/srv/irclogs.ubuntu.com/2020/07/07/#juju.txt

=== ec0_ is now known as ec0
stickupkidachilleasa, CR if you're around? fixes your concerns around reading stdin https://github.com/juju/juju/pull/1179609:32
achilleasastickupkid: looking09:34
achilleasastickupkid: got more questions10:38
stickupkidachilleasa, because it's expected to use the very annoying options in the command10:39
stickupkidachilleasa, that parseOneArg does more than you think10:40
stickupkidachilleasa, also your code - os.Stdin isn't an io.Reader10:43
achilleasait's not?10:45
achilleasait should be an *os.File10:45
stickupkidlet me try10:46
achilleasastickupkid: so, does the ignoreagentversion bit also work for stdin inputs? I am confused :D10:47
stickupkidachilleasa, works for everything10:47
stickupkidachilleasa, if you don't add the ignore flag you get this error message10:53
stickupkidjuju model-config --format=yaml | juju model-config -10:53
stickupkidERROR "agent-version"" must be set via "upgrade-model"10:53
stickupkidachilleasa, you free for a sec?10:58
achilleasastickupkid: sure; meet you in daily11:00
zeestratHey Juju folks, I have an existing model (v 2.5.8) with MAAS backend where I just added a new AZ and machines in that AZ but when I try to deploy new units it fails when provisioning as it does not recognise the new AZ and the machines in it. I tried `juju update-cloud <nameofmymaascloud>` but that didn't do the trick. How do I get Juju to recognise/update AZs from MAAS?11:41
stickupkidachilleasa, ^ that's reload spaces right?11:42
zeestratI presumed reload spaces only reload the network spaces from MAAS and not the AZ?11:43
achilleasastickupkid: I think that reload-spaces only refreshes subnets and spaces. Not sure about AZs though11:48
stickupkidyeah, that's why I asked, I'm not sure either11:48
manadart_zeestrat: We are working on independent modelling of AZs this cycle, but at present AZs are stored as part of subnet data, which is discovered during `reload-spaces`.11:48
manadart_zeestrat: If that does not work, it might be a MAAS provisioning bug.11:50
zeestratmanadart_: I see. I'll give it a try. Thank you. P.S Is there a way to inquire what AZs juju knows about?11:50
manadart_zeestrat: That is precisely the work we have on the table, including notification when a new one becomes available :)11:51
zeestratmanadart_: Great :) Any bugs or PRs I can subscribe to to follow along?11:51
zeestratmanadart_: Unfortunately it looks like `reload-spaces` didn't do the trick.12:01
manadart_zeestrat: Got some logs you can put in a pastebin?12:02
zeestratmanadart_: I tried scouring the machine logs, but not a lot to find. Here's what I have: https://paste.ubuntu.com/p/QhzdvNwBs7/ Let me know if there is something else I should look at12:14
manadart_zeestrat: I need to look at the (old ;)) 2.5 code to get the real work on this. But have you tried using placement instead of a constraint? I.e. `--to zone=x`.12:21
zeestratmanadart_: Much appreciated! Yes, I tried `juju deploy ubuntu ubuntu --to zone=hcc-dev-rack4` for example, but that just hangs on pending: https://paste.ubuntu.com/p/MDWhGt5tPh/12:31
manadart_zeestrat: Are you running HA?12:47
zeestratmanadart_: If you mean HA controllers, yes, 3x.12:48
manadart_zeestrat: OK, so what you need to do is restart each of the controller machine agents.12:49
manadart_zeestrat: This is a bug in the provisioner, which has a cache of distribution groups for distributing new machines across AZs.12:49
manadart_zeestrat: So as per the cycle work, we need to make it aware of the fact that new AZs can come online.12:50
manadart_I replicated your situation and restarting my controller fixed it.12:50
zeestratmanadart_: I can work with that :) Is the `reload-spaces` still required in addition to the restarts?12:52
manadart_zeestrat: No.12:52
manadart_zeestrat: If you create a bug for this, I'll triage it and add it to the feature task list.12:53
zeestratmanadart_: Gotcha, will do. I'll give it a spin in dev here just to check. Thank you very much for digging in! (And apologies for the old version. Had this one since 2.0 and we're slowly catching up)12:54
* manadart_ nods.12:55
zeestratmanadart_: Worked as a charm in dev, thanks again! For writing up the issue, is it a general provisioner issue or limited to HA and/or MAAS?13:15
manadart_zeestrat: Glad you're up. It's a general provisioner error.13:15
zeestratmanadart_: Created https://bugs.launchpad.net/juju/+bug/1886666 Please edit as you see fit.13:51
mupBug #1886666: juju 2.5.8 does not recognise new AZs created in cloud after deployment of controller <juju:New> <https://launchpad.net/bugs/1886666>13:51
manadart_zeestrat: Great, thanks.13:52
jamstickupkid, thanks for the review14:44
stickupkidjam, nps14:44
manadart_achilleasa: Got time to look at some more link-layer mechanics? https://github.com/juju/juju/pull/1180716:16
achilleasamanadart_: looking16:20
achilleasamanadart_: got a small nit about DocID(). Take a look at my comment and let me know what you think16:27
stickupkidhml, I updated my PR to rename charm to entity16:56
hmlstickupkid: ack16:56
stickupkidhml, but I left the json tag for the client as "charm" so it will still work with their API16:57
hml:-)16:57
stickupkidsee here https://github.com/juju/juju/pull/11793/commits/a0a9060221f382d83648a63427d513094114b543#diff-0c3b53c2399932098ef150096b7df011R1516:57
hmlstickupkid: i’ve started working on the review16:57
stickupkidi'm still unsure if we want an entity or a charm or bundle type.16:58
stickupkidbut for now let's do this16:58
hmlstickupkid: in my rework of where info convert is done… i used entity inside the apiserver, charm/bundle outside16:58
manadart_achilleasa: Replied to your comment. Have to head home now, so will look tomorrow.17:14

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!