[00:05] Bug #1587644 changed: jujud and mongo cpu/ram usage spike

[00:43] wallyworld: you said you reviewed this PR but I don't see a review from you. https://github.com/juju/juju/pull/6261 [00:43] menn0: never mind it's on RB [00:43] wallyworld: ^^^ [00:44] wallyworld: there was no RB link on the PR so I presumed that RB had missed it. [00:45] juju model-config logging-config="=TRACE;" [00:46] menn0: oh, i just went stright to rb dashboard [00:46] wallyworld: all good, my bad [00:49] menn0: if you get a chance sometime, here's a quickie for mark http://reviews.vapour.ws/r/5701/ no rush [00:52] menn0: what's that, a bug with RB? [00:52] ;) [00:53] natefinch: well our integration with RB at least [00:54] wallyworld: what does "inactive" mean in the HA column? (vs "-") [00:54] menn0: that HA is not enabled. "-" means we don't have any info at all. that only occurs just after bootstrap before we update the controllers yaml with machine and agent info [00:55] wallyworld: ok right [00:55] so right after bootstrap we have a skeleton yaml [00:55] with not much [00:55] so a few things show as "-" [00:55] then as bootstrap ends, we update with address ingo etc [00:55] including machine and model counts [00:55] wallyworld: and "1/3" means HA with 3 nodes is configured but only one node is up/active? [00:56] yep [00:56] juju show-controller has really cool info [00:56] what is in list is a short summary [00:56] wallyworld: is there something to distinguish between HA nodes being on their way up verse a formally functioning node having died? [00:57] that's in show-controller [00:57] list just has that little summary [00:57] this pr doesn't change that [00:57] it just shows HA column all the time [00:57] even when there's no HA [00:57] hence the addition of "inactive" [00:57] wallyworld: ok sounds good [00:58] personally, I'd prefer N/A to inactive... I might second guess what inactive means, but N/A is pretty unequivocal [00:59] wallyworld: natefinch has a good point [00:59] wallyworld: aside from that ship it (I did it on GH) [00:59] i was following mark's suggestion [00:59] i can change it [00:59] :-( gh [00:59] thanks for review [01:00] we can try N/A and get feedback i guess [01:06] wallyworld: was going to ask during standup but got distracted by shiny things.. what was the outcome of upload-tools discussion at release call? [01:07] anastasiamac_: they were doing it wrongly :-) [01:07] we can't allow binaries to lie about their version [01:07] wallyworld: \o/ best kind of case [01:07] it was a shortcut that no longer works for good reason [01:08] :) [01:38] veebers: any idea what happened here? http://juju-ci.vapour.ws:8080/job/github-merge-juju/9255/ [01:41] veebers: looks like something is up with the windows build host or something. the last 2 merge attempts failed in the same way [01:44] menn0: I'll look now [01:45] menn0: I see a series of messages of 'undefined ...' then an error: http://juju-ci.vapour.ws:8080/job/github-merge-juju/9255/artifact/artifacts/windows-err.log [01:47] veebers: yeah it looks like the run of `python ci/gotesttarfile.py` failed. do you know what that does? [01:48] menn0: not yet, looking now :- [01:48] :-) [01:48] veebers: thank you [01:49] menn0: from the docstring: Run go test against the content of a tarfile. [01:49] veebers: ok, well it's failing to even run the test for some reason [01:50] menn0: So the job copies across the .tar.gz then runs go test against it, but fails at some point [01:50] yeah [01:52] menn0: it gets as far as "Building test dependencies" before barfing [01:54] menn0: from what I can decipher the build command is: 'powershell.exe', '-Command' go.exe, 'test', '-i', './...' [01:55] veebers: there must be some output from that which will provide the reason for failure [01:56] hmm.. [01:56] what a surprise [01:56] look [01:56] another rabbit hole [01:56] menn0: I think that's what the undefined errors are right? [01:56] menn0: http://pastebin.ubuntu.com/23184603/? [01:57] veebers: ah right... missed that [01:57] veebers: sorry, I was being dense. that's certainly a problem related to this change. [01:57] * menn0 fixes [01:57] menn0: nw [02:04] thumper: super quick one: https://github.com/juju/utils/pull/238 [02:04] * thumper looks [02:04] * thumper is hungry [02:04] need lunch [02:05] thumper: I was and then I had lunch.... way too much lunch. Feel a tad sick now. [02:07] found an annoying bug in testing package [02:07] submitting a fix now [02:10] * thumper thrashes the cpu for a bit [02:14] thumper, wallyworld: do we actually support