[00:17] does anyone have familiarity with windows ? [00:17] I'm trying to get our windows CI machine to have go installed [00:17] as for some reason it seems to have lost it [00:17] I'm currently struggling with two things: [00:18] when I ssh from goodra the windows shell doesn't scroll, so after a page of output it just overwrites the last line [00:18] I managed to download the go 1.11.12 msi [00:18] but I can't seem to get it to install... [00:25] thumper: maybe this? https://powershellexplained.com/2016-10-21-powershell-installing-msi-files/ [00:25] in my frustration I rebooted the machine [00:25] I hope [00:27] hpidcock: let me try that... [00:31] hmm... nope, that didn't seem to help [00:31] man I hate windows [00:32] * thumper emails crew [00:46] thumper: were you trying to fix the RunUnittests-win2012 job? [00:47] yep [00:47] and failing [00:47] ok I'll check it out this morning [00:49] thanks hpidcock [02:59] hpidcock: thanks for fixing the go part. curious, what did you need to do? [03:00] I notice that mongo doesn't appear to be installed as all the tests failed [03:00] I just fixed mongo [03:00] but they ended up looking for /usr/local/bin/mongo [03:00] ah sweet [03:00] so what was the go problem? [03:00] was it actually installed and somehow out of the path? [03:01] thumper: for some reason it never installed, I had to use that guys powershell script to install it [03:01] probably would have been easier in the end to use RDP to do it via the gui [03:02] also to work around the ssh issue I did `mode con lines=9999` to fool the windows conhost.exe thinking the window height is 9999 [03:03] thumper: trivial https://github.com/juju/juju/pull/10505 [03:03] also had to fix up all the env vars, they were wrong and weird [03:05] babbageclunk: timClicks: how goes the vsphere cloud-init fun and games? [03:05] wallyworld: looking [03:06] have figured out how to customise the new instances' hostnames [03:06] testing to see if that's sufficient [03:07] timClicks: ok, good to unblock if it works, but it would be preferred to follow up with a bespoke cloud init per instance [03:08] wallyworld: thanks, lgtm [03:08] thumper: the race wasn't even in test code so it seems it could occur in production [03:08] * thumper nods [03:08] not sure of the effect etc [03:09] wallyworld: to be honest, unlikely to be a big problem [03:09] however, good catch [03:09] we'll need to be careful adding new entities if we also need to use a hashcache foro those [03:09] wallyworld, timClicks: how can the hostname impact the delivery of the full cloud init user data? [03:10] thumper: the import process and the vm cloning process have different args [03:10] thumper: it's more so that all cloned instances had the same hostname [03:10] because cloud init data was set once only in the template and cloned [03:11] hmm.. [03:11] ok [03:11] so setting hostname properly is necessary but not sufficient [03:11] still need to sort out cloud init properly [03:11] but it may unblock [03:11] looking for quick win to start with [03:12] small steps, just fix one thing that is feasible [03:13] thumper: i *think* andrew's approach also had the cloud init issue (not 100%) but the hostname was not an issue - that was introduced by the alternative workflow [03:20] thumper: do we really care about maas 1.9 now that trusty is esm? perhaps we can remove that ci test. we don't really have the means to properly run it anyway IIANM [03:21] No... I don't think we care about maas 1.9 any more [03:21] yay, one less red dot [03:21] we know that juju 2.3 and 2.4 work fine and can be a jump point if necessary [03:21] yup [03:22] we should remove the 1.9 code paths from juju for 3.0 [03:22] yep, on the metaphorical list [04:05] hey timClicks, how's it going? [04:17] babbageclunk: hey sorry, was outside [04:17] timClicks: no worries/same [05:53] wallyworld: approved [05:53] awesome ty [06:58] kelvinliu: trivial forward port from 2.6 if you are around [06:58] https://github.com/juju/juju/pull/10507 [06:59] yep looking [07:01] wallyworld: done! thanks! [07:01] ty