| *** yadnesh4 is now known as yadnesh | 08:57 | |
| *** tonyb_ is now known as tonyb | 08:57 | |
| *** zxiiro_ is now known as zxiiro | 08:57 | |
| *** toabctl_ is now known as toabctl | 08:57 | |
| *** erbarr_ is now known as erbarr | 08:57 | |
| *** ricolin_ is now known as ricolin | 08:57 | |
| *** mnasiadka_ is now known as mnasiadka | 08:57 | |
| *** simondodsley_ is now known as simondodsley | 08:57 | |
| *** clarkb is now known as Guest5393 | 08:59 | |
| *** diablo_rojo_phone is now known as Guest5405 | 08:59 | |
| *** clarkb is now known as Guest5418 | 10:45 | |
| opendevreview | Stephen Finucane proposed openstack/pbr master: Add support for PEP-660 https://review.opendev.org/c/openstack/pbr/+/899848 | 11:18 |
|---|---|---|
| opendevreview | Stephen Finucane proposed openstack/pbr master: build: Use kwargs https://review.opendev.org/c/openstack/pbr/+/899849 | 11:18 |
| *** dansmith_ is now known as dansmith | 13:40 | |
| opendevreview | Stephen Finucane proposed openstack/pbr master: Update integration test https://review.opendev.org/c/openstack/pbr/+/899862 | 13:58 |
| opendevreview | Stephen Finucane proposed openstack/pbr master: Add support for PEP-660 https://review.opendev.org/c/openstack/pbr/+/899848 | 14:01 |
| opendevreview | Stephen Finucane proposed openstack/pbr master: build: Use kwargs https://review.opendev.org/c/openstack/pbr/+/899849 | 14:01 |
| *** Guest5418 is now known as clarkb | 15:03 | |
| clarkb | sfinucan: fungi: I left a comment on 899848 indicating why I think what is done there is sufficient for old python. Something we should be careful around, but also something that our testing should provide sufficient coverage for. | 15:12 |
| clarkb | I think the tempest-full issue might be that there is a python3 specific tempest full job? I'm still digging into that problem | 15:25 |
| clarkb | oh! it is specifically installing stable/train | 15:26 |
| fungi | but it's not installing stable/train of neutron for some reason | 15:26 |
| fungi | it's installing master | 15:26 |
| clarkb | which was the last python2.7 proper release? And now is EOL? | 15:26 |
| clarkb | fungi: ya because the branch doesn't exist anymore and the fallback for a missing branch is to use the default branch | 15:27 |
| clarkb | in this case master | 15:27 |
| fungi | aha! it's branch deletion, yep. that needs to install neutron from tag | 15:27 |
| clarkb | ya I think override-checkout can take a table that specifies refs for each project | 15:28 |
| clarkb | in this case it looks like nova and neutron at least need to be from train-eol but other projects like glance and keystone still have the branch | 15:28 |
| clarkb | or we can stop trying to test python2.7 integration all together and rely on unittests | 15:30 |
| clarkb | I'm working on a change | 15:31 |
| fungi | thanks! | 15:33 |
| opendevreview | Clark Boylan proposed openstack/pbr master: Fix stable/train testing with tag checkouts https://review.opendev.org/c/openstack/pbr/+/899881 | 15:40 |
| clarkb | something like that maybe | 15:40 |
| *** sfinucan is now known as stephenfin | 16:15 | |
| clarkb | if ^ that works we may need to squash it with 899862 to make things mergeable | 16:18 |
| fungi | i expect so, yes | 16:18 |
| fungi | i don't think either will pass on their own | 16:18 |
| frickler | clarkb: I collected some data regarding rsync vs find+mv for the /opt stuff at https://etherpad.opendev.org/p/uJl4PI8M3AIbecv74i0m to me it really looks like there is no significant difference | 16:42 |
| clarkb | frickler: ya I think that was more hopeful than anything else. Thank you for confirming. I think we can abandon both of those changes | 16:44 |
| clarkb | frickler: do you want to do that since you collected the data? | 16:44 |
| frickler | clarkb: ack. what about https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/865383, do we still want that in? | 16:45 |
| clarkb | frickler: probably need to get results from it to see how much it pollutes the logs | 16:46 |
| clarkb | my concern with that one is it may be a ton of files because git | 16:47 |
| frickler | clarkb: well this is what the logs in the etherpad all contain. not much logging, just a couple of line with stats | 16:48 |
| *** promethe- is now known as prometheanfire | 16:49 | |
| *** prometheanfire is now known as Guest5452 | 16:49 | |