controllers running CentOS or is that only for workload machines? [02:14] workload only [02:14] we only support ubuntu controllers [03:35] * thumper sighs [03:35] seems I've traded one intermittent failure for another [03:37] * thumper goes to make coffee while stress test runs [04:06] menn0: ping [04:07] menn0: I need another brain [04:11] thumper: mine is fairly much mush by this stage of the week but I will try :) [04:11] menn0 thumper anastasiamac_: if any of you have the time, I would appreciate a review on https://github.com/juju/juju/pull/6265 [04:13] menn0: 1:1 [04:13] thumper: give me a sec, just got a phone call [04:14] thumper: coming [04:39] Bug #1578059 changed: Default route not coming up with juju 1.25.5 and bonding

[04:40] menn0: i go this govet error [04:40] cmd/juju/controller/listcontrollersformatters.go:38: Fprintln call ends with newline [04:40] axw_: I'll look soon [04:40] but that was deliberate as i wanted an extra newline [04:41] and didn't see the need for a whole extra println call [04:41] it also seems our landing bot ignores govert errors [04:41] wallyworld: I've fixed that in my PR [04:42] axw_: awesome ty. but i still don't think it should be flagged as an error [05:21] wallyworld: I see your point, I guess a \n at the end could often be a mistake though [05:22] wallyworld: it might clearer to use Fprintf with \n\n on the end anyway [05:22] that would work [05:28] axw_: i've changed the PR to add the cred watcher to environ tracker. i added the WatchCredential() api to the ModelConfig facade as it was convenient to do so (since environ tracker uses that already) and it is sort of related. but maybe that's stretching it a bit. http://reviews.vapour.ws/r/5691/ [05:34] wallyworld: hrm. I'd prefer if it were elsewhere, but it's at least just about watching them, and not obtaining them. [05:35] wallyworld: on a side note, seems that we're not using environ-tracker in the firewaller yet [05:35] axw_: it can be moved, was just convenient [05:35] so firewaller won't be updated unless we add watching there, or move it over to using environ-tracker (which would be preferable) [05:35] i couldn't see a good existing facade [05:35] yeah, that latter [05:36] wallyworld: it's on the agent facade, which is fine. I just don't want it mixed in wiht the common ModelWatcher [05:36] that's not a facade, it's a mixin [05:36] fair point, i can add to agent facade directly [05:37] axw_: did you tlk to QA about public clouds yaml? [05:39] wallyworld: I sent an email to aaron, curtis and torsten that it needs to happen. I asked to hold off until at/around RC1 though, as it's a breaking change [05:39] ok, ta [05:39] wallyworld: are you doing the change to WaitForEnviron in a follow-up ? [05:40] axw_: i should do it now before landing [05:40] wallyworld: what's there looks fine, apart from the location [05:40] yeah [05:40] just reviewing your branch [05:40] then will try and do it quickly [05:41] i took short cut on the location [05:41] menn0: https://github.com/juju/testing/pull/110 [05:44] thumper: looking [05:52] thumper: done [05:52] ta [05:57] wallyworld: do you have time to check if your azure account is active and usable? I've just pushed a juju snap that has interactive auth in it [05:57] axw_: my trial expired, i need to sign up with a cred card etc [05:57] wallyworld: no worries then, I'll ask rick [05:57] ok, i can see if i have time in a bit maybe [05:58] wallyworld: no it's ok, too distracting [06:01] fingers crossed [06:01] thumper: omg, if we fix "no reachable server", we'd address most of these at least ... https://bugs.launchpad.net/juju?field.searchtext=no+reachable&search=Search&field.status%3Alist=NEW&field.status%3Alist=INCOMPLETE_WITH_RESPONSE&field.status%3Alist=INCOMPLETE_WITHOUT_RESPONSE&field.status%3Alist=CONFIRMED&field.status%3Alist=TRIAGED&field.status%3Alist=INPROGRESS&field.status%3Alist=FIXCOMMITTED&field.omit_dupes=on& [06:01] fixed the intermittent pinger test failures [06:01] anastasiamac_: I'm not sure I've got all of them [06:01] thumper: I said "most" to leave a little bit of doubt :-P [06:02] :) [06:02] axw_: move watch to apiserver/agent facade [06:02] *moved [06:03] looking [06:03] axw_: damn, forgot a bit [06:03] just need to move the api bit [06:04] wallyworld: and the "TODO(wallyworld) - pass in credential watcher" ? [06:04] sigh, yeah [06:05] axw_: actually, that bit is correct [06:05] we pass in the api for watching model config [06:05] but not credentials [06:05] yet [06:05] maybe i should rename w [06:08] wallyworld: as in, you're going to