| *** mhen_ is now known as mhen | 01:58 | |
| opendevreview | Daniel Bengtsson proposed openstack/oslo.service master: Add threading backend implementation using cotyledon and standard threads https://review.opendev.org/c/openstack/oslo.service/+/945720 | 10:26 |
|---|---|---|
| opendevreview | Daniel Bengtsson proposed openstack/oslo.service master: Add threading backend implementation using cotyledon and standard threads https://review.opendev.org/c/openstack/oslo.service/+/945720 | 10:50 |
| gibi | ralonsoh: hberaud[m] damani[m]: have you tried the behavior of the periodic_task decorator with the threading backen? It seems to me that the periodic is not running with threading backed but it works as usual with eventlet backend | 11:42 |
| ralonsoh | gibi, I think we have this in the L3 agent | 11:45 |
| ralonsoh | this is the only one I've tested locally with the threading backend | 11:45 |
| ralonsoh | let me check again, one sec | 11:47 |
| ralonsoh | gibi, yes, it is working. In the L3 agent we have https://github.com/openstack/neutron/blob/5090d675be06fb6242485a245aeefa640e9bbc58/neutron/agent/l3/agