2020/06/15/#smooth-operator.txt

[05:18] <MarkMaglana> jam: thanks for commenting on the Existing Tools section! Very good comments. Helps with thinking about and evaluating possible solutions.
[05:19] <jam> MarkMaglana, happy to. I think we're also very invested in recommendations for how to get the dependencies into the charm and into the deployment
[05:20] <jam> MarkMaglana, I should mention that the current plans are looking to have a build step for charms that can be run on a Launchpad build farm (like snaps and debs) so that charms end up series/architecture specific for deployment.
[05:20] <jam> We've tried to avoid that in Juju for a long time, but it is just a practical matter that charms depend on python extensions
[05:20] <jam> (even PyYAML uses libyaml)
[05:21] <jam> note that the charm store and Juju don't currently support looking up charms by series+architecture so a lot of that ends up in 'room for growth'
[05:33] <MarkMaglana> jam is this build step plan documented anywhere? I'm curious and would like to reference that in my document.
[07:33] <davigar15> Hey jam!
[07:34] <davigar15> Ian Booth told me that the issue https://bugs.launchpad.net/juju/+bug/1882127 is fixed in 2.7/edge
[07:35] <jam> davigar15, well, it now properly reports an error rather than a nil panic, we still need to understand what the underlying error is
[07:36] <davigar15> jam: okay! great! I will investigate if I have time. Let me know if you find out anything
[07:36] <davigar15> have a nice day :-)
[07:38] <jam> davigar15, you too
[08:13] <Chipaca> good morning people!
[08:52] <Chipaca> jam: should I tell people to give #323 a try if they're interested in it?
[08:52] <mup> PR #323: 317 state get <Created by jameinel> <https://github.com/canonical/operator/pull/323>
[08:52] <jam> It isn't wired up yet, so I still need to do a bit
[08:52] <jam> morning Chipaca
[08:52] <Chipaca> jam: ok, i'll just mention it's a WIP and on target for 0.7 then :)
[08:58] <vgrevtsev> hi jam - any chance you can take a one more look into the https://github.com/canonical/operator/issues/328? Still have no idea what could be wrong :/
[09:05] <jam> vgrevtsev, is src/charm.py executable and have the right #! line?
[09:06] <jam> maybe python isn't found at the path you expect inside the application container?
[09:07] <jam> replied to the bug
[09:11] <vgrevtsev> jam: yes it's exec-able, the charm itself is running fine
[09:15] <Chipaca> vgrevtsev: do you get debug logs?
[09:22] <jam> vgrevtsev, that is running in the operator pod, while the actions run in the application pod
[09:22] <jam> vgrevtsev, different python
[09:22] <Chipaca> jam: but the charm ran in the application pod as well, otherwise what created the symlink
[09:23] <vgrevtsev> jam: yeah but it didn't raise the bad interpreter or smth similar
[09:23] <jam> Chipaca, juju when it copies the charm directory from the operator pod to the application pod during init
[09:23] <Chipaca> ahhh
[09:24] <jam> vgrevtsev, $ ./foo
[09:24] <jam> bash: ./foo: /bin/ash: bad interpreter: No such file or directory
[09:24] <jam> that does say "No such file or directory"
[09:24] <vgrevtsev> but it also says "bad intepreter"
[09:24] <jam> vgrevtsev, yeah, and there is some sort of python that is trying to exec the script
[09:25] <vgrevtsev> well wait
[09:26] <vgrevtsev> i think i found smth
[09:26] <vgrevtsev> `/bin/sh: /usr/bin/env: not found`
[09:26] <vgrevtsev> (in the app container)
[09:26] <vgrevtsev> while my `charm.py` has `#!/usr/bin/env python3`
[09:34] <Chipaca> hehe
[09:34] <Chipaca> /usr/bin/env bites again
[09:34] <Chipaca> (this wrt: people doing '/usr/bin/env bash' to make things more portable --- despite env being in /usr/bin less frequently than bash is)
[09:35] <Chipaca> s/(?<=than bash is)/ in \/bin/
[09:36] <vgrevtsev> yeah, and here is another problem: my app container doesn't have python installed :\
[09:36] <vgrevtsev> so that basically means that I can't run actions against it
[09:36] <Chipaca> yeah
[09:37] <jam> vgrevtsev, you can write non-python actions, it depends what you want them to do
[09:43] <vgrevtsev> true
[09:43] <vgrevtsev> just wondering why the actions are executed at the app pod but charm itself runs on the operator pod?
