[03:02] <veebers> wallyworld: to confirm, if we have a unit status of Blocked and a container status of Blocked, which wins? Show the unit blocked message of container?
[03:04] <wallyworld> veebers: if the container itself reports as blocked, we need to show that as that will surface the more relevant error. consider the gitlab charm case. gitlab will report blocked as db relation is missing, but gitlab pod may not be able to be started
[03:07] <veebers> wallyworld: ack. Makes sense
[04:04] <babbageclunk> wallyworld: hey, do I need to merge that change forward from 2.3 to 2.4, or will that get done periodically?
[04:04] <wallyworld> babbageclunk: only if someone does it; feel free to do it this time
[04:05] <babbageclunk> ok, so a full forward merge from 2.3 to 2.4, right?
[04:07] <babbageclunk> wallyworld: ^
[04:07] <wallyworld> yeah
[04:07] <wallyworld> sorry, otp, distracted
[04:29] <veebers> wallyworld: I'll need to revert my change to my caas charm where it sets active before setting podspec, as we never overwrite that unless the cloud container is error, blocked or allocating
[04:31] <babbageclunk> kelvinliu__: did you get anywhere with the flag problem?
[04:33] <kelvinliu__> babbageclunk, I manually chown the log files to solve the problem. but I am still solving a different error now
[04:33] <veebers> hmm, actually is this a different issue
[04:34] <kelvinliu__> babbageclunk, {"request-id":2,"error":"unknown version (2) of interface \"Agent\"","error-code":"not implemented","response":"'body redacted'"} Agent[""].GetEntities, seems the Agent facade is not registered correctly.
[04:35] <veebers> wallyworld: no, I think it's my crappy active message that has me wrong. Charm will set active, pod starts up, if it encounters errors they are displayed (and in  history) then once resolved the charm active status is used (including in history)
[04:35] <babbageclunk> kelvinliu__: I wouldn't have thought the log file would cause any problem - it tries to create and chown the file but doesn't worry if it fails.
[04:36] <babbageclunk> kelvinliu__: weird, that sounds like a mismatch between what the client and the server thinks that facade version should be.
[04:36] <babbageclunk> If you turn on trace-level logging for juju.apiserver, can you see the contents of the login response?
[04:36] <babbageclunk> That should show all of the registered versions of all facades.
[04:41] <kelvinliu__> babbageclunk, let me have a try,
[04:46] <anastasiamac> babbageclunk: do u have a chance to review this little gem? https://github.com/juju/juju/pull/9175
[04:46] <babbageclunk> anastasiamac: of course!
[04:46] <anastasiamac> babbageclunk: \o/
[04:49] <wallyworld> veebers: sorry, otp. seems like you got it sorted
[04:49] <veebers> aye, sorry for the noise
[05:00] <babbageclunk> anastasiamac: sorry, got distracted - reviewed!
[05:01] <babbageclunk> anastasiamac: did you tag the wrong PR for your trello task?
[05:03] <anastasiamac> thnx!
[05:03] <anastasiamac> and yes but untagged within seconds :(
[05:03] <babbageclunk> anastasiamac: do we normally need to get merge PRs reviewed? This one was small and pretty straightforward. https://github.com/juju/juju/pull/9181
[05:04] <anastasiamac> babbageclunk: no
[05:04] <babbageclunk> sweeeeet
[05:04] <anastasiamac> and fwiw i had a pcik at ur merge and +1
[05:04] <babbageclunk> Thanks!
[05:04] <kelvinliu__> babbageclunk, NewConnFacade.caller.BestFacadeVersion('Agent') --> 2         CallNotImplementedError -> &rpcreflect.CallNotImplementedError{RootMethod:"Agent", Version:2, Method:""}
[05:04] <kelvinliu__> , and we have only one v2 facade for Agent on apiserver   do u think what's wrong here
[05:04] <anastasiamac> babbageclunk: i have not clicked the button but will for ur peace of  mind
[05:05] <babbageclunk> anastasiamac: thanks! Then it's diffuse responsibility when it breaks everything,
[05:06] <anastasiamac> babbageclunk: certainly, especially in that universe where we held reviewers accountable as much as devs :)
[05:06] <babbageclunk> :D
[05:07] <anastasiamac> babbageclunk: funny. i did originally do everything in the loop but though it'd be cleaner to separate individual logical steps.. i'll revert
[05:10] <babbageclunk> anastasiamac: thanks!
