[01:06] <CarlFK> I want to test a dnsmasq config: server in one container, dhcp client in a 2nd, how do I ... do that?
[01:11] <CarlFK> macgreagoir:  here is how I am driving the ubuntu minimal:   one container for each host: https://github.com/xfxf/video-scripts/blob/master/carl/ansible-misc/slurp-hosts.py
[01:12] <CarlFK>         print('juju deploy {charm} {host}'.format(charm=charm, host=h))
[01:12] <CarlFK>         print('juju config {host} hostname={host}'.format(host=h))
[01:13] <CarlFK> and once they area all ready (about4 ;min)          cmd=['ssh', 'ubuntu@{host}'...."sudo", "cp", ".ssh/authorized_keys", "/root/.ssh"]
[01:14] <CarlFK> https://github.com/xfxf/video-scripts/blob/master/carl/ansible-misc/mk-hosts.py#L72
[01:14] <CarlFK> Is there a better way I should be doing that?
[04:33] <hygl> hi, i have installed a local kubernetes cluster using these instructions https://github.com/castrojo/kubernetes.github.io/blob/ubuntu-lifecycle-revisions/docs/getting-started-guides/ubuntu/local.md . How do I uninstall or stop it, if I do not need it anymore?
[08:15] <macgreagoir> CarlFK: You might be trying to configure the machine, where `config` means the application https://jujucharms.com/docs/2.0/charms-config.
[08:15] <macgreagoir> For ssh, this might help: https://jujucharms.com/docs/2.0/charms-working-with-units
[08:17] <macgreagoir> CarlFK: I'm not sure if bundles will help you, but be aware with scripts that Juju returns asynchronously, so you need to give the application time to deploy before you'll see the status you might expect.
[08:23] <CarlFK> macgreagoir: yeah, takes about 5 min for all the instances? to get an IP.  I just manually watch juju status and check every min or so
[08:36] <macgreagoir> CarlFK: I see a new release announcement for https://pythonhosted.org/juju/ on the juju list. I'm not sure it this could be interesting to you.
[10:37] <anrah> quick question about actions: Is it so that actions are not part of hookenv? I mean that if I want to use resources as part of action download etc. it seems like the get-resource command does not work
[10:39] <anrah> I wouldn't want the juju attach command straigthly to do stuff with update-config hook that but make separate action to apply the changes on nodes
[11:24] <stub> anrah: To my understanding, actions run in the same sort of context as hooks and the resource commands should work fine.
[11:26] <stub> anrah: You should even be able to test the resource commands using 'juju run --unit=foo "get-resource ..."', as that also uses the same context
[11:29] <anrah> I tried to run my command through debug-hooks
[11:29] <anrah> and that went fine
[11:29] <anrah> but when running the action through juju run-action <action> etc. results failed
[13:48] <petevg> bcsaller, cory_fu: I updated the crashdump PR with a small tweak (checking "is None" for env, just in case somebody wants to pass a blank dict in). Unfortunately, I also came across a bug: crashdump works fine in my isolated "test_crashdump.matrix", but it seems to be breaking in the context of the full default test :-/
[13:48] <petevg> Will dig into it after I finish doing morning routine stuff ...
[13:51] <stub> anrah: I'd need to see the error message from your logs to guess what is going on.
[14:30] <petevg> bcsaller, cory_fu: I figured out the problem. It's not fun :-( Basically, "juju ssh" doesn't work from the command line in the models that we've created. I've noticed that when manually troubleshooting. I wasn't running into the problem in my standalone test, because I was running crashdump right away, before machines had come up, rather than waiting for
[14:30] <petevg> things to deploy (that's what I get for writing tests that run fast to help meet deadlines ...).
[14:32] <petevg> bcsaller, cory_fu: it looks like "juju ssh" doesn't have the right credentials; I'm assuming that, when we create the model inside of python-libjuju, we're not handing something essential back to the system env ...
[15:02] <petevg> cory_fu, bcsaller: make a ticket (https://github.com/juju-solutions/matrix/issues/59).
[15:31] <cory_fu> petevg: Have you noticed that Matrix doesn't work on Juju 2.0.2 (current stable) because Controller.connect_current() doesn't work (though it does on the latest beta)?
[15:38] <cory_fu> petevg: Although, I just realized that might be due to me bootstrapping with 2.1-beta3 and then trying to interact with the 2.0.2 client.  I'm actually surprised it worked at all now
[15:44] <petevg> cory_fu: I hadn't noticed, though I am on the beta, with a controller created by the beta.
[15:58] <anrah> stub: got it working apparently i forgot to add the +x bit for the script before the build
[16:28] <cory_fu> petevg: You were asking yesterday about finding the name of the current controller with libjuju.  There's juju.client.connection.JujuData().current_controller(), but I should note that it doesn't seem to work with registered shared controllers
[16:28] <petevg> cory_fu: still useful, though. Thank you.
[16:28] <cory_fu> Because it seems that `juju show-controller` doesn't give any info if you've registered a shared controller
[16:28] <petevg> Interesting.
[16:32] <cory_fu> petevg: Oh!  show-controller requires admin perms
[16:32] <petevg> Of course it does :-)
[16:32] <petevg> it makes sense that it does, too.
[16:45] <cory_fu> petevg: https://github.com/juju/python-libjuju/pull/39
[16:46] <cory_fu> petevg: It doesn't make sense that it doesn't say anything and just returns an empty dict.  It could at least return as much info as list-controllers, or just complain that you're not a superuser
[16:47] <petevg> cory_fu: raising a sensible error in python-libjuju makes sense to me. +1 on that PR, based on code readthrough.
[17:12] <petevg> cory_fu, bcsaller: after some finagling, I think that https://github.com/juju-solutions/matrix/pull/56 is good to merge (it deliberately skips running juju-crashdump, because that is blocked by the bug with juju ssh, but it does successfully zip up the matrix logs, so that the CI can get at them).