57] https://github.com/juju/juju/pull/5389 <- easy review [03:00] mwhudson: lgtm. it is really only that one arg that needs the =? I guess is for on/off values? [03:01] wallyworld: there are about three other ones that would need it i think but we don't pass any of them [03:01] mwhudson: maybe add a comment to the code? [03:01] wallyworld: i guess it can't land because master is hilariously blocked? [03:02] yeah, we are about to do a beta7 even though not all races are done. should be unblocked real soon [03:02] mwhudson: btw, using your go 1.7 ppa. can't belive how farking slow the 1.6 compiler is [03:02] 1.7 is a *bit* better [03:02] wallyworld: oh god eyah [03:02] wtf went wrong? [03:03] why did it get so slow? [03:03] the conversion to go mostly [03:03] but the 1.7 *linker* is faster than the 1.4 linker now [03:03] so Go is a worse compiler than C? [03:03] (the compiler is still slower) [03:04] maybe partly that, but also C mechanically translated to Go is probably not optimal Go [03:04] the go compiler is also better in 1.7, so that helps [03:05] at least they are dog fooding now [03:05] wallyworld: i added a comment [03:05] ty [03:06] the explanation I heard is that the compiler is kind of a bad use case for garbage collection, it produces tons of objects and throws them away. if you set GOGC=800, I think it's like 20% faster, but it'll use gobs of memory [03:06] yeah, a lot of work went into making the compiler a more natural Go program this cycle [03:06] natefinch: that's part of it [03:06] natefinch: although not quite right, the problem is that it allocates tons of objects and they *never become garbage* [03:06] ahh [03:06] so the gc is just spinning cpu and not acheiving anything [03:07] (some stuff becomes garbage of course, but relatively little) [03:08] wallyworld: so when can this land? when master is unblocked, but do i have to pay attention and notice when that's happened? [03:09] mwhudson: sadly you need to pay attention. but there's normally an email to juju-dev [03:09] mwhudson: is it holding up anything urgent? [03:10] wallyworld: it's holding up the migration from -proposed to -release in yakkety [03:10] wallyworld: but i can smuggle it in as a distro patch there [03:10] mwhudson: ok, in that case - add the blocker tag to the bug [03:10] and land with $$fixes-xxxxx$$ [03:10] where xxxx is the bug number [03:11] mwhudson: so long as you don't break xenial builds [03:11] wallyworld: not sure i have permission to do that? [03:12] it's bug 1581284 fwiw [03:12] Bug #1581284: --sslPEMKeyPassword parameter must be passed with = on yakkety

[03:12] it's already marked as a blocker [03:12] i just did that now [03:12] Bug #1581284 opened: --sslPEMKeyPassword parameter must be passed with = on yakkety

[03:12] but i don't think i can do the $$fixes-$$ thing? [03:12] you can [03:12] maybe i can, let's find out i guess [03:13] if you can $$merge$$ you can [03:13] i don't think i can $$merge$$... [03:13] ok, i'll do it, we'll have to fix those permissions [03:15] mwhudson: i just added you, can you try to merge? [03:15] wallyworld: so i just comment "$$fixes-1581284$$"? [03:15] yup [03:20] wallyworld: i guess i should file another bug about how utterly unobvious it all is when the juju-db systemd unit fails to start properly... [03:20] that would be good [03:20] our observability in that area is shite [03:23] * mwhudson can only agree [03:24] particularly awesome is the fact that after the bootstrap fails, /etc/whatever/juju-db.service is a dangling symlink [03:28] wallyworld: huh i guess the bot hadn't noticed the tag yet? [03:29] mwhudson: it didn't like your tag for some reason. it acted on mine, but hasn't picked up that that bug number is valid [03:30] hmm [03:30] ah, needs to be crtical [03:31] i think, i'll try that [03:36] Bug #1581297 opened: confusing failure mode if juju-db systemd unit fails to start up [03:36] oh probably [03:37] https://launchpad.net/ubuntu/+source/juju-core/2.0~beta6-0ubuntu2.16.10.2 fwiw [04:23] mwhudson: in case you didn't notice, your fix landed [05:34] wallyworld: it occurred to me that we can manage ZFS from within the controller with the LXD provider, since the controller can manage its own config [05:34] wallyworld: so we could have a ZFS storage provider [05:34] hmmmm [05:34] or LVM [05:34] interesting === blahdeblah_ is now known as blahdeblah [06:44] natefinch https://bugs.launchpad.net/juju-core/+bug/1581337 [06:44] Bug #1581337: natefinch/npipe/v2: Closing a pipe is not safe in a threaded environment [07:04] Bug #1581337 opened: natefinch/npipe/v2: Closing a pipe is not safe in a threaded environment === frankban|afk is now known as frankban [08:53] dimitern: ping [08:53] voidspace: pong [08:57] dimitern: I've sent you a question about transactions in private chat [08:59] voidspace: 'cause it had swears in it. [08:59] voidspace: yeah, I'm replying [09:02] babbageclunk: I'm quite happy swearing here [09:12] Hi, lads! Which mongod version is juju 2.0 going to use? [09:13] Mmike: it's using 3.0 now, but efforts are made to make it use 3.2 before the release [09:14] dimitern: ack! dimitern is it using the default storage engine for used version? mmap in 3.0 and Tigersomething in 3.2 ? [09:15] Mmike: correct (IIRC) - mmap (..something) in 3.0 and wiredtiger in 3.2 [09:15] dimitern++ ack, thnx [09:28] dimitern, pushed tests for http://reviews.vapour.ws/r/4796/ -- not *super* happy with them, just because there's so much *to* test, would very much appreciate your opinion [09:30] fwereade: thanks, will have a look [09:30] dimitern, ...no rush though [09:30] dimitern, maybe best wait, they could use some docs and I should address your comments on the rpeviosu version too [09:32] fwereade: ok, won't be able to get to it for a few hours anyway I'm afraid [09:33] dimitern, perfect ;p [09:54] dimitern, btw, re Annotate/Annotatef, do you feel particularly strongly? [09:55] dimitern, the way I see it Annotatef-everywhere is easier to tweak and has no real disadvantages [10:02] it is stable enough to upgrade to 16.04 lts ubuntu server with maas2.0 for juju? [10:03] fwereade: not strongly, and you have a point - we should likely replace Annotate with Annotatef across the codebase [10:06] fwereade, dimitern either of you know of a juju client api that I can use to get the CharmURL of a given unit?, there appears to be one for services, I'd like to find one for units [10:07] mattyw, isn't the charm url a service level property? [10:07] mattyw, i.e. if you know the unit's service, the charm url will be that of the service [10:08] mattyw: it's been a while since I was in those parts of the API I'm afraid [10:08] tasdomas, correct, I could make a number of calls to get the answer, I'm interested to know if there's a single one that will do all the work I want [10:08] tasdomas, mattyw: the service's charm url might be different during upgrades [10:08] dimitern, ok no problem - I pinged because I saw you were around [10:08] dimitern, yeah, I think that's ok for these circumstaces [10:16] mattyw, tasdomas: a unit's charm url should only set once the charm has been downloaded and verified, iirc -- I don't think there's a specific external api for it, but I'm pretty sure it's available via status [10:17] dimitern, cheers [10:17] fwereade, ok thanks [10:22] mattyw, tasdomas: (also applies when upgrading -- in general, a new value will be set when the charm has overcome all remote hurdles to installing a new charm) [10:27] dimitern, hey, Worker should be called Mortal [10:27] * fwereade wonders if there's a clean anglo-saxon word for waiting until something dies [10:27] fwereade: lol why not? [10:28] dimitern, I think it's actually closer [10:29] dimitern, I just replied pointing out that "any type of task" isn't quite right because all worker *actually* signifies is something whose validity is bounded in time [10:30] fwereade: