| *** mhen_ is now known as mhen | 03:04 | |
| opendevreview | Xiang Wang proposed openstack/oslo.messaging master: Implement ConsulOperator as a service broker to support HTTP Driver https://review.opendev.org/c/openstack/oslo.messaging/+/912499 | 06:11 |
|---|---|---|
| opendevreview | Xiang Wang proposed openstack/oslo.messaging master: Add HTTP driver to oslo.messaging.drivers https://review.opendev.org/c/openstack/oslo.messaging/+/914840 | 06:11 |
| opendevreview | Xiang Wang proposed openstack/oslo.messaging master: Implement HTTPListener and HTTPClient and add OpenAPI specs https://review.opendev.org/c/openstack/oslo.messaging/+/915056 | 06:11 |
| zigo | Is there any fix for this one ? https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1089050 | 07:27 |
| zigo | hberaud[m]: ^ | 07:28 |
| zigo | Eventlet related?!? | 07:28 |
| NickKush | hi o/ would like to get early feedback about this https://review.opendev.org/c/openstack/oslo.messaging/+/936847 | 07:48 |
| NickKush | also got Zuul errors that are likely unrelated to the patch. need help with it | 07:48 |
| hberaud[m] | zigo: Apparently at first glance they do not seems related to eventlet | 08:25 |
| zigo | hberaud[m]: Looks like it's an issue with threading, right? | 08:32 |
| zigo | In fact, there's 42 unit test failures in this package. | 08:32 |
| zigo | I tried rebuilding with eventlet 0.38.0, that doesn't fix the problem. | 08:32 |
| hberaud[m] | yeah I see | 08:32 |
| hberaud[m] | Without digging too much neither, I didn't see sign of Eventlet problem | 08:33 |
| *** ralonsoh_ is now known as ralonsoh | 09:21 | |
| zigo | hberaud[m]: Any idea what's happening then? | 09:21 |
| hberaud[m] | nope | 09:25 |
| zigo | :/ | 09:31 |
| frickler | well iiuc eventlet 0.38.0 is required if you want to run on py3.13. and otoh 0.38.0 is failing in the reqs cross jobs, in particular the tempest jobs is failing with something that looks related to oslo.messaging. | 09:37 |
| opendevreview | Dr. Jens Harbott proposed openstack/oslo.messaging master: DNM: testing eventlet bump https://review.opendev.org/c/openstack/oslo.messaging/+/937337 | 09:39 |
| frickler | maybe ^^ gives some more indication | 09:39 |
| frickler | oh, that's not even using 0.38.0, only bumping from 0.36.1 to 0.37.0, but anyway | 09:43 |
| frickler | hmm, at least no fast failure in the tox job, as I had kind of hoped for | 09:48 |
| frickler | zigo: at least I can reproduce getting 10 unit test failures with eventlet 0.38.0 and py3.13 and the errors look similar to your bug report. with 0.36.1 things break down already at the test discovery stage | 10:17 |
| zigo | frickler: My version of eventlet has python 3.13 patches cherry-picked from upstream, so it should work. | 10:55 |
| zigo | (even if it's a 0.36.1) | 10:56 |
| frickler | ok, so that then would seem to undermine the assumption that it is not related to eventlet | 11:00 |
| opendevreview | Hervé Beraud proposed openstack/oslo.messaging master: Deprecate the Eventlet executor https://review.opendev.org/c/openstack/oslo.messaging/+/933399 | 14:13 |
| hberaud[m] | zigo: so if you think that it is related to eventlet please fulfill a new github issue on the eventlet side, once done I'll attach it to the python 3.13 tracker. | 14:21 |
| hberaud[m] | I wonder if we want to also track it on the oslo.messaging side | 14:22 |
| hberaud[m] | According to your observations, I'd argue that that's more an eventlet thing | 14:23 |
| hberaud[m] | The problem you observe could be an occurence of https://github.com/eventlet/eventlet/issues/998 | 14:32 |
| hberaud[m] | your errors seems to be around threads and listeners status | 14:33 |
| zigo | Indeed. | 14:33 |
| zigo | What you just pointed at looks like my exact issue. | 14:34 |
| zigo | Could that be that oslo.messaging isn't doing monkey patching when running its tests? | 14:36 |
| hberaud[m] | no idea | 14:36 |
| zigo | Fun... | 14:36 |
| * zigo .oO(Looking forward we get rid of eventlet...) | 14:36 | |
| hberaud[m] | I personally think that it is more an incomplet support of python 3.13 on the eventlet side | 14:37 |
| hberaud[m] | I think eventlet needs more adjustments to support python 3.13, so pieces are surely missing | 14:37 |
| hberaud[m] | because the eventlet issue initially reported was not related to oslo.messaging | 14:38 |
| hberaud[m] | and the major update for python 3.13 was around threading https://github.com/eventlet/eventlet/pull/988/files | 14:39 |
| hberaud[m] | you already know the story | 14:39 |
| zigo | Yeah, I got that patch in my package! :) | 15:00 |
| zigo | That's the one I was referring to, but couldn't find it again in github. | 15:02 |
| * zigo fixes the Debian patch header | 15:02 | |
| opendevreview | Dr. Jens Harbott proposed openstack/oslo.messaging master: DNM: testing eventlet bump https://review.opendev.org/c/openstack/oslo.messaging/+/937337 | 15:08 |
| opendevreview | Dr. Jens Harbott proposed openstack/oslo.messaging master: DNM: test condition locking https://review.opendev.org/c/openstack/oslo.messaging/+/937374 | 15:08 |
| hberaud[m] | zigo: we should notice that the amqp1 driver has been removed so, normally we can ignore the errors related to this driver https://opendev.org/openstack/oslo.messaging/commit/7997a199acada382b0c61dc84aa956e4c65de986 | 17:04 |
| hberaud[m] | in parallel, this driver's tests previously used daemonized threads, like the PoC given in the github issue, so I wonder if daemon threads do not need additional refactors | 17:05 |
| frickler | hberaud[m]: zigo: fyi https://review.opendev.org/c/openstack/oslo.messaging/+/937337 seems to look much better than before in terms of running oslo.messaging with eventlet 0.38.0. the remaining failure seems related to swift, maybe timburke can take a look there | 17:58 |
| frickler | I must admit I have no idea whether my fix is actually correct, I just came up with it based on what the threading docs say. I have found no indication as to why this hasn't been broken forever, though | 17:59 |
| frickler | actually my patch is nonsense, I totally missed the "with self._cond" wrapper | 18:15 |
| frickler | after reading more docs, I think my patch should essentially be a no-op, but since it isn't, there's bound to be a bug somewhere? | 18:39 |
| timburke | oh, that's fun: https://github.com/eventlet/eventlet/commit/ed743d75ec67d20610ab51e19c272baab1efafec | 23:05 |
| timburke | idk why CONTENT_LENGTH vs Content-Length should be such a worry when there are so many other fun ways to break header parsing: https://github.com/python/cpython/issues/81274 | 23:21 |
| timburke | https://review.opendev.org/c/openstack/tempest/+/937416 | 23:33 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!