[00:24] Bug #1484308 opened: MachineSuite fails

[00:27] davechen1y: yeah, i should reply to that === meetingology` is now known as meetingology [01:48] Hey guys I am going to sleep I won't make it awake to the team meeting [01:48] perrito666: heh, good night. [01:54] ericsnow: you around? [02:21] natefinch: I am if that helps [02:29] wwitzel3: there's a line in the hook context's flush that looks dangerous/bad [02:30] wwitzel3: https://github.com/juju/juju/blob/feature-proc-mgmt/process/context/context.go#L206 [02:30] wwitzel3: when we flush, we set updates to nil, but the rest of the code expects that updates will be non-nil [02:32] natefinch: I thought we explictly checked for len(c.updates) in all those places? [02:33] natefinch: or we have a !ok check for the key accessor [02:33] natefinch: if not, we should [02:33] wwitzel3: we don't, so we should. Or just not set it to nil, just set it to an empty map. [02:37] natefinch: yeah, might as well just set it to an empty map [02:49] wallyworld: re your comment in http://reviews.vapour.ws/r/2330/. I would rather not add that assertion. it's done in other tests already. these tests are big and unwieldy already, I'm trying to make them a bit more targeted [02:50] axw: sure, np [03:18] wallyworld: did you want to review http://reviews.vapour.ws/r/2332/ (azure volume source), or should I go ahead and land it? [03:18] axw: looks like it's had 2 others, so go for it i reckon [03:20] I do so like an unblocked master [03:37] wallyworld: 2nd point in follow-ups in description: "- update storage provisioner to set status when detaching or destroying" [03:38] wallyworld: I can do it in this branch if you prefer [03:38] ah [03:38] tis ok, sorry i missed that [04:00] thumper: Quick question about jujulib, I'm trying to find the package for python-websocket. I couldn't see anything on pypi. [04:00] :) [04:00] huwshimi: websocket_client [04:01] thumper: Ah hah! [04:01] huwshimi: quick hangout? [04:01] thumper: Sure [04:01] huwshimi: https://plus.google.com/hangouts/_/canonical.com/onyx-standup [05:25] davechen1y: https://github.com/juju/juju/pull/2977 please [07:43] TheMue: morning === anthonyf is now known as Guest27914 [08:07] Bug #1484403 opened: provider/openstack: default block source not set to "cinder"

[08:55] Bug #1484419 opened: Local provider: fail to download charm from state server when using an isolated network

[09:05] dooferlad: ping === urulama_ is now known as urulama [10:16] yeeeehaw, mini provider mock and envrion work for test [10:29] Bug #1307445 changed: openstack: instances with multiple nics on the same network don't have deterministic private-addresses

[10:29]

