[00:16] perrito666: ping [00:27] uh, https://github.com/juju/juju/pull/6938 [00:27] tough one there, I might self review [00:29] * anastasiamac_ looking [00:29] redir: LGTM \o/ [00:43] tx anastasiamac_ [01:21] wallyworld: re https://bugs.launchpad.net/juju/+bug/1662587, any thoughts on what we can do that doesn't involve going back to what we were doing? [01:21] Bug #1662587: 2.1beta5 breaks lxd shared controller [01:24] axw: so the bug is saying that adding a user, and then having someone else register doesn't work? because that new person doesn't have the credentials cached? [01:28] hello, what can i do if i want to merge 163a221,773fb90 to 2.0.x? lp said only 2.1 and 2.2 https://bugs.launchpad.net/juju/+bug/1622024 [01:28] Bug #1622024: destroy-unit help text mentions nonexistent remove-services [01:30] seyeongkim: if you have 2.0.x checked out, you can cherry-pick the commit [01:31] i see wallyworld , will try [01:31] seyeongkim: it's a rather cosmetic bug fix though. do you need the CLI help text changed urgently? [01:31] 2.1 will be out within a couple of weeks [01:32] no urgent [01:33] thanks wallyworld [01:33] sure, anytime [01:35] wallyworld: basically yeah. it works if you're the same $USER, because it'll pull the certs out of your $HOME/.config/lxc dir. but in their case, it's an entirely different $USER [01:35] wallyworld: I don't think we can support what they were doing without going backwards... [01:36] axw: should we consider a controller endpoint where cched creds in controller can be checked? [01:36] * wallyworld waves hands [01:36] wallyworld: sorry, don't understand [01:36] wallyworld: what creds, checked for what? [01:36] HO? may be easier to discuss [01:37] wallyworld: ok, in 5-10 mins; just preparing lunch [01:37] sure no rush [01:46] * redir eods [01:51] wallyworld: are you free now? 1:1? [01:51] yep [02:06] wallyworld: it will work, but only if they're using the 2.1-beta5 client. I suspect they're not [02:07] axw: good pickup, sounds like it can be closed as "use the current client" [02:07] wallyworld: it's not great that it will fail with the old client, but I think it's better than reverting? [02:14] axw: i think we can justify it (IMO). there's a work around we can document and it's not a super common use case i don't think [02:16] wallyworld: actually I think what they're doing is running from within a container, or some other machine on the same network; not on the host itself [02:17] wallyworld: so might need to just make "autoload-credentials" a bit easier [02:34] axw: yeah, autoload creds does need some polish in a few places [02:35] it was all rushed at last minute to get 2.0 out [02:35] axw: could you validate my choice of new skew value on this small PR? https://github.com/juju/juju/pull/6940 [02:38] wallyworld: seeing as this is only run on linux, can we just bite the bullet and use a monotonic clock? [02:38] axw: i didn't realise there was such a thing [02:39] wallyworld: man clock_gettime [02:39] axw: so how would that work clustered? [02:39] all machines need the same idea of time [02:39] ntp achieves that [02:40] but it needs make -ve corrections potentially [02:40] wallyworld: it's for measuring elapsed time [02:40] wallyworld: which IIANM is what we're doing in that code [02:40] elapsed time local only [02:41] ok, i'll take a look [02:48] wallyworld: https://github.com/davecheney/junk/blob/master/clock/clock_linux.go -- just need to include the LICENSE if you use that [02:49] axw: MonotonicClock "but is affected by the incremental adjustments performed by adjtime(3) and NTP" [02:50] i guess it only goes foreard though [02:52] wallyworld: http://stackoverflow.com/questions/14270300/what-is-the-difference-between-clock-monotonic-clock-monotonic-raw [02:52] explains it better than I could [02:53] axw: so it seems monotonic changes frequency to adjust for NTP and eventually syncs up [02:53] wallyworld: yep, but it won't jump around [02:54] it just adjusts rate of change [03:10] thumper: menn0: jam: tech board? [03:11] we are still dealing with the fallout from this prod issue [03:12] ok [03:54] axw: That was a nice link [03:55] FTR, I would be very nervous about juju attempting to go the CLOCK_MONOTONIC_RAW route, given some of the bad clocks I've seen around, esp. in Azure. [03:56] blahdeblah: yeah, I don't think it's appropriate for our use case [04:10] Bug #1662599 changed: Change maas api and password controller in model already created [04:13] Bug #1662599 opened: Change maas api and password controller in model already created [04:25] Bug #1662599 changed: Change maas api and password controller in model already created [06:04] Hi all, when logging into mongodb shell per https://github.com/juju/juju/wiki/Login-into-MongoDB, I get a number of recommendations: https://pastebin.canonical.com/178617/ [06:05] Is it worth creating a bug for that? [06:05] I wonder some of those things mightn't be contributing causes for our mongodb syncing issues (bug 1647238) [06:56] blahdeblah: axw and wallyworld were talking about disabling huge pages yesterday... maybe they can help? [07:00] anastasiamac: I believe wallyworld was talking about *hug* pages, and I was merely a an absent target ;p [07:01] axw: :) [07:16] axw: I can think of better things to hug, personally. :-) [07:17] axw: (Than pages, I mean. I'm sure you're a great huggee.) [07:17] blahdeblah: :p [08:19] Bug #1510787 changed: juju upgrade-charm errors when it shouldn't

