| opendevreview | Rajat Dhasmana proposed openstack/devstack-plugin-ceph master: Add Ceph post-deployment health checks https://review.opendev.org/c/openstack/devstack-plugin-ceph/+/1009412 | 06:54 |
|---|---|---|
| *** gibi_ is now known as gibi | 08:11 | |
| tkajinam | hi it'd be nice if https://review.opendev.org/c/openstack/devstack/+/1009327 can get quick attention to unblock CI job update in python-cinderclient | 10:11 |
| frickler | tkajinam: I'm not sure whether that is a really good idea, you'll have tox running with py3.14, but all devstack services still use p3.13. likely you'll want to use resolute as nodeset for your python-cinderclient func job instead? | 10:26 |
| tkajinam | for that specific tests we only care about the version used to run tox (because it's testing behavior of python-cinderclient as a pure client) so that's "mismatch" is expected IMHO | 10:31 |
| tkajinam | we can eventually get rid of the problem by moving the job to resolute but we can face the same problem when we bump the upper job to py315 next | 10:31 |
| tkajinam | unless we get a handy new base os whose default python is 3.15 | 10:32 |
| tkajinam | frickler, ^^^ | 10:34 |
| tkajinam | frickler, maybe I can drop the version-specific functional tests from cinderclient following the job defs in the other clients such as novaclient | 10:43 |
| frickler | tkajinam: the TC only makes python versions mandatory if they are available as default version in a supported distro. in your case the mismatch may be expected and fine, but I'm worried about other consumers who might get confused. maybe let's wait for other reviewers then (cc gmaan) | 10:48 |
| tkajinam | ok | 10:49 |
| tkajinam | I've updated the cinderclient change to use unversioned job. if that's accepted then I'll abandon that job update, but also will wait for input from the others here, too | 10:49 |
| fungi | debian forky, some time next year, is expected to use python 3.15 by default | 15:35 |
| fungi | the 3.15 default transition is already underway in unstable/testing | 15:36 |
| clarkb | tkajinam: frickler: I think that handling that should be done outside of devstack | 15:48 |
| tkajinam | ok that's a good news | 15:49 |
| clarkb | devstack shouldn't be concerned about managing random python versions for specific test casese | 15:49 |
| clarkb | you can pretty trivially have your own jobs do that after devstack has deployed. I'm pretty sure we did/do that with zuul-launcher and nodepool tests and maybe even dib | 15:49 |
| tkajinam | clarkb, but the job is meant to run tox tests against devstack and IMHO that specific job should take care what is needed for tox | 15:49 |
| clarkb | tkajinam: yes but those tox tests have nothing to do with devstack | 15:49 |
| tkajinam | I totally agree with that for the base devstack job. | 15:50 |
| clarkb | you should treat devstack as an opaque box that givesyou a cloud | 15:50 |
| tkajinam | what I'm trying to fix is devstack-functional-tox | 15:50 |
| clarkb | then you can consume that cloud for your tests cases. disk image build does this as does zuul | 15:50 |
| clarkb | we basically run devstack. Then we do our test things as separate tasks. We don't rely on devstack to set up the test env for our test cases | 15:51 |
| tkajinam | no I'm not trying to change the python version used in devstack. it's totally fine that devstack, specifically speaking services deployed by devstack uses the default python | 15:51 |
| tkajinam | what I'd like to change is the version used to run tox command to run some tests against apis after devstack completes | 15:52 |
| clarkb | yes I know | 15:52 |
| clarkb | and I'm saying other people do that sort of thing all the time today by managing that externally beacuse devsatck shouldn't really care about your dependencies | 15:52 |
| clarkb | devstack should care about its dependencies | 15:52 |
| clarkb | I'm not completely opposed to this. I'm just trying to illustate that it isn't necessary and you can solev this without changing devstack and others already do | 15:53 |
| clarkb | we have a bad habit of putting everything into devstack whcih amkes it harder to maintain over time | 15:53 |
| tkajinam | I can override the job for sure but I don't really want to maintain a single ansible playbook replacing the one with just one line added there | 15:53 |
| tkajinam | then we should update it when the base playbook is updated in devstack side | 15:54 |
| tkajinam | It's totally acceptable though that the job shouldn't be used to use specific python versions in tox and that's how we should basically implement non-versioned tests (I mean the job using only the default version) | 15:55 |
| tkajinam | this is a kind of conflict between cinderclient CI and how devstack expects it to be used (at least currently) so we can fix one of these to proceed. | 15:56 |
| clarkb | I'm not sure devstack ever really had a strong opinion on how people show use it which lead to it being used for everything. I'm just suggesting that if your jobs need special casing like that having your own run.yaml or pre.yaml handle it might help make devstack easier to maintain long term. In this case I think the cost is somewhat low because that ensure-python role is | 15:57 |
| clarkb | pretty well behaved. So I'm speaking more to the general case | 15:57 |
| clarkb | where we've had a bad habit since forever of putting everything in devstack (or devstack-gate in the paste) | 15:57 |
| clarkb | s/paste/past/ | 15:57 |
| tkajinam | yes I understand that | 15:58 |
| tkajinam | point | 15:58 |
| clarkb | in my mind the utility of having ensure-python in devstack would be if we want to run devstack with non default python versions. Not for libraries that test against the cloud (rather tahn built into the cloud) to select a specific version | 15:58 |
| clarkb | I guess in this case there may be utility for having it in devstack if used for that case ^ and then it can be used for the other case too if everyone shares a common non default python | 15:59 |
| tkajinam | ok | 16:00 |
| clarkb | to be clear I'm not -1 on this. I just want to make the point of trying to be more disciplined about what we add to devstack | 16:00 |
| tkajinam | yeah I think that's basically same as the initial suggestion by frickler > that case | 16:01 |
| tkajinam | ok | 16:01 |
| clarkb | I'll leave a note that ensure-python should noop unless you set vars asking it to take action so it should be safe in the default case (this is what I meant by well behaved role aboev) | 16:04 |
| tkajinam | thx and agree | 16:06 |
| *** mhen_ is now known as mhen | 16:20 | |
| *** chEwBaCca is now known as AIromito | 17:10 | |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!