Ephemeral ? [10:31] dimitern, Mortal.Kill > Ephemeral.Dissipate, or whatever it'd have to be for the metaphor to be consistent [10:31] :D [10:31] .Transmigrate() [10:31] haha [10:33] dimitern, and it goes a lot better with tomb and catacomb ;) [10:34] fwereade: yeah [10:35] dimitern, (hmm, also, mortal.Reap vs worker.Stop) [10:35] fwereade: I love how you always strive to find meaningful and unambiguous names, which (unlike me) are also usually nice and short [10:36] * fwereade had to fight himself not to call RestartWorkers a CruciformWorkers, on the basis that dan simmons references are probably a bit obscure [10:36] dimitern, ty :) [10:36] fwereade - vigil? [10:36] babbageclunk, nice! [10:36] fwereade: Ha, that would have been great. [10:37] fwereade: although vigil isn't really specific to waiting for someone to die, so maybe clear enough [10:37] babbageclunk, it's the best yet for sure [10:38] babbageclunk, I was mostly failing to come up with vulture metaphors [10:38] fwereade: eww [11:15] frobware, voidspace, dooferlad, babbageclunk: http://reviews.vapour.ws/r/4824/ fixes bug 1580417; please take a look (at least some of you :) [11:15] Bug #1580417: Allow a different VPC ID to be used for hosted models [11:16] dimitern: need to finish up on my PR, then will take a look. Might be a little while. === rodlogic is now known as Guest45407 [11:17] frobware: no worries, when you have time - thanks! === rodlogic is now known as Guest44812 === rodlogic is now known as Guest56723 [13:40] anyone know where i can find a description of the configuration attributes accepted by the gce provider? [13:40] juju generate-config used to be a thing but doesn't seem to exist any more [13:41] and there don't seem to be any docs for >=2.0 in https://jujucharms.com/docs [13:42] Bug #1580501 changed: cloudimg-base-url parameters not in Juju2 anymore

[13:42] Bug #1581537 opened: Can't destroy service -- juju status "failed" & unknown relation 0 error in log [13:54] Bug #1540394 changed: Juju does not handle 429 "too many requests" from Azure <2.0-count>

[13:54] Bug #1579051 changed: Race in juju/controller/destroy and TestDestroyCommandConfirmation

[13:55] cherylj: o/ [13:56] hey hey marcoceppi [13:56] cherylj: we're doing some office hours in like 4 mins if you want to drop by [13:56] I can't join the office hours party today :( [13:56] aww, okay :) [13:56] I'm stuck at the dealership waiting for my car to get fixed :/ [13:57] bleh! anything I should pass along? beta7 teaser maybe? [13:57] MAAS2 support [13:57] no longer behind a feature flag [13:57] and create-model will be renamed add-model [13:57] \o/ [13:57] we're hoping to release today, but we'll see how things go :( [14:00] fwereade, ping? [14:01] cherylj: hey, do you mean are we releasing 2.0 today or it's the next beta? [14:01] next beta [14:01] lol [14:01] beta7 [14:01] ah :) [14:02] sorry :) [14:02] ok, got worried for a sec.. [14:02] yeah, that would be bad [14:03] mattyw, pong [14:03] fwereade, I'm doing some stuff that involves calling this api https://github.com/juju/juju/blob/master/api/service/client.go#L121 [14:04] fwereade, after I do I start getting errors about there being an upgrade in progess [14:04] mattyw, from that call? or from elsewhere? [14:04] fwereade, I'm tracking down if there's a bug in some of our stuff, but basically that api shouldn't be responsible right? [14:04] fwereade, I don't know, it's likely from somewhere else [14:04] mattyw, yeah, I'm pretty sure it should not do that :) [14:05] fwereade, ok cool thanks [14:08] frobware, voidspace, dooferlad, babbageclunk: hate to be a pest guys, but can at least someone have a look at http://reviews.vapour.ws/r/4824/ ? [14:10] dimitern: looking now, sorry. Will be a nice change from trying to disentangle this db creation and population stuff. [14:10] frobware: I've found an issue with the lxd broker that causes e.g. http://reports.vapour.ws/releases/3971/job/functional-container-networking-lxd/attempt/160, testing