| jam | morning all | 06:41 |
|---|---|---|
| Chipaca | jam: morning! | 07:02 |
| jam | Chipaca, I didn't expect to see you here so early :) | 07:02 |
| Chipaca | jetlag :-| | 07:02 |
| Chipaca | (turns out, it was stress all along) | 07:02 |
| Chipaca | jam: working on the slides for our plenary now (sent you a link in the canonical irc) | 07:04 |
| Chipaca | probably going to step away and have breakfast and other non-sprint stuff in a bit | 07:04 |
| jam | Chipaca, sounds like a musical "it was stress, all a long" | 07:04 |
| Chipaca | doo be doo be dooo | 07:05 |
| jam | Chipaca, https://youtu.be/yOeUXEpxzcc?t=104 | 07:21 |
| niemeyer | Morning folks | 08:39 |
| facubatista | Muy buenos días a todos! | 11:39 |
| jam | morning facubatista | 11:47 |
| facubatista | hola jam | 11:47 |
| Chipaca | facubatista: EventSetBase?? | 11:53 |
| Chipaca | facubatista: you probably want to merge master into #233, i guess? | 11:53 |
| mup_ | PR #233: Execute the registered methods to events under PDB if proper envvar is set <Created by facundobatista> <https://github.com/canonical/operator/pull/233> | 11:53 |
| facubatista | oh, yes | 11:53 |
| facubatista | Chipaca, pushed | 11:56 |
| Chipaca | charmcraft#1 | 12:26 |
| mup_ | PR charmcraft#1: Draft to show several characteristics <Created by facundobatista> <https://github.com/canonical/charmcraft/pull/1> | 12:26 |
| Chipaca | niemeyer: dunno if mup follows renames; if not, can you update its conf to look at ^? | 12:27 |
| niemeyer | Chipaca: Will do, thanks for the note | 12:49 |
| axino | hi | 15:26 |
| axino | should I expect juju to fire a leader-elected hook when it first deploys a unit ? | 15:26 |
| axino | it apparently did in 2.7.3, and doesn't do anymore in 2.8 beta1 | 15:27 |
| niemeyer | axino: Sounds like a bug.. | 15:27 |
| niemeyer | axino: First unit is the leader in general | 15:27 |
| axino | (and I relied on that via self.framework.observe(self.on.leader_elected, self.configure_pod)) | 15:27 |
| niemeyer | axino: Is that you mean? The first unit of an app? | 15:27 |
| axino | niemeyer: yeah the first unit of an app | 15:27 |
| niemeyer | axino: It should work.. jam? | 15:27 |
| axino | juju deploy'ing a new app basically | 15:27 |
| axino | my charm worked in 2.7.3, but not with 2.8 beta1 | 15:28 |
| jam | axino I would expect leader-elected to fire, if something has changed there I would definitely file a bug w/ Juju | 15:28 |
| axino | jam: ack, I will | 15:28 |
| niemeyer | jam: Sounds worth looking deeper into it with axino.. it's a bug worth catching if something went wrong there | 15:29 |
| jam | axino, I would expect relation-created to fire before leader-elected, in case you were looking for an exact order of hooks | 15:29 |
| jam | and relation-created is new in 2.8 | 15:29 |
| axino | jam: nah, no relation here, this is a very basic charm without relation | 15:29 |
| jam | axino, link and context in a bug is very much appreciated | 15:29 |
| axino | jam: I know I know | 15:30 |
| jam | axino, you've always done well, just helps to remind sometimes :) | 15:30 |
| axino | :) np | 15:30 |
| axino | jam: https://bugs.launchpad.net/juju/+bug/1875675 I think this is "just" hook ordering | 15:45 |
| jam | axino, for juju 2.7.? (I think .6) and 2.8 Juju will now fire the Install hook, which is a point where the initial handlers can be registered. | 15:46 |
| Chipaca | jam, facubatista, niemeyer, i think i'm going to cancel the wordpress k8s review session | 15:47 |
| jam | I think firing leader-elected before 'start' is more expected behavior, as it is how non-k8s has worked for a long time | 15:47 |
| facubatista | Chipaca, or move it | 15:47 |
| jam | Chipaca, k. I was going to ask for a bit of a breather, but doing the review still seems worthwhile | 15:47 |
| Chipaca | moved! | 15:47 |
| axino | jam: hook ordering changed between 2.7.3 and 2.8 beta1 | 15:47 |
| axino | jam: at least one of leader-elected or config-changed (or upgrade-charm) used to fire after start, and now fire before | 15:48 |
| jam | axino, non-k8s charms have always fired install, leader-elected, config-changed, start | 15:48 |
| jam | I think 2.7.3 to 2.8b1 is making k8s act more like the rest of the system | 15:48 |
| axino | jam: so should k8s charm use the "install" hook to "bootstrap" themselves ? | 15:49 |
| jam | axino, however, #juju is better place to discuss this. | 15:49 |
| jam | axino, yes 'install' should be used if it fires. | 15:49 |
| axino | ok | 15:49 |
| niemeyer | Chipaca: Thanks | 15:49 |
| jam | axino, operator framework itself has a even for 'install' to do so | 15:49 |
| niemeyer | Chipaca: The deck review today is on the late side.. can we put it inside the sprint window today or tomorrow? | 15:50 |
| axino | jam: I think we have slides somewhere that lead us to use "start" | 15:50 |