do it in a followup? [06:08] menn0: here is the other half https://github.com/juju/juju/pull/6266 [06:09] axw_: as in whoever does the next bit of work to actually use the new watcher in environ tracker; might be me when i get back [06:09] wallyworld: ok, got it [06:09] oh you mean rename w? [06:09] nope, that's what I meant, all clear [06:09] ok [06:12] there, moved the api bit [06:13] hmm... [06:13] just realised that I don't need those wait for alarms [06:13] since the test is now more robust [06:14] menn0: it's updated [06:15] wallyworld: perhaps you could cast your eyes over? [06:15] https://github.com/juju/juju/pull/6266 [06:15] ok [06:15] * thumper goes to make some dinner === thumper is now known as thumper-spamalot === frankban|afk is now known as frankban [09:56] Is 1 GB the lowest amount of RAM juju expects when provisioning a machine? It the ram could be expressed in MB? [09:58] frankban: Another small review: removing some trivial uses of url.Channel in charmstore code. https://github.com/juju/charmstore/pull/681 [09:59] babbageclunk: looking [09:59] frankban: thanks! [10:00] babbageclunk: lgtm [10:00] \o/ [10:00] anyone? [10:01] the ammount of ram could be expressed in MB inide juju ? I'm talking about HardwareCaracteristics inside the juju/instance pkg. [10:10] ping [10:10] hoenir: I'm not familiar with that code, but from the look of it Mem is in bytes - it uses parseSize to turn M/G/T/P units into a float64 number of bytes (I guess so you can say 1.5G), and then multiply to a number of bytes. === thumper-spamalot is now known as thumper [10:10] hoenir: Sorry, meant to say a uint64 of bytes. [10:15] babbageclunk, thanks really appreciate it [10:15] hoenir: :) === cueb6526 is now known as chris38 === cueb6526 is now known as chris38 === freyes__ is now known as freyes [13:39] man, it's kind of sad that it took us 4 years to get a Go program to run on all flavors of linux. But at least it does, now. [14:25] babbageclunk, ping [14:48] easy review anyone? +12 -12 https://github.com/juju/version/pull/2 [15:07] alexisb: pong, sorry [15:29] * rick_h_ runs for lunch with family biab === DaveJ is now known as dave-johnston [15:43] man, there's like nobody online today. [15:49] natefinch: lgtm [15:49] waht mgz said natefinch [15:50] thanks guys [15:50] redir: you closed the PR :/ [15:52] no citizenM, I do not want your newsletter :/ [16:20] natefinch: wrong button somehow [16:40] rick_h_: hey [16:40] rick_h_: question, does application dict entries support a series key === frankban is now known as frankban|afk [17:05] natefinch: help? [17:05] availability-zone [17:05] how do I set that as a constraint [17:05] rick_h_: ^? [17:10] Can anyone help us with availability zones? We are in an openstack environment if that matters [17:10] beisner: ^ ? [17:12] Juju tried to spread servers on different zones and picked up a zone that was invalid. We want to specifically put a new machine on a zone we know is good. [17:22] hi mbruzek - i've not exercised juju against multiple AZs with the juju openstack provider [17:22] OK [18:41] marcoceppi: sorry, had to run an errand, still need help? [18:42] availability zone is not a constraint it's a placement directive --to zone=foo [18:44] (admittedly, there's a fine line there) [19:05] natefinch, his q looks to have been answered in another channel, you can thank balloons [19:10] cool [20:32] alexisb: yt? [20:32] redir, yes sir [20:32] wuz up? [20:32] HO for a minute? [20:33] alexisb: ^ in the standup room [20:33] sure [21:08] Hey, working with juju/httprequest, how do I send CORS headers for a service? All my usual resources are exhausted or on vacation [21:09] wait, juju add-unit --to ? [21:09] that seems awkward [21:15] marcoceppi, awkward how? that is how we to direct machine placement everywhere [21:16] alexisb: not sure, I didn't expect it and it wasn't in the help output for add-unit [21:17] I am seeing it [21:17] in the help output [21:17] alexisb: I'm on beta18 and don't see it, grep "zone" reveals nothing [21:17] alexisb: maybe it's an rc thing ;) [21:18] I suppose you could argue that machine characteristics like instance-type, cpu, mem, etc are also "placements" which is why I lumped zone with constraints [21:18] but having explained it now, it makes sense [21:18] no it should be there [21:18] pasted what I am looking at in private chat [23:26] Bug #1624579 opened: error bootstrapping rackspace provider [23:32] Bug #1624579 changed: error bootstrapping rackspace provider [23:35] Bug #1624579 opened: error bootstrapping rackspace provider