| opendevreview | Katarina Strenkova proposed openstack/tempest master: Fix cleanup of keypairs for --prefix option https://review.opendev.org/c/openstack/tempest/+/924919 | 08:52 |
|---|---|---|
| opendevreview | Karolina Kula proposed openstack/devstack master: Replacing usage of rdo-release rpm https://review.opendev.org/c/openstack/devstack/+/925368 | 10:36 |
| opendevreview | Stephen Finucane proposed openstack/hacking master: Add missing Python classifiers https://review.opendev.org/c/openstack/hacking/+/925379 | 11:27 |
| opendevreview | Stephen Finucane proposed openstack/hacking master: Bump flake8 version https://review.opendev.org/c/openstack/hacking/+/925380 | 11:27 |
| opendevreview | Alfredo Moralejo proposed openstack/devstack stable/2024.1: Replacing usage of rdo-release rpm https://review.opendev.org/c/openstack/devstack/+/925400 | 14:08 |
| dansmith | gmann: I really want to be able to validate image formats against glance (and maybe nova) to make sure we're accepting and rejecting the formats we should | 16:34 |
| dansmith | it seems like tempest should be able to do that in most cases. is that reasonable to expect to add that? | 16:34 |
| dansmith | if not, I can do it maybe in whitebox or something | 16:34 |
| dansmith | so much of the confusion around these CVE fixes came from things like iso being not tested anywhere | 16:35 |
| dansmith | so I've got this now: https://opendev.org/openstack/os-test-images | 16:36 |
| dansmith | which has a manifest explaining which images should be usable and not and I'd really like tempest to be able to fetch those from somewhere and throw them at glance to confirm that they are accepted or rejected as expected | 16:36 |
| opendevreview | Rodolfo Alonso proposed openstack/devstack master: Move the check of "rpc_workers" after the post-config phase https://review.opendev.org/c/openstack/devstack/+/925412 | 16:51 |
| gmann | clarkb: ack, will check. thanks | 17:43 |
| gmann | dansmith: if I am not wrong the accepted/rejected format will be known via upload API only right? | 17:43 |
| dansmith | gmann: not sure what you mean | 17:44 |
| dansmith | after changes to glance, if we say an image is vmdk but we upload a qcow2, we should get a fail | 17:44 |
| gmann | dansmith: I mean how you will get to know from glance if format is accepted or rejected? | 17:44 |
| dansmith | if we say it's a vmdk and upload a vmdk but the wrong subformat, it should fail | 17:44 |
| dansmith | if we upload a qcow2 but with a data-file set, it should fail | 17:44 |
| gmann | yeah, then it is good to add in tempest itself | 17:44 |
| dansmith | okay cool, | 17:45 |
| dansmith | so the options for that are: | 17:45 |
| dansmith | 1. Fetch the sample images from http somewhere, maybe during some setup phase? | 17:45 |
| dansmith | 2. use the repo above to generate them locally | 17:45 |
| dansmith | to be clear, #2 could be a devstack thing that drops them somewhere and tells tempest where to find them locally | 17:46 |
| gmann | dansmith: yeah, it make sense to do #2 in devstack. this lgtm overall | 17:48 |
| dansmith | gmann: okay so for non-tempest people, it will be something like "point tempest.conf to a local directory with manifest.yaml in it" right? | 17:49 |
| dansmith | sorry, | 17:49 |
| dansmith | non-devstack people I mean | 17:49 |
| dansmith | i.e. cloud ops that want to validate that all the CVE images are rejected after they apply patches | 17:50 |
| gmann | dansmith: yes, config can hold the place same way we do for cirros image also | 17:50 |
| dansmith | okay | 17:51 |
| dansmith | if we do it this way, then it would be nice if a change to add a new check could depends-on a change to os-test-images as well | 17:51 |
| gmann | bcz test generating make test long as well as not configurable if other generated images needs to be tested | 17:51 |
| dansmith | I don't know what is involved in making that linkage work though | 17:52 |
| gmann | for depends-on things, we still need to add os-test-images as required-project | 17:52 |
| dansmith | gmann: the generate step in os-test-images is very fast, but makes sense to do it in devstack as part of tempest setup or something | 17:53 |
| gmann | k | 17:53 |
| gmann | I am thinking if it is ok to add os-test-images as required-project in base job if you want to run it in all jobs | 17:53 |
| gmann | if just single job is fine then adding it in tempest-full-py3 only make sense | 17:54 |
| dansmith | idk, probably no reason to make neutron run these | 17:54 |
| gmann | yeah | 17:54 |
| dansmith | nova, glance, cinder in various configs, probably but dunno that others need to really | 17:54 |