[12:08] marcoceppi: ping? [12:09] perrito666: yooooo [12:09] liberal icmp protocol [12:09] icmp for the 21st centry [12:11] marcoceppi: can you take a look at this review comment and give me your opinion on how we should display the new available version on tabular format? http://reviews.vapour.ws/r/2314/#comment15670 [12:11] ill give some context [12:11] I have added a new tools version availability check [12:11] context is for chumps, I dive right in and make uninformed decisions on everything I do ;) [12:11] that will display, in status, whenever a new version is available for upgrade [12:13] so, that will only display when you have a newer version of tools available? [12:13] yep [12:16] perrito666: aesthetically I'd prefer it if had a header, something like [Juju] and then just a Running: and Available: key, but that's really just personal opinion to keep everything inline style wise. No one is going to be parsing the tabular output so I'm fine whatever people choose. My only real concern is the amount of realestate it takes, if it's going to consume 4 lines at the top it better be damn important because in deployments of > [12:16] 5 services and 10 units info on normal displays start to truncate while running watch [12:16] perrito666: also, how does this look in yaml and json? is it parsable there? [12:18] marcoceppi: in yaml and json is parsable [12:19] it is at the same level as environment [12:19] cool [12:20] perrito666: so, tl;dr; I'd prefer a heading and simple output (just the keys) but I don't have a strong opinion on the subject, I'd just find it pleasing to the eye [12:20] marcoceppi: well you are more of a user than we are [12:20] and position, top? [12:25] perrito666: hum, that's a good question. Considering most people won't be upgrading the moment a new release comes out (IS, for example) having that nag as the top item seems wasteful. I'd probably possition it between machines and units but that feels weird too [12:27] perrito666: having it at the bottom is fine too. I think it's a bit foolish to assume juju is the most important thing to display in the output. To us it may be, to the user I think they're more concerned about their workload [12:27] well shame on them [12:28] * perrito666 is about to go fix mramm's connection [12:28] marcoceppi: thanks for your imput [12:57] dimitern: TheMue: if you have time http://reviews.vapour.ws/r/2360/ [12:57] relatively straightforward === anthonyf is now known as Guest46427 [12:58] dimitern: TheMue: dammit, it's not ready yet - belay that === Guest46427 is now known as anthonyjf [13:04] dimitern: TheMue: ok, it's ready now... [13:04] (again...) [13:23] why do we hardcode series in Juju? [13:28] marcoceppi: just to piss everyone off [13:28] marcoceppi: including ourselves [13:28] natefinch: but like, I'm serious. What's the use for it in juju? I'm trying to understand the justification [13:29] (without reading a bunch of go code) [13:30] marcoceppi: ostensibly it's so we can know if someone is using a series that's too old. Though I would think the lack of tools would be a perfectly acceptable way to stop them from deploying to a machine where it wouldn't work. [13:30] * marcoceppi considers ripping all this out and just submitting a patch [13:30] yeah so, we're broken on OSX and Windows because of this [13:31] natefinch: what do you need from us to help? [13:31] honestly, what we should do is hardcode the series that are too old, since those are static. [13:31] * marcoceppi hates the idea of hard coding anything [13:31] jcastro: what problem are you seeing? [13:31] Simplestreams seems to be the best means [13:31] doesn't work on OSX El Capitan or Windows 10 [13:32] le sigh [13:32] like, we don't work at all currently [13:32] for the client... we shouldn't look at series at all. If the code can run enough to look at series, it's probably fine. [13:33] +50 [13:34] natefinch: in the meantime: https://github.com/AdamIsrael/juju/commit/909bf8e3074c0bae25579e84f41a71996fe61a3f [13:36] jcastro: gah, yeah. Send a PR [13:37] natefinch: https://github.com/juju/juju/pull/2969 [13:38] natefinch: we'll need a fix for Windows 10 as well, which is actually released and in the wild. [13:40] natefinch: incoming from marco in a few minutes [13:40] jcastro: cool. I told CI to merge the OSX one [13:48] OK [13:49] I understand why we have series matching for client, it appears to be able to map caveats for things? [13:49] like on Linux do this on Windows do this and on OSX do this? [13:50] marcoceppi: we don't need series for that, just OS, and Go gives us OS detected (i.e. darwin vs linux vs windows). What it doesn't give us is, for example, trusty vs Wily or centos vs ubuntu [13:50] natefinch: I'm just trying to figure out how to remove hard coding versions all together [13:50] like, why do we need to know what version of osx or windows you're running at that point? [13:50] marcoceppi: I am all for that [13:50] natefinch: well, you' [13:50] ll love this patch [13:51] marcoceppi: one trick is local provider [13:51] igaf about local provider, since it's only Ubuntu and we should be able to detect ubuntu fairly well given we created the OS [13:52] idgaf* [14:03] * katco sitting in the standup all by my lonesome, wwitzel3 natefinch ericsnow [14:04] katco: introspective standup :p [14:11] anybody in for a fast one? http://reviews.vapour.ws/r/2361/ [14:11] natefinch: Sorry, I fixed the failing test for that osx patch [14:11] aisrael: cool [14:13] dimitern: I'm postponing the test of isNetworkingEnvironment() in cmd/jujud/agent/manchine.go. as we talked about this morning SupportsNetworking() does not rely on the feature flag but on a type assertion of the environment [14:14] dimitern: will add a card for it [14:23] aisrael: can you add back a test for version 16 producing series unknown etc? [14:24] natefinch: done [14:39] cmars: I'd like to have a chat about dependency engine at some point today if you have some time [14:47] wwitzel3: you going to be around for our 1:1? [15:29] katco: hey, sorry, just woke up [15:39] wwitzel3: no worries... in a meeting atm [15:58] TheMue, sounds good [16:30] wwitzel3: k sorry... are you up for a 1:1? or is your throat too sore? [16:54] Bug #1484606 opened: bootstrap fails if control bucket not specified and exists [17:45] Bug #1484617 opened: Juju doesn't honour agent-stream: released in the config

=== natefinch is now known as natefinch-afk [19:33] Bbl [20:45] Bug #1484688 opened: [Juju HA] Past logs do not get synced to new slave state servers [21:05] cherylj, you around? [21:05] o you are in standup [21:05] alexisb: in standup [21:05] nm, ping me post standup please [21:05] alexisb: will do [21:24] Bug #1478232 changed: juju 1.24 poor performance

[21:33] mramm: ping? [22:12] Bug #1450092 changed: juju fails to bootstrap local provider with vivid and upstart

[22:12] Bug #1460087 changed: quickstart deployment fails to add relations when bootstrap goes "down"

[22:12] Bug #1484718 opened: String config values can't be set as all integers [22:42] Bug #1410876 changed: Error executing lxc-clone: lxc_container: utils.c: mkdir_p 220 Not a directory - Could not destroy snapshot %s - failed to allocate a pty; Insufficent privileges to control juju-trusty-lxc-template

[22:42]

[22:42] Bug #1473450 changed: upgrade-juju from 1.18 to 1.20.14 leaves some agents down

[23:43] mm, gce as a provider works really well, cheers wwitzel3 and eric