[05:11] <anastasiamac> babbageclunk: nws, the difference of course is that u will always get an additional pass through the loop
[05:11] <anastasiamac> babbageclunk: so in most cases, instead of going once, u'll loop twice
[05:12] <anastasiamac> babbageclunk: since the loop will be do-while instead of while-do
[05:13] <babbageclunk> anastasiamac: I think that's ok, but you could just have an if-break in the middle if you wanted it to be the other way.
[05:14] <babbageclunk> kelvinliu__: sorry, I have to go for a bit, but I'll be back online later on
[05:14] <anastasiamac> babbageclunk: i could but in terms of lean loop, the ifs and other logic stmts, make it yuck...
[05:15] <kelvinliu__> babbageclunk, no worries. cu later
[05:15] <babbageclunk> anastasiamac: I'm not sure - it's not a big deal either way
[07:01] <babbageclunk> kelvinliu__: did you get that facade issue sorted out? want any help?
[07:03] <kelvinliu__> babbageclunk, just got some hints from Ian, im testing on it now. Thank you
[07:05] <babbageclunk> kelvinliu__: cool cool
[07:06] <kelvinliu__> babbageclunk, have a good one, cu tmr
[07:06] <babbageclunk> kelvinliu__: you too! :)
[07:08] <wallyworld> kelvinliu__: if you get a chance between testing, here's a small PR which fixes a critical k8s agent issue affecting CI and end users which I'd like to land https://github.com/juju/juju/pull/9182
[07:08] <kelvinliu__> wallyworld, looking now
[07:09] <wallyworld> ty
[07:22] <kelvinliu__> wallyworld, LGTM thanks
[07:22] <wallyworld> tyvm
[07:22] <wallyworld> how's the jujud? did you find any state errors?
[07:30] <kelvinliu__> wallyworld, found a mongo auth issue, looking on it now
[07:30] <wallyworld> kelvinliu__: ah good! at least that explains the error
[07:30] <kelvinliu__> wallyworld, yeah, thanks!
[09:03] <boritek> hello, after a machine restart "juju status" hangs, and the gui is not reachable
[09:03] <boritek> also the IP not pingable
[09:03] <boritek> how can I start it again?
[09:03] <boritek> it seems the controller is not running
[09:08] <stickupkid> if you've lost your controller and you no longer have access to it, then you need to bootstrap and start from scratch...
[09:13] <boritek> stickupkid: how can i lose it?
[09:13] <boritek> nothing special happened. is it not production ready?
[09:13] <stickupkid> boritek: which provider did you use, lxd, aws, azure, etc?
[09:14] <boritek> stickupkid: I configured my own maas
[09:15] <boritek> and maas-controller is on the same host where juju controller was
[09:15] <stickupkid> boritek: i don't know enough about setting maas up to be more helpful, some people will come on later today that will be able to help you more
[09:16] <boritek> stickupkid: does juju controller not run in a container?
[09:17] <boritek> how could I manually start it up?
[09:17] <boritek> is it in lxc, lxd ?
[11:32] <rick_h_> boritek: so the controller runs on top of hardware registered in MAAS
[11:32] <rick_h_> boritek: the controller runs on the cloud with the software you want to run/manage so that we know it can reach everything network-wise and such
[12:42] <rick_h_> TFW your print says it's out of paper and you can't recall the last time you put paper in it or where fresh paper might be...
[21:16] <veebers> so with the fixes for 1.11 going into 2.3 and 2.4 I suspect that we're looking at doing the next release with go 1.11?
[21:30] <rick_h_> veebers: just trying to test windows
[21:31] <rick_h_> veebers: we're not updating everything to 1.11. the issues fixed are valid 1.10
[21:31] <rick_h_> veebers: my understanding is 1.11 is a sprint topic
[21:36] <veebers> rick_h_: ack I see, thanks for confirming