| opendevreview | Stephen Finucane proposed openstack/hacking master: Ignore SyntaxError exceptions https://review.opendev.org/c/openstack/hacking/+/925418 | 17:54 |
| gmann | I think doing it in those jobs only make sense. tempest-full-py3 and nova, glance, cinder integrated jobs | 17:55 |
| stephenfin | gmann: frickler: Workaround for an interesting little flake8 bug there (plus an earlier flake8 bump patch), if you have time over the rest of this week ^ | 17:55 |
| dansmith | gmann: ack, well, let me figure out the mechanics and then we can apply to jobs | 17:55 |
| gmann | ++ | 17:56 |
| dansmith | gmann: how does devstack (or that job) know to use a gerrit review instead of opendev for a git repo if depends-on is used? | 17:56 |
| gmann | dansmith: devstack clone those repo from source in playbook pre run i think | 17:57 |
| gmann | dansmith: here -> and all required-projects is appended in LIBS_FROM_GIT by default https://github.com/openstack/devstack/blob/master/roles/write-devstack-local-conf/library/devstack_local_conf.py#L260 | 17:58 |
| dansmith | ah okay | 17:59 |
| dansmith | so that puts a thing in libs_from_git, but..we must get the url from depends-on somewhere | 18:00 |
| dansmith | but, if libs_from_git is the devstack fu required maybe that's all I need to know... | 18:01 |
| gmann | stephenfin: ack, will check | 18:01 |
| gmann | dansmith: ohk, I need to check if that is in devstack or zuul does that? | 18:02 |
| dansmith | yeah, idk | 18:03 |
| gmann | I think its zuul when it prepare the repo clone. I am sure it is not in devstack | 18:04 |
| dansmith | okay but devstack needs to know "get this from /opt/stack/$proj instead of http://opendev.org/openstack/$proj" right? | 18:06 |
| dansmith | or /home/zuul or wherever they go | 18:06 |
| gmann | I think the way it handle is /opt/stack/$proj itself will be cloned from the master (if it is required project) and then checkout depends-on as head | 18:08 |
| gmann | and then proj get installed from there | 18:08 |
| gmann | and checkout to depends-on is done by zuul | 18:09 |
| dansmith | after devstack has done the clone to /opt/stack/$proj? I figured that was done before devstack even starts | 18:10 |
| gmann | dansmith: it is done before only but by zuul in one of it roles which prepare/clone all the required repo. This is one example I was just checking when oslo.policy change as depends-on was cloned with depends-On commit id and then devstack installed it from there https://zuul.opendev.org/t/openstack/build/c77d94815a35426b82136ede32bf453c/log/job-output.txt#1578 | 18:14 |
| dansmith | okay and it does that if required_projects:$proj? | 18:16 |
| gmann | so zuul keep repo clone ready as per depends-on/required-projects etc | 18:16 |
| gmann | dansmith: yes. if only in required-project otherwise it is not even installed from git/master | 18:17 |
| gmann | it will be released version installed and no checkout etc | 18:17 |
| dansmith | okay, so devstack's git_clone must just run `git checkout` if the repo already exists? | 18:17 |
| gmann | yes | 18:18 |
| dansmith | okay cool, I expected it was going to be more complicated than that | 18:19 |
| gmann | only key thing is if you want to test os-test-images as released version (if you are planning to release it) or master | 18:20 |
| dansmith | not planning to release, so I think we can always just clone from master (or branch if zuul checks out one for us) | 18:21 |
| gmann | but both can be tested like we do for olso. released version everywhere but master version in *oslo-master job | 18:21 |
| gmann | ok | 18:21 |
| opendevreview | Dan Smith proposed openstack/devstack master: Add os-test-images support in lib/tempest https://review.opendev.org/c/openstack/devstack/+/925425 | 19:05 |
| opendevreview | Merged openstack/hacking master: Ignore SyntaxError exceptions https://review.opendev.org/c/openstack/hacking/+/925418 | 19:33 |
| opendevreview | Dan Smith proposed openstack/tempest master: WIP: Test image formats https://review.opendev.org/c/openstack/tempest/+/925431 | 19:48 |
| dansmith | gmann: we'll see how this does ^ | 19:48 |
| dansmith | works for me locally and confirms that glance currently rejects the gpt image as expected | 19:48 |
| dansmith | glance doesn't do nearly enough of the checking it should, so there's not a ton of those tests we can write today, except maybe as xfail | 19:49 |
| dansmith | gmann: I'll have to go hunt for this I guess: https://zuul.opendev.org/t/openstack/build/f99cf0dd6bf24eaeb03cb33a718a536b | 20:43 |
| dansmith | but fairly close I think | 20:43 |
| opendevreview | Merged openstack/tempest master: Fix cleanup of keypairs for --prefix option https://review.opendev.org/c/openstack/tempest/+/924919 | 22:29 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!