[00:00] <kelvinliu> wallyworld: what do u mean rename env config, HO?
[00:01] <wallyworld> ok
[00:03] <babbageclunk> launchpad down?
[00:03] <wallyworld> yup
[00:04] <babbageclunk> I've got the page now, looks like it was just slow
[00:04] <babbageclunk> I mean, very slow
[00:04] <anastasiamac> yes.. very very slow
[00:10] <tlm> slow here also
[01:24] <wallyworld> kelvinliu: i merged your ppa packaging branch
[01:26] <kelvinliu> ty, so i probably doesn't have permission to merge.
[01:26] <wallyworld> maybe yeah, not 100% sure off hand
[01:56] <hpidcock> wallyworld: thanks for the approval on the 2.7 merge
[02:30]  * thumper screams into the ether
[02:30]  * thumper digs for old PR to put back something
[02:31] <wallyworld> tlm: would prefer not to export the k8s client struct. the pattern we use is to declare at the point of use an interface that represents the expose methods we want
[02:31] <wallyworld> make testing much cleaner and easier
[02:32] <wallyworld> reduces explicit coupling between layers
[02:32] <tlm> I have removed most of those couplings this morning for testing, may need to discuss this more though
[02:36] <wallyworld> ok, i'll wait for pr to be fully ready before poking my nose in
[02:37] <tlm> no feedback is great. I realised this morning that mistake. I still have the export there so may need a hand to remove it later this afternoon as there is three cases it's still solving
[05:30] <thumper> two PRs up for review: https://github.com/juju/juju/pull/11248 and https://github.com/juju/juju/pull/11247
[05:30] <thumper> I know they'll clash, but I'll deal with that
[05:30] <thumper> these are needed for new model summary watcher for dashboard
[05:30]  * thumper EODs
[08:50] <achilleasa> jam: thanks for the feedback. I will push extra commits to address your points
[09:40] <jam> achilleasa: sorry it took so long
[10:18] <stickupkid> fixing the issue where integration tests fail on exit
[11:11] <achilleasa> jam: pushed additional commits and answered some of the comments; can you please take another look?
[11:57] <achilleasa> can someone point me at the code that cleans up the agents/unit-XYZ-W folder once the unit is removed?
[14:25] <hml> Anyone know what Commands are in terms of the uniter resolving hooks?
[14:50] <stickupkid> manadart, rick_h got microstack working in a vm
[14:50] <stickupkid> muhahaha
[14:50] <manadart> stickupkid: Noice.
[14:50] <rick_h> stickupkid:  with vm provisioning?
[14:50] <rick_h> stickupkid:  the nested virt I was expecting to fail to get VMs in the microstack
[14:50] <stickupkid> rick_h, we'll soon see
[14:50] <rick_h> stickupkid:  ah ok gotcha
[14:51] <rick_h> stickupkid:  good stuff on progress
[14:51] <stickupkid> rick_h, going to launch a cirros image and see what happens
[14:52] <stickupkid> works for me
[14:52] <stickupkid> hey it's fast if you give it 18cores
[14:52] <rick_h> stickupkid:  lol
[14:53] <stickupkid> right i'm a bit shocked, wasn't expecting it to work... now how to make this usable
[14:54] <rick_h> stickupkid:  still might be worth keeping an eye on the snapshot idea (though maybe a lxd snapshot) to not have to wait for it to install/setup each time
[14:54] <rick_h> stickupkid:  but optimization I guess at that point
[15:01] <manadart> stickupkid: https://github.com/juju/juju/pull/11249
[15:04] <stickupkid> manadart, why not 2.7?
[15:05] <manadart> stickupkid: I goes on top of https://github.com/juju/juju/pull/11208.
[15:05] <stickupkid> manadart, fair
[16:05] <skay> ouch. I am grepping through the postgresql-charm src and cannot find where it sets any flagsother than postgresql.upgrade.systemd. where are db.master.available and db.connected set?
[16:13] <skay> oh, it's in the interface codebase
[16:49] <achilleasa> rick_h: Is the reboot detection logic meaningful for CAAS workloads? I think not but looking for +1s
[16:53] <rick_h> achilleasa:  no, since pods are expected to die/come back all the time not sure it's in juju's hands but more the k8s hands as the official scheduler of things
[16:53] <achilleasa> rick_h: my thoughts exactly.
[16:53] <rick_h> I'd think juju doing things would mess with k8s scheduler expectations
[17:15] <achilleasa> to answer my previous question, the agent data gets purged via a call to something that implements worker/deployer/deployer.go:47
[17:15] <achilleasa> of course it would be called "RecallUnit" to make searching for it harder :D
[17:17] <achilleasa> (the actual implementation lives in simple.go in the same package)
[17:21] <rick_h> lol "RecallUnit" that's new
[17:22] <rick_h> "come home unit, come home!"
[17:41] <achilleasa> stickupkid: is there a place to put common dependencies for workers?
[17:42] <achilleasa> this is for something that will be exposed via the dep engine but don't want to put in workers/
[17:43] <stickupkid> achilleasa, not sure tbh, not that I'm aware of
[17:47] <achilleasa> stickupkid: got a few min for a quick ho?
[17:47] <stickupkid> sure
[21:04] <anastasiamac> babbageclunk: there is a PR for juju-restore to stop/start agent
[21:45] <babbageclunk> anastasiamac: oops, missed that - looking now
[21:46] <anastasiamac> babbageclunk: \o/