| *** Guest5452 is now known as prometheanfire | 16:51 | |
| clarkb | oh I see its ansible's representation of the copies not a massive wall of text | 16:52 |
| clarkb | frickler: I'd say we keep it in that case as it may help us better understand the /opt copies and their cost giving us hints to other improvements | 16:53 |
| clarkb | frickler: I +2'd it. Feel free to convert your -W to a +A if you agree | 16:54 |
| frickler | ack | 16:59 |
| opendevreview | Merged openstack/openstack-zuul-jobs master: Add some output to the `find' command https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/865383 | 17:07 |
| clarkb | the tempest-full job for pbr got far enough to run tempest. If that passes I'll squash it into stephenfin's change and get things rebased on top of that hopefully making the whole stack landable | 17:11 |
| clarkb | frickler: sean-k-mooney that pbr test fixup for train seems to mostly work except for two cirros kernel panics: https://storage.bhs.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_d85/899881/1/check/tempest-full/d858a97/testr_results.html | 17:46 |
| clarkb | any idea what is going on there? Maybe we have reach the point where a better fix is to drop the python2.7 integration testing and rely on unittests instead | 17:46 |
| frickler | clarkb: iiuc it is a known instability with older cirros versions. for that particular use case one could try to look into disabling the affected tests. or only unit tests, yes | 17:48 |
| clarkb | frickler: reading https://lists.openstack.org/pipermail/openstack-discuss/2020-December/019250.html sean-k-mooney seems to think it is general kernel instablity for older kernels. I worry that we'd turn off these tests and then it will fail on some other tests. But let me recheck to see if this is a consistent set of tests that fail | 17:50 |
| clarkb | if so I guess we can start there | 17:50 |
| sean-k-mooney | is this cirros 5.x acpi kernel crash? | 17:56 |
| sean-k-mooney | end Kernel panic - not syncing: IO-APIC + timer doesn't work! Boot with apic=debug and send a report. Then try booting with the 'noapic' option. | 17:56 |
| sean-k-mooney | frickler: clarkb so that should not happen on 6.x | 17:56 |
| sean-k-mooney | but that was a patched kernel bug in ubuntu i tjust was never pulled into th 0.5.x seriese in cirros | 17:57 |
| clarkb | sean-k-mooney: no this is stable/train era cirros whatever that is | 17:57 |
| sean-k-mooney | as in later ubuntu 18.04 lts kernels had the fix | 17:57 |
| clarkb | sean-k-mooney: the background on the job is that PBR has had a python2.7 tempest job as a sanity check that PBR updates don't break old openstacks since you can't really constraint PBR (until pyproject.toml came along) | 17:58 |
| clarkb | if the recheck is clean then maybe it doesn't occur often enough to be a problem for PBR. If it does recur in the same tests we can exclude the tests. If it happens in other tempest tests I'm going to drop the test entirely | 17:58 |
| sean-k-mooney | so i have been noticing a really weried behaivor with the pyproject.toml patches | 17:59 |
| sean-k-mooney | only the tox py39 jobs are treating the lack fo editiable support as an error | 17:59 |
| sean-k-mooney | https://zuul.opendev.org/t/openstack/build/9a3790cdc70a4c36b543bd002f38acf0/log/job-output.txt#1566 | 18:00 |
| clarkb | sean-k-mooney: when I see tox errors I can't help myself: does nox fail similarly :) Part of the problem I have with tox is that it implements its own pep 517/660 implementation so when this stuff fails in tox you are kinda at their mercy | 18:01 |
| clarkb | they interpret those peps far more strictly than `build` and `setuptools` and `pip` do as well | 18:01 |
| sean-k-mooney | ya so i realsied htat i cant use depend on to test https://review.opendev.org/c/openstack/pbr/+/899848 | 18:01 |
| sean-k-mooney | with the nova change | 18:02 |
| sean-k-mooney | but i was going to manually install that and see if it help | 18:02 |
| clarkb | sean-k-mooney: I think the reason py39 jobs fail is taht they are running tox under python3.8 on that platform (focal?) and tox must have python-requires that prevent latest from installing there? | 18:02 |
| sean-k-mooney | i have been busy with other thigns today so havent tried it but i have a centos 9 host i can check this on quickly | 18:02 |
| clarkb | hrm no latest tox is python>=3.8 | 18:03 |
