| amorin | hello team, mistral is now evenlet free, YAY! | 07:20 |
|---|---|---|
| hberaud[m] | \o/ | 07:20 |
| hberaud[m] | Congratulations amorin | 07:20 |
| amorin | but still I question, when mistral is installing in a venv, eventlet is also installed, due to oslo.service[threading] | 07:20 |
| amorin | is that expected? | 07:20 |
| hberaud[m] | IIRC yes | 07:20 |
| hberaud[m] | that's cotyledon/futurist who are optional | 07:21 |
| hberaud[m] | eventlet remained the default requirements | 07:21 |
| amorin | I did try to uninstall it manually after that, my wsgi service is then refusing to start (related to init backend not possible or something like this), is it also expected? | 07:21 |
| hberaud[m] | yes | 07:21 |
| hberaud[m] | see https://review.opendev.org/c/openstack/oslo.service/+/955199 | 07:22 |
| amorin | perfect, I was surprised by this behavior since we patch the BACKEND very early in the starting procedure | 07:22 |
| hberaud[m] | amorin: and more especially see https://opendev.org/openstack/oslo.service/commit/5b71dae6a54b7b9c844d1d30c12f5227dd4a7811 | 07:23 |
| amorin | hum, but should I have this issue only if I rely on oslo.service wsgi? Because I am seeing the issue with uwsgi | 07:24 |
| hberaud[m] | you do not need the router or the loader, etc... ? | 07:25 |
| amorin | I use service and sslutils | 07:26 |
| amorin | from oslo_service import service | 07:26 |
| amorin | from oslo_service import sslutils | 07:26 |
| amorin | here: https://review.opendev.org/c/openstack/mistral/+/958595 | 07:26 |
| hberaud[m] | we be good to inspect your a bit more your error | 07:26 |
| amorin | in mistral/api/service.py | 07:26 |
| amorin | BUT, I also prepared this: https://review.opendev.org/c/openstack/mistral/+/959603 | 07:27 |
| hberaud[m] | ack | 07:27 |
| amorin | so I can afford using the service itself | 07:27 |
| hberaud[m] | ok | 07:28 |
| amorin | but this is still broken when I start with eventlet uninstalled, maybe I can grab the stack trace and paste somewhere, that would help | 07:28 |
| hberaud[m] | yes please | 07:28 |
| hberaud[m] | I think there is a corner case | 07:28 |
| amorin | I dont like corner cases :( | 07:32 |
| hberaud[m] | if you are able to reproduce that bug would be worth to fill a launchpad bug | 07:33 |
| *** ralonsoh__ is now known as ralonsoh | 13:36 | |
| amorin | hberaud[m]: take a look at the stacktrace here: https://plik.arno.ovh/file/WarQpOC0zEnmW0GL/4amfPoOUPQp7ABNl/t | 15:12 |
| hberaud[m] | OMW | 15:13 |
| amorin | it's on the init of the backend that it fails | 15:13 |
| amorin | let me try to try / except this | 15:13 |
| amorin | ok, maybe that's my fault actually, I created the wsgi without invoking the init_backend initially | 15:20 |
| amorin | I think this is the solution: https://review.opendev.org/c/openstack/mistral/+/959603 | 15:24 |
| amorin | calling init_backend | 15:24 |
| amorin | I also proposed on ironic side https://review.opendev.org/c/openstack/ironic/+/959870 but maybe that's not relevent | 15:35 |
| hberaud[m] | awesome, thank you Arnaud | 16:13 |
| JayF | I'm really confused about ironic. | 17:07 |
| JayF | afaict you are right | 17:07 |
| JayF | but I know we explicitly tested to ensure we were eventlet-free | 17:32 |
Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!