[08:45] axw: Could you take a look at my openstack provider changes, pretty please? https://github.com/juju/juju/pull/6943 [08:45] babbageclunk: sure [08:45] axw: thanks! === frankban|afk is now known as frankban [08:49] Bug #1632909 changed: ceilometer-api fails to start on xenial-newton (maas + lxd)

[08:53] babbageclunk: don't worry about that VolumeSource thing too much. the API design is a bit broken, I will fix it at some point [08:53] what you have is fine [08:55] axw: great, thanks [09:11] axw: i've pushed up a juju/utils PR and also tweaked the juju PR for clock skew. got to fix a few tests and add deps change, but have to head out for a bit. i also need to live test etc but the basis should be there. the monotonic clock implementation doesn't return a true time.Time but it does work cross platform [09:12] axw: oh saw your comment - i just didn't want the hassle of 2 implementstions, or an implementation on linux and a no-op elsewhere [09:13] nanotine() is implemented cross platform in the run time isn't it? [09:13] wallyworld: I think so. but there's more than one compiler, and more than one runtime... [09:13] seems to be a popular approach to the problem [09:13] ok, i do the linux only thing then [09:14] wallyworld: never mind, it's probably academic [09:15] wallyworld: just leave it and we'll change it if necessary [09:16] ok [09:16] wallyworld: utils PR is approved [09:16] looking at other one now [09:17] ty, am late bbiab, will fix test setc then === StoneTable is now known as aisrael [10:31] Hi guys [10:32] So for a day I'm trying to make the oracle provider add custom cloud definitions. In order to communicate with the oracle rest api the client needs to specify the Identity Domain name and I'm trying to make the add-cloud command add this yaml field. [10:33] So I've tried to write the CloudSchema return the correct pointer to jsonsmecham.Smecha struct [10:34] https://paste.ubuntu.com/23953504/ [10:34] What I'm I missing here? [10:35] When I launch the add-cloud command it's printing me to add the identity domain and after the commands finished writing the clouds.yaml it does not add this attribute. [10:36] Also I found this https://github.com/juju/juju/blob/staging/cloud/clouds.go#L106 === frankban is now known as frankban|afk [10:37] Is there a way how to write in this attribute when the provider reads the clouds definition on bootstrap? [10:38] Is there anyone that knows or can point me into the right direction? [10:41] here is an example interacting with the add-cloud cmd [10:41] https://paste.ubuntu.com/23953549/ [10:42] sorry for the spelling issues. As you can see these does not write the identity-domain provided in the cmd interaction. [10:42] It is because the CloudSchema definition provided is somehow off? [10:46] Or there is now wat to do this and I should add some custom interactive code into the FinalizeCredential? [10:46] like in the azure provider? [11:37] Bug #1662857 opened: cannot go get the source code === cmars` is now known as cmars [11:55] Bug #1662857 changed: cannot go get the source code [11:58] Bug #1662857 opened: cannot go get the source code [12:28] morning [12:38] morning perrito666 [13:19] is there a problem when a normal machine and the controller machine are on different addresses? [13:19] im unable to ssh into a machine that is on a different ip-range using GCE as provider [13:20] with `juju ssh` that is, ssh:ing into the machine normally works fine [13:21] nvm, that wasnt the problem, now i cant ssh into a machine on the same range - really odd because other machines in the same model works fine [13:23] http://paste.ubuntu.com/23954247/ [13:23] any ideas? === frankban|afk is now known as frankban === stokachu_ is now known as stokachu [14:15] SimonKLB, cloud provider? [14:15] deanman: google [14:18] SimonKLB, can you access any app on that model ? [14:19] SimonKLB, i had almost a similar problem with openstack where additional units of the same app weren't assigned a public IP, which resulted only being able to ssh into only one at a time. [14:20] deanman: that shouldnt be the problem, i can see the public ips fine [14:20] deanman: not sure what you mean by accessing the app though? [14:21] SimonKLB, this was my issue, https://bugs.launchpad.net/juju/+bug/1655383, seems though unrelated to yours [14:21] Bug #1655383: When deploying multiple units of the same charm on top of openstack mitaka only one is assigned a floating IP [14:24] deanman: might be related, but in my case i all the machines get assigned a floating/external ip [14:24] im also able to ssh to them via the GCE console without any