| Chipaca | niemeyer: I didn't realise you were invited to that | 15:50 |
| Chipaca | niemeyer: that might be an oversight :) | 15:50 |
| niemeyer | I'm surprised you think I wasn't :) | 15:50 |
| Chipaca | niemeyer: i thought it was just me and sohini | 15:51 |
| axino | jam: this appears to be https://bugs.launchpad.net/juju/+bug/1854635 ! | 15:53 |
| niemeyer | Chipaca: Can we get it moved inside the sprint window? | 15:54 |
| * facubatista -> lunch | 16:08 | |
| Chipaca | niemeyer: trying to figure it out, there might be some confusion, sorry | 16:27 |
| crodriguez | Can I get some advice about debugging k8s models with the operator framework? I'm not sure what is causing my charm to crash on the install hook, I only see this "application-mssql: 11:45:32 ERROR juju.worker.uniter.operation hook "install" (via explicit, bespoke hook script) failed: exit status 1" in the juju debug-log, and I can't really debug with kubectl commands because the pods do not come up since it | 16:51 |
| crodriguez | fails on install | 16:51 |
| crodriguez | I looked this up https://discourse.juju.is/t/live-debugging-at-python-level/2914/4 but that's not provided on k8s models | 16:52 |
| Chipaca | crodriguez: what juju version are you running? | 16:54 |
| crodriguez | Chipaca: 2.8.beta-1 | 16:56 |
| Chipaca | jam: did the change to logs land in that? | 16:56 |
| jam | crodriguez, I would imagine the operator pod is running, if you wanted to connect to it, since it isn't part of the application pod | 16:57 |
| jam | Chipaca, the debug level changes did land in 2.8.beta1 | 16:58 |
| jam | crodriguez, so if you're using the latest operator framework, you should be able to set "juju model-config logging-config=DEBUG" and juju should show you more information about what was going wrong | 16:58 |
| crodriguez | okay I'll try that thank you | 16:59 |
| Chipaca | crodriguez: the if at the beginning of jam's statement is important :-D | 17:01 |
| Chipaca | latest as in master from 6 days ago or newer | 17:01 |
| jam | Chipaca, taking the dog out on a walk during the lunch break, bbiab | 17:02 |
| Chipaca | jam: 👋 | 17:02 |
| crodriguez | lol it changes quickly uh! I'll update that too... Idk, I was deploying the same piece of code a month ago and it wasn't failing on install. I wonder if something in my code might not be compatible with an update or juju and/or framework | 17:03 |
| jam | crodriguez, juju <=2.7.5 didn't run install | 17:04 |
| Chipaca | niemeyer: moved | 17:05 |
| facubatista | jam, may you know why/if "juju debug-code is not supported on kubernetes models"? https://discourse.juju.is/t/live-debugging-at-python-level/2914/5 | 17:28 |
| jam | facubatista, I believe SSH into an operator container isn't supported, which is what juju debug-hooks and debug-code use to get into the machine | 17:36 |
| jam | facubatista, it seems like it could be implemented on top of something like kubectl exec, but it is a matter of implementing support for it | 17:37 |
| crodriguez | I love the new debug logs <3 | 18:05 |
| jam | crodriguez, did you work out the problem? | 18:14 |
| crodriguez | yeah, it showed that my submodule wasn't found, so the import of ops.whatever was failing. I reinitialized it, updated it to latest, and I think I'm back on track now. Thanks! | 18:15 |
| facubatista | crodriguez, this is related to your comment here? https://discourse.juju.is/t/live-debugging-at-python-level/2914/5 | 18:31 |
| crodriguez | facubatista: I posted it when I was blocked earlier, yes, but it is more of a general inquiry. I feel that we're not as well equipped to debug with juju for k8s models compared to openstack/vmware/baremetal. | 18:33 |
| mup_ | PR operator#245 closed: use path.open() instead of open(str(path)) <Created by chipaca> <Merged by chipaca> <https://github.com/canonical/operator/pull/245> | 18:35 |
| Chipaca | niemeyer: wrt plastic in your mate, don't google "food defect action levels" | 18:39 |
| Chipaca | e.g. chocolate is fine until there is an average of 60 or more insect fragments per 100 grams | 18:39 |
| * Chipaca omnoms chocolate | 18:40 | |
| niemeyer | Chipaca: Ouch.. let me guess.. FDA doesn't consider plastic to be an issue | 18:40 |
| niemeyer | Okay.. I did Google.. apparently it's much worse | 18:42 |
| facubatista | crodriguez, we will improve there! glad to know you're not blocked | 18:47 |
| crodriguez | facubatista: I'm sure! :) | 18:57 |
| mup_ | PR operator#233 closed: Execute the registered methods to events under PDB if proper envvar is set <Created by facundobatista> <Merged by facundobatista> <https://github.com/canonical/operator/pull/233> | 19:24 |
| jam | good night all | 20:08 |
| Chipaca | EOD here as well | 20:11 |
| Chipaca | 👋 | 20:11 |
| facubatista | bye jam | 20:13 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!