Friday, 2026-10-16

zigotkajinam: puppet-openstack-integration-8-scenario001-tempest-debian-trixie is already in success ! :)06:25
zigoOthers are affected by the '<frozen ._bootstrap>.log' and just fail the "save log" thing.06:26
zigoI'm trying to figure this out, ie: why placement behaves like this even though I've set --log-file in PYARGS.06:27
zigotkajinam: This get worsen by this issue that I also am fixing (upstream and in Debian): https://review.opendev.org/c/openstack/placement/+/101082307:04
zigoSo, Placement has a bug that one cannot set any args, and so --log-file is never forwarded to oslo.log, so it gets the frozen garbage.07:06
zigoOne more point why my oslo.log patch is needed.07:06
tkajinamit makes me feel like that we should probably consider stop using log files and direct all log outputs to systemd/journald at some point07:19
tkajinamthough that likely causes ugly duplicated timestamp/levels/etc07:19
zigoSame issue in Zaqar: https://review.opendev.org/c/openstack/zaqar/+/101082407:24
zigoFrom my perspective: I do like --log-file, that's very convenient and help differenciate daemons, and I also love the fact that oslo.log offers any facility that anyone wants (ie: rsyslog as well if someone wants).07:25
opendevreviewMerged openstack/puppet-ceph master: Handle secrets as Sensitive values  https://review.opendev.org/c/openstack/puppet-ceph/+/100668507:45
tkajinamzigo, I wonder if you tried [uwsgi] procname = heat-api in uwsgi.ini ?07:47
tkajinamzigo, maybe you want a bug report at least for placement assuming you want that to be backported ?07:48
tkajinamzaqar is applying quite loose policy but for placement they may want that07:48
zigotkajinam: Again, the issue is *NOT* my Debian packaging or how I start daemons, it's about the default falling back to <frozen importlib._bootstrap>.log07:49
zigoI'll follow your advice and file a bug against placement, this sounds good.07:49
tkajinamagain I'm fine with replacing None by a better string07:50
tkajinamI'm concerned with adding uwsgi specific thing there, because that allows people to ask "hi this doesn't work for our deployment using something new so add this logic for us"07:50
tkajinamthat's what I meant to mention in my comments there07:51
zigoSure, but that's protected by reading if uwsgi is the thing that starts the API.07:53
zigoIf you don't like it, I'm fine removing that part, though what do you suggest so that it's not always the same basename(sys.argv[0]) for all daemons?07:54
zigoIt must come down to something.07:54
tkajinamlet me dig into the history to find out why we ended up using that stack thing08:07
tkajinamzigo, can you try something like this https://paste.opendev.org/show/bTDxUooLn6UGT3n3QUL9/ ?08:21
zigotkajinam: Well, NO ! :)08:23
zigoThis will fallback to always having uwsgi_python3 for all APIs.08:24
tkajinamzigo, for per-file separation you should use log-file or procname in uwsgi08:24
zigoI agree, but we should also have a nicer fallback than just sys.argv[0].08:24
tkajinamif uwsgi doesn't tell proc name correctly then adapt your configuration so that the logic should work with that pattern08:24
tkajinam--log-file is that nicer fallback I'm saying and is more robust08:25
zigoYeah, that's not mutual-exclusive than having better fallbacks.08:25
zigoI'd prefer "please-set-your--log-file-correctly.log" ...08:25
zigoI had head-scratching because of this for months, I'd prefer if other downstream do not.08:26
zigoAs you saw, there are many cases where it's easy to have log_file not working correctly.08:27
zigoI uncovered 2 issues at once about this.08:27
tkajinamzigo, I'm not fully following you. if sys.argv[0] doesn't work for them then they can override it by whatever you want and using argv[0] isn't that bad idea, because it's predictable and human readable10:12
zigoHaving /var/log/placement/uwsgi_python3.log will not help operators to understand "oh, I didn't set --log-file", that's what I'm saying ...10:29
zigotkajinam: I don't mind much the details as long as we fix the <frozen importlib... thing. Do you feel like taking over the patch then ? :)10:48
tkajinamzigo, if you don't mind I can take it over11:22
tkajinami noticed that we don't document the behavior with log-dir but without log-file and we should explain the behavior there. I'm on it, too.11:22
zigoThanks !11:23
opendevreviewThomas Goirand proposed openstack/puppet-openstack-integration master: WIP / DNM: Add Debian 13 (Trixie) integration job  https://review.opendev.org/c/openstack/puppet-openstack-integration/+/98908511:28
opendevreviewMerged openstack/puppet-neutron master: Deprecate support for [oslo_policy] enforce_scope  https://review.opendev.org/c/openstack/puppet-neutron/+/99708716:15
zigotkajinam: Scenarios 4 and 5 are gree on my Debian WIP / DNM patch ! :)18:43
zigogreen18:43
zigo3 is just a <frozen log file issue.18:44
zigo(tests are passing otherwise)18:44
zigoSo it's looking good ...18:44
opendevreviewThomas Goirand proposed openstack/puppet-ironic master: Fix ironic_api_uwsgi_config idempotency  https://review.opendev.org/c/openstack/puppet-ironic/+/101100519:46
zigo^ This one is one of the 2 reasons why scenario002 isn't green19:47
opendevreviewThomas Goirand proposed openstack/puppet-openstack-integration master: WIP / DNM: Add Debian 13 (Trixie) integration job  https://review.opendev.org/c/openstack/puppet-openstack-integration/+/98908520:14
zigoOk, ALL 5 Debian scenarios will pass once that Ironic uwsgi_config thing is merged !  :)21:47
opendevreviewThomas Goirand proposed openstack/puppet-openstack-integration master: Add Debian 13 (Trixie) integration job  https://review.opendev.org/c/openstack/puppet-openstack-integration/+/98908522:12
zigoRemoved "WIP / DNM" from patch title ! :)22:12

Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!