[09:44] <Chipaca> vgrevtsev: I think there's an ongoing conversation about that
[09:44] <Chipaca> or a few actually?
[09:44] <Chipaca> seems a lot more work needs to be done still
[09:44] <Chipaca> mostly juju-side aiui
[09:45] <jam> vgrevtsev, a lot of it depends on what you want out of an action. many types of actions require running with the appliaction (eg, backup the database)
[10:32] <Chipaca> BTW, PROMPT_COMMAND='printf "⏎%$((COLUMNS-1))s\\r"' FTW
[10:55] <Chipaca> jam: in zookeeper_cluster, am i reading it right that it uses both framework.model.get_relation(...) and framework.relations[...] for the same relation, despite the usage of the first meaning that any checks on the length of the second are spurious?
[10:56] <Chipaca> because get_relation raises an exception if it's not unique
[11:01] <jam> Chipaca, meeting?
[11:59] <Chipaca> jam: another one?
[11:59] <Chipaca> oh wait
[12:00] <jam> Chipaca, https://github.com/canonical/operator/pull/327 needs your review
[12:00] <mup> PR #327: many: move Model.name to be read from _ModelBackend <Created by jameinel> <https://github.com/canonical/operator/pull/327>
[12:00]  * Chipaca gets out the red pen
[12:01] <jam> Chipaca, 'oh wait' ?
[12:01] <jam> Chipaca, if you can find a way to get your red pen marks online, I'm all for it
[12:01] <Chipaca> jam: saw your 'meeting?' message _after_ the meeting :)
[12:03] <Chipaca> stub: filed #329
[12:03] <mup> Issue #329: dependencies <enhancement> <🧠 big brain> <Created by chipaca> <https://github.com/canonical/operator/issues/329>
[12:03] <mup> Issue operator#329 opened: dependencies <enhancement> <🧠 big brain> <Created by chipaca> <https://github.com/canonical/operator/issues/329>
[12:04] <Chipaca> jam: i'd actually gotten half way through reviewing that on Friday
[12:04] <Chipaca> I lie
[12:04] <Chipaca> Thursday
[12:04] <jam> Chipaca, I can imagine it was "I reviewed it but didn't hit send"
[12:05] <Chipaca> i hadn't gotten to the bottom :)
[12:08] <mup> Issue operator#223 closed: Operator fails on Juju 2.7.0 and below <bug> <Created by chris-sanders> <Closed by chipaca> <https://github.com/canonical/operator/issues/223>
[12:08] <mup> PR operator#324 closed: model: don't use --app for relation-get/-set before 2.7.0 <Created by chipaca> <Merged by chipaca> <https://github.com/canonical/operator/pull/324>
[12:08] <mup> PR operator#326 closed: make abc TestMain private to help pytest <Created by chipaca> <Merged by chipaca> <https://github.com/canonical/operator/pull/326>
[12:26]  * Chipaca stepping out for 30'
[12:33] <mup> Issue operator#328 closed: k8s charm throws a "not found" error when invoking the action <Created by exceptorr> <Closed by exceptorr> <https://github.com/canonical/operator/issues/328>
[13:00] <Chipaca> jam: i'll be along as soon as i've figured out why i have a conflicting meeting
[13:00] <jam> ok
[13:02] <jam> Chipaca, so far nobody else is here
[16:37]  * Chipaca gives up
[16:37] <Chipaca> slightly early EOD from me
[16:38] <Chipaca> ttfn! might pop in later, but i'll go run now