issue [14:24] SimonKLB, next thing i could think is that somehow "security groups" are different for machine #3 and #4 [14:25] could it be that #4 has different security groups and somehow the ssh port is blocked? Can you ssh from #3 to #4 ? [14:27] deanman: i have no idea why, but now, an hour later they are all accessable [14:27] havent changed a thing [14:27] have even worked in a different provider since then [14:27] maybe infra glitch ;-) [14:27] haha, let's hope so :) [14:27] i'll re-run the tests [14:39] Got a question for the devs, I'm running juju agent inside corporate network and i have to go through proxy. I'm using nc for most of my ssh needs but don't know how to make it work with juju so i can access my aws models. Any hints? [14:57] can anyone advice me, If I have a constrains let's say cpu-power and the provider does not support passing constraints like cpu how can I manage with the constraints.Validator methods this issue? I've already used the Unsupported methods but it seems this does not do anything to alert the client in someway if the constraint passed is valid or not for the provider [14:57] constraint* [14:57] can anyone provide some insight in this situation? [14:58] hoenir: I believe we are currently silent if a constraint is ignored with a provider [14:58] sinzui: you have tried all providers, what can you tell about this? [14:59] So informing the user with warningf is bad technique? [15:00] perrito666: your are correct, juju siltenly ignore constraints that are not relavant. [15:00] so it's ok if I inform the user? [15:00] of this? [15:00] in the bootstrap process [15:00] ? [15:00] with warningf? [15:01] hoenir: yes, absolutely [15:01] but we wont stop juju for it [15:01] yeah:) [15:02] just warn or even info about [15:02] thanks ! [15:02] I think warn is better suited for this situation [15:02] hoenir: mm, I think you are correct, if the user is dealing with constraints its most likely watching warn [15:03] sorry for the delay I was afk [15:03] thanks ! [15:07] smt like this. [15:07] https://paste.ubuntu.com/23954775/ [15:49] perrito666: if you get a chance, a review of https://github.com/juju/juju/pull/6944 [16:30] postponing the air conditioning in my office was the worse idea I had In my life [16:35] jam: ship it [16:36] jam: special kudos for the doc in NewLXDBroker [16:39] morning juju-dev o/ [16:40] redir: morning [16:44] morning redir [16:45] \o [17:48] Does Juju (2.1-beta5) support running lxd on a port other than 8443? === frankban is now known as frankban|afk [17:57] Specifically, I'm trying to use the snap to run beta5 alongside stable (via apt), and am hitting a bind conflict: http://pastebin.ubuntu.com/23955676/ === rogpeppe1 is now known as rogpeppe [19:31] cory_fu: sorry, I am unfamiliar with the lxd code. [20:03] morning people! [20:27] o/ [20:30] morning thumper [20:47] morning thumper [20:55] thumper: btw, I addressed your comments in https://github.com/juju/juju/pull/6763 and mentioned you to re check it, could you take a look when you are not working on urgent stuff? [20:55] * perrito666 knows he just wrote a blank check to thumper [20:56] perrito666: ok, will try to look shortly [20:56] * perrito666 is fighting to take his PR queue to 0 === frankban|afk is now known as frankban [22:31] perrito666, ping [22:31] alexisb: pong [22:32] release call [22:32] ah sorry was concentrated on a critical I just introduced [22:32] going === plars is now known as plars-away [22:49] anyone https://github.com/juju/juju/pull/6947 [23:23] Are old revisions of a resource cleaned up ever? I'm revving on a charm with a fairly sizable resource and after a couple of times updating the resource, the controller seems to run out disk [23:25] perrito666: if you're not EOD after you finish your HO can you please ping me about this migration issue. I have some thoughts. [23:28] menn0: certainly, I foresee that my EOD is far :) [23:29] menn0: I might ask you to HO , I am a bit sick of writing :p[ [23:32] perrito666: sure [23:32] perrito666: now [23:32] ? [23:33] menn0: this is the cause https://github.com/juju/juju/commit/49316b2 I am running this test to try to understand how this change caused this [23:33] menn0: no, still on a call [23:33] perrito666: let me know when you're off [23:34] menn0: I will [23:40] menn0: standup call? [23:40] perrito666: yep [23:40] cory_fu: there may have been an issue at one point.... what version of juju are you running? menn0 do you know if we fixed tht? [23:42] wallyworld: 2.0.2 I can re-test on 2.1-beta5 tomorrow if it's likely to have been fixed [23:43] cory_fu: I *think* so but will need confirmation as I didn't work on it myself [23:43] Ok. Thanks [23:43] menn0 is on a call, he will respond soon [23:45] wallyworld: reviewed your PR [23:45] ty [23:46] axw: end-beginning check should never fail now - i just thought defensively best to keep it? [23:47] axw: standup? [23:47] wallyworld: oops, coming [23:50] wallyworld: feel free to leave it, just my preference === frankban is now known as frankban|afk