| sean-k-mooney | clarkb: the py38 job passed | 18:03 |
| sean-k-mooney | so it should be using hte same base image and tox version | 18:03 |
| sean-k-mooney | https://zuul.opendev.org/t/openstack/build/31c25b583d0d4081900cef987499b342 | 18:04 |
| clarkb | huh I don't know then. But I'll continue to strongly recomment no one use tox for this particular reason. They use their own implementation and you run into random errors like this | 18:04 |
| sean-k-mooney | so we could jsut turn off that job in nova but one thing i want to check is on centso which has py39 as its native python what actully happens | 18:05 |
| sean-k-mooney | clarkb: we are doing some funky things like useing py38 wheels in the py39 venv i think | 18:06 |
| sean-k-mooney | so you coudl be right that somethign is clamping to an older version | 18:06 |
| sean-k-mooney | ok so if i do ".tox/.pkg/bin/python3 -m pip install -e ../pbr/" | 18:23 |
| sean-k-mooney | with stephens patch then it silences the warning | 18:23 |
| sean-k-mooney | ie the one that pbr does not supprot build_editiable | 18:23 |
| sean-k-mooney | but its the pbr in the .pkg venv not the venv used to run the test that matters | 18:24 |
| sean-k-mooney | and also not the pbr in the env used to execute tox | 18:24 |
| clarkb | ya because it uses isolated build envs | 18:24 |
| clarkb | we should be able to land stephenfin's updates soon and make a release which will work around that | 18:25 |
| sean-k-mooney | ya so unless we pin it in the tox.ini | 18:25 |
| sean-k-mooney | we have no way to actully contol that | 18:25 |
| clarkb | I've got an update to the python2.7 job to only run smoke tests which I hope will make it more reliable | 18:25 |
| clarkb | but I don't want to push that until the current run finishes for data collection purposes | 18:25 |
| clarkb | maybe I should just skip the line though and if this works it is good enough | 18:26 |
| sean-k-mooney | ack | 18:26 |
| opendevreview | Clark Boylan proposed openstack/pbr master: Fix stable/train testing with tag checkouts https://review.opendev.org/c/openstack/pbr/+/899881 | 18:26 |
| clarkb | if that passes I'll squash it into stephenfin's stack and things will hopefully be mergable | 18:27 |
| fungi | would it make sense to go ahead and squash that with the sqlam->alembic fix? | 18:27 |
| clarkb | fungi: yes thats where ti will be squashed. I don't want to do that yet because if I do and need to do further adjustments it will rebase the rest of the stack each time | 18:28 |
| clarkb | I want to see this work first | 18:28 |
| fungi | otherwise it's going to have to be tested all over again when you squash them togerther anyqay | 18:28 |
| fungi | well, technically you don't need to rebase the rest of the stack until the squashed fix runs clean | 18:28 |
| clarkb | thats true | 18:28 |
| clarkb | ok you convinced me | 18:29 |
| opendevreview | Clark Boylan proposed openstack/pbr master: Fix PBR integration testing https://review.opendev.org/c/openstack/pbr/+/899862 | 18:32 |
| clarkb | fungi: ^ that detaches it from the rest of the stack and squashes the current state of the two chagnes together | 18:32 |
| clarkb | I'll keep an eye on it and rebase the rest of the stack if this goes green | 18:33 |
| fungi | clarkb: stephenfin: the squashed job fixes change passes check, i guess the rest can be restacked onto that | 19:51 |
| clarkb | on it | 19:52 |
| opendevreview | Clark Boylan proposed openstack/pbr master: Add support for PEP-660 https://review.opendev.org/c/openstack/pbr/+/899848 | 19:53 |
| opendevreview | Clark Boylan proposed openstack/pbr master: build: Use kwargs https://review.opendev.org/c/openstack/pbr/+/899849 | 19:53 |
| clarkb | I went ahead and approved the testing fix change. I'll defer to fungi to approve the pep 660 update once satisfied it doesn't break older pythons | 19:55 |
| fungi | thanks! | 19:55 |
| fungi | i'm heading out to an early dinner momentarily, but will probably be back before testing for that finishes anyway | 19:56 |
| opendevreview | Merged openstack/pbr master: Fix PBR integration testing https://review.opendev.org/c/openstack/pbr/+/899862 | 21:30 |
| opendevreview | Merged openstack/pbr master: Add support for PEP-660 https://review.opendev.org/c/openstack/pbr/+/899848 | 23:28 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!