[00:23] q: how long does it take to run zero juju tests [00:23] real 26m57.934s [00:24] user 76m59.990s [00:24] sys 9m22.711s [00:24] 26 minutes to test nothing!! [00:24] go test -i ./... && time go test ./... [00:31] davecheney: that cannot be the build time of the test suite [00:35] that is the time to run all the tests, excluding actually running tests [00:35] sorry, hang on [00:35] i posted the rwong numbers [00:35] * perrito666 is expectant [00:59] blahdeblah: \o/ thank yuo. I saw :D [01:00] Looks like it's some sort of leadership election strangeness [01:02] blahdeblah: m working on 1.25 and the bug 's on the radar. cannot tell u right now when it'll get some love but should happen \o/ [01:02] real 15m30.976s [01:02] user 56m52.867s [01:02] sys 2m50.591s [01:02] lucky(~/src/github.com/juju/juju) % go test -i ./... && time go test ./... -run=XXX [01:02] anastasiamac: cool - thanks [01:02] 15:30 to build and run zero tests [01:03] I should document my workaround on that bug [01:03] blahdeblah: tha would b fantastic! [01:03] anastasiamac: Not really; it's just "restart all the things" :-\ [01:04] blahdeblah: :D [01:04] thumper: it seems that the share model functionality does not support setting readonly = true. there is a read only flag but it's always false. is that your recollection? and we never access the flag anyway [01:04] blahdeblah: it would be great to isolate the bug, i.e. minimum setup and steps to reproduce :D [01:05] wallyworld: hey [01:05] wallyworld: yes, very much so, but cmars's team is taking that on [01:05] I wrote the bit that made read only possible [01:05] ah, righto, thanks [01:05] but we don't yet support it at the cli, or client api [01:05] anastasiamac: I just deployed my env & waited. Yesterday morning I was working on it just fine; came back yesterday evening and it was broken. [01:05] to set / get / list [01:05] yep [01:06] wallyworld: was there another question about that? [01:07] thumper: there's been discussion about create-model and who should be allowed to do it etc, as if i can create a model in your controller using your creds, i can cost you money. so i'm just seeing what's been done in that area [01:07] nothing [01:08] the current expectations are that creds are passed up explicitly [01:08] since that is to change [01:08] yes, in a --config file [01:08] there needs to be discussion [01:09] yup. i think it should be if you are the admin user of a controller you can create a model using teh saem creds [01:09] otherwise you need to supply them === pmatulis_ is now known as pmatulis [03:15] Bug #1544847 changed: configSuite.TestNewModelConfig test failure lxd [03:15] Bug #1548799 changed: bootstrap from windows fails with no such file [03:15] Bug #1550810 changed: TestDetectCredentialsKnownLocationWindows missing region [03:21] Bug #1544847 opened: configSuite.TestNewModelConfig test failure lxd [03:21] Bug #1548799 opened: bootstrap from windows fails with no such file [03:21] Bug #1550810 opened: TestDetectCredentialsKnownLocationWindows missing region [03:33] Bug #1544847 changed: configSuite.TestNewModelConfig test failure lxd [03:33] Bug #1548799 changed: bootstrap from windows fails with no such file [03:33] Bug #1550810 changed: TestDetectCredentialsKnownLocationWindows missing region [03:51] thumper: the apiserver/client package imports juju/agent just to get one constant, which brings in the entire api client ... [03:51] what's the constant? [03:51] const SystemIdentity = "system-identity" [03:52] why does the apiserver want that constant? [03:52] for reasons that I do not understand [03:52] the apiserver has a client [03:52] and that client does things, on remote machines ?!?! [03:52] i have NFI why the apiserver has a client [03:52] and why that client is juju/api [03:53] https://github.com/davecheney/prdeps [03:53] wrote a simple tool to print out dep graphs [03:53] the shit it's turning up is making my hair turn white [03:53] the server has a client??? [03:54] yes [03:54] apiserver/client [03:54] um... [03:54] that is the implementation of the client facade is it not? [03:54] it isn't an api client [03:54] but it imports the api client [03:55] and then uses the api client to do stuff [03:55] O_o! [03:59] oof [05:41] anastasiamac: if you get a moment sometime, could you please look at http://reviews.vapour.ws/r/3999/ [05:51] wallyworld: [05:51] looking :D [09:05] jam: ping, 1:1 today? [09:09] frobware: sorry. I'm omw [09:11] frobware, morning [09:11] dimitern: morning [09:11] frobware, I have figured out how to do the multi-nic containers backed by devices [09:12] frobware, I have 2 machines running - each with 2 nics, first disabled, the other one with 4 vlans [09:12] frobware, and on each of these nodes, I have 2 containers running with 5 nics each, using the respective br-eth1[.xx] bridges [09:13] pinging works, connecting with nc shows correct source and destination addresses [09:14] so we need to make devices with physical nics, generated macs, and linked to the correct maas vlan and subnet [09:14] can't do vlan interfaces on devices, but it wasn't necessary [09:40] Bug #1551141 opened: Juju bootstrap local - cannot get replset config: not authorized for query on local.system.replset [09:43] Bug #1551141 changed: Juju bootstrap local - cannot get replset config: not authorized for query on local.system.replset [09:43] frobware, FYI, I've updated the spike card assigned to me with all the relevant info [09:52] Bug #1551141 opened: Juju bootstrap local - cannot get replset config: not authorized for query on local.system.replset [10:01] frobware: hangout time === dooferlad_ is now known as dooferlad [12:05] dooferlad: frobware: dimitern: http://reviews.vapour.ws/r/4000/ [12:05] all tests pass and manual testing works [12:19] voidspace, reviewed [12:20] voidspace: I had one question [12:21] frobware: we still do name translation because maas can still use names with spaces (etc) [12:21] frobware: once maas is fixed we can remove name translation [12:21] dimitern: thanks [12:22] frobware: we will kill it with fire as soon as possible :-) [12:23] voidspace: ok, gotcha [12:36] frobware: ping [12:40] voidspace: pong [12:41] * dimitern whew ... ok vlans work across hosts and containers - tagged traffic gets untagged before reaching the container [12:46] dimitern: :-) [13:48] morning === lazypower is now known as lazyPower [14:43] voidspace, dimitern, frobware: http://reviews.vapour.ws/r/4002 -- The API server + mongo space change is ready for review. [14:43] ericsnow: you around? [14:45] dooferlad: will look in a bit, just finishing up a review too. [14:45] dooferlad, cheers, will look at it soon [14:46] frobware, dimitern: thanks. It looks big, but there is a huge name change in one of the test files that takes up a big chunk of it - don't be put off! [15:26] Bug #1551276 opened: 2.0-beta2 stabilization [15:32] Bug #1551276 changed: 2.0-beta2 stabilization [15:38] Bug #1551276 opened: 2.0-beta2 stabilization [16:25] ericsnow: ping? [16:25] natefinch: hey [16:26] ericsnow: hey... I'm working on that upgrade charm, pull metadata from charmstore card [16:26] ericsnow: there's a task that says "Set available field in resource doc." ... is that out of date? [16:26] natefinch: yep [16:27] natefinch: really, I'd expect that we'd be doing exactly the same thing as deploy for this [16:29] natefinch: along with an extra step to release the stored resource blob, if necessary [16:30] ericsnow: well, serverside, I don't think we added anything for deploy about the charmstore at all [16:30] ericsnow: there's just NewResolvePendingResourcesOps [16:30] natefinch: that's what I'm talking about [16:30] ericsnow: ok, so we're relying on the client to push up the new resource metadata? [16:31] natefinch: ah [16:32] natefinch: given that we now have --resources (like deploy), we should follow suit [16:32] ericsnow: ok, that's easy enough (and in fact I've already done that part). The talk about hooking into the same machinery as resource poller threw me off [16:33] ericsnow: I'm worried that this will break if you upgrade using the GUI, though [16:33] natefinch: yeah, this was before --resources [16:33] natefinch: we have the same problem with deploy though, no? [16:35] ericsnow: then we should fix it for both, right? :) [16:36] ericsnow: I was thinking of just adding a charmstore.ListResources and Resources.SetCharmStoreResources in the state code that does deploy and upgrade-charm [16:36] ericsnow: (for charmstore charms) [16:37] natefinch: if I recall correctly, for resources we chose to follow the precedent of charm handling for deploy [16:38] natefinch: ergo, pull info on client-side [16:39] (from charm store) [16:40] ok.. maybe we can discuss with katherine when she gets back later... we obviously need clientside support for --resources, but it would be nice to also have serverside support for clients that don't do that. [16:43] natefinch: k, though I expect we'll punt on it for 2.0 :) [16:50] frobware: dimitern: note that Jay is out today and the networking hangout is not on (according to the calendar anyway) [16:51] voidspace: ack [16:52] frobware: I've completed my self review (written it anyway) [16:52] frobware: would you like to see it and comment before I submit it, or shall I just submit? [18:34] katco: let ericsnow and I know when you're back. We need to discuss some scope issues around my current task [19:00] ericsnow: I was going to skip uploading metadata for resources whose revision hasn't changed... but that requires an extra round trip to the charmstore... do you think it's worth it? [19:01] natefinch: nope [19:05] ericsnow: well, good :) [19:35] natefinch: ericsnow: o/ [19:36] katco, ericsnow: moonstone? Just want to clarify some questions [19:36] natefinch: sure [21:31] natefinch: standup [21:35] should i be able to bootstrap lxd on xenial now, with juju built from master? [21:36] cmars`: Makyo mentioned that he got it running today [21:36] Wellll. [21:36] It wiped out my /etc/sudoers and now I can't do anything. [21:36] ohhh [21:37] what?!! that sounds bad. for me, it seems to hang during bootstrap, and i wondered if anyone else had seen it [21:37] hangs after this: https://paste.ubuntu.com/15247197/ [21:37] rumor is master + lxd stable ppa is supposed to work? [21:37] cmars`: same problem, then times out after 10 minutes [21:37] which is weird, i can lxc exec into the container and it's just run cloud-init [21:37] Makyo: you're more patient than me :) [21:38] but good to know [21:38] Makyo: using lxd with zfs? [21:38] i almost wondered if it was so fast, we hit a race? :) [21:39] cmars`: ext4 [21:39] + encrypted home [21:39] Er... whole disk encryped, I guess. [21:40] Makyo: same here (LVM, ecryptfs), but i'm using the zfs block storage for lxd, just a big file, I think. it really does help -- though i bet it'd be even faster if it was native [21:42] well, i'm going to open a bug then [21:42] I'll add my head [22:02] cmars`: that sounds like the image problem that serge is working with Tom Barber on in the list [22:04] cmars`: have a chance to poke at this ? https://lists.ubuntu.com/archives/juju/2016-February/006699.html [22:05] or Makyo it soudns like? [22:05] Sure [22:09] Bug #1551440 opened: Can't bootstrap LXD on Xenial [22:12] rick_h__: yep, i think that explains it [22:12] thanks [22:18] Bug #1551440 changed: Can't bootstrap LXD on Xenial [22:48] katco: FYI, the new cards are up [22:48] ericsnow: tyvm [23:56] rick_h__: cmars` the problem I ran into with sudoers was due to permissions installing lxd images. Disregard it.