a fix now [14:11] babbageclunk: tyvm! === rodlogic is now known as Guest81137 [14:26] dimitern: can we sync please [14:26] frobware: ok [14:27] Bug # changed: 1571917, 1571932, 1571933, 1574649, 1578456 [14:28] dimitern: 1:1 ho [14:28] frobware: omw [14:33] HOME=/nonexistent juju version [14:33] error: cannot load ssh client keys: mkdir /nonexistent: permission denied [14:33] this is unexpected ^ [14:34] context: running tests on the lp buildfarm [14:35] actually, with pbuilder at the moment [14:39] well this works: HOME=$(mktemp -d) juju version [14:39] so, workaround achieved maybe === rodlogic is now known as Guest5666 [15:33] Bug #1581579 opened: resource-get fails with duplicate key error [15:39] * dimitern really hates it when adding logging to an otherwise intermittently racy state method makes the race disappear [15:40] hi, i seem to be unable to connect to the controller mongo database lately.. are there some instructions somewhere on how to do that? [15:40] thinking it might be related to the move to mongodb 3.2? [15:41] by connect, i mean with the mongo client, to poke around and debug things [15:41] cmars: I wasn't able to since the upgrade to 3.2 [15:42] dimitern, ack, thanks for confirming. i'll email the list === rodlogic is now known as Guest63438 [15:45] dimitern, voidspace, dooferlad, babbageclunk: if you have some cycles... http://reviews.vapour.ws/r/4826/ [15:46] frobware: looking [15:51] fwereade: yt? [15:57] dimitern, dooferlad: halp! Why might my state.runTrasactions call be hanging? [15:57] :( [15:58] babbageclunk: hanging how? [15:58] babbageclunk: what are you trying to do? [15:58] dimitern: Yeah, that probably needs some context... [15:58] :) [15:59] don't panic ;) [15:59] dimitern: I'm pulling apart the initialisation so that we don't recreate the collections and indexes for each test. [16:00] But creating a model isn't just the schema, it's also some data in the collections. [16:00] So now the teardown clears out the collections and puts that data back in. [16:01] babbageclunk: right.. [16:01] dimitern: So basically that's this bit: https://github.com/juju/juju/blob/master/state/open.go#L135 [16:02] Pulled out into a separate function. [16:02] babbageclunk: ok [16:02] But when it gets to the st.runTransaction call (L175) it hangs. [16:03] babbageclunk: but if you're just moving all that code from L135 to the end, keep in mind the defer-s above [16:04] babbageclunk: show me your diff perhaps? [16:05] dimitern: hmm. I hadn't really paid attention to that. I don't *think* it matters - the test has an open state at this point and a cleanup calling state.Close(). [16:05] but I might be wrong. [16:05] Ok, hang on - I'll commit and push. [16:05] babbageclunk: yeah, looking at it closely the defer shouldn't cause it to hang [16:05] babbageclunk: it's all a bit rough at the moment. [16:06] duh, typed my nick [16:06] babbageclunk: for some extra debugging from mgo/txn in tests, try txn.SetLogger(c) and txn.SetDebug(true) [16:06] feeling very Friday afternoony. [16:06] dimitern: Ooh, that sounds extremely useful! [16:13] frobware: LGTM [16:13] dimitern: ty [16:15] dimitern: http://reviews.vapour.ws/r/4827/ [16:15] \o/ not such a bad Friday, 13 after all - the VPC fix is confirmed to work [16:15] dimitern: I think it's fully tested by the existing tests (I had failing tests along the way) [16:16] dimitern: yay \o/ [16:16] voidspace: yay indeed :) looking at your PR [16:17] dimitern: very simple for spaces, assuming I didn't miss anything... [16:23] perrito666, so the connection info hasn't changed or anything -- still the same username/password and everything? [16:24] exactly [16:24] tag for the user and password for password [16:24] cmars: install the shell and check just in case [16:25] perrito666, ok great, thanks! [16:25] cmars: ping me if you need a hand [16:25] voidspace: reviewed [16:26] dimitern: thanks [16:26] could anyone sign a forwardport? http://reviews.vapour.ws/r/4828/ [16:27] dimitern: I don't think there's any point in adding a new spaceGlobalKey name for a single use [16:27] dimitern: maybe entityGlobalKey taking entity name and provider id [16:27] networkEntityGlobalKey maybe [16:28] dimitern: and do you want a ":" separating entity name and provider id? Your email used "#". [16:28] dimitern: it probably doesn't matter but given you've explicitly said ":" in your comment I'd like to check [16:29] voidspace: a generic helper to construct providerIDDoc.ID is better, yeah - taking a globalKey and providerID [16:30] voidspace: as for ":" vs "#" - I've realized some entities use multiple # in their global keys (e.g. LLD and their addresses), hence the ":" [16:30] dimitern: ok, cool - ta [16:31] perrito666: reviewed [16:31] dimitern: tx [16:31] ah tx for that spotting :D [16:32] cherylj: can I $$merge$$ some outstanding PRs now that beta7 appears to be cut? [16:32] perrito666, got connected with the shell. ssl didn't verify though. nbd for poking around in the shell, but where would I find the certs to connect properly? [16:33] cmars: out of the top of my head /var/lib/juju/agent/blablabal/cert [16:33] perrito666, ok, thanks! [16:34] Bug # changed: 1321442, 1543283, 1557143, 1564054, 1564577, 1567518, 1569467, 1572746, 1574607, 1575332, 1575760, 1575768, 1575808, 1576270, 1576376, 1578373, 1579127, 1581284 [16:34] Bug #1581600 opened: should be able to host multiple controllers in a single mongodb [16:43] dimitern: Sorry, got sidetracked, here's where it's hanging in my version: https://github.com/babbageclunk/juju/blob/state-tests/state/open.go#L151 === rodlogic is now known as Guest63003 [16:44] dimitern: Hmm, probably clearer to see the commits: https://github.com/babbageclunk/juju/commit/aa4d79e1bb8a47ac6b558c36ac9a39378d75f56a [16:44] babbageclunk: will have a look in a bit, otp [16:45] dimitern: And then this change to juju/testing: https://github.com/babbageclunk/testing/commit/40e8403b275140383d044253c8d437b77d221d60 [16:45] dimitern: No worries - I'm stopping now anyway, brain is too fried. [16:46] babbageclunk: have a nice weekend :) will post comments on your commit if needed [16:48] dimitern: Thanks! You too. [16:49] babbageclunk: o/ [16:50] dimitern: PR updated http://reviews.vapour.ws/r/4827/ [16:51] still running tests [16:51] bbiab [16:59] voidspace: provisional LGTM - please double check [17:00] dimitern: kk [17:04] Bug #1581612 opened: api: Connection interface not well documented [17:19] ericsnow: natefinch: redir: o/ [17:19] katco: \o [17:20] frobware: I've unblocked master, so you can land [17:21] cherylj: ack. thx [17:27] dimitern, ^^^ [17:27] alright all, I am going to be down for a bit replacing some parts on my labtop [17:27] cherylj, alexisb: thanks! [17:27] biab, hopefully :0 [17:32] alexlist: gl! [17:32] oops wrong alex postfix :) [17:33] ericsnow: natefinch: don't forget, last day to get cards on the tanzanite board + clean up specs [17:35] katco: crud, thanks, been heads down looking at a bug. Will clean up my spec now. [17:52] Bug #1581627 opened: functional-container-networking-lxd fails with invalid parent device name

[18:06] yay, old style lenovo mouse buttons I missed you so! [18:18] alexisb: lol, new laptop? [18:19] rick_h_, no, just updated track pad [18:19] oic === akhavr1 is now known as akhavr [18:21] if i add some extra fields to a bundle.yaml, will juju still deploy it, ignoring them? [18:21] just wondering, for science [18:21] cmars: I *think* so but have to verify. [18:21] i can find out soon enough :) === frankban is now known as frankban|afk [18:34] could someone review my patch ? http://reviews.vapour.ws/r/4809/ === rodlogic is now known as Guest72878 === rodlogic is now known as Guest18774 [19:30] fwereade: let me know if you happen to be around and available... I know it is EoW there. Otherwise I'll hit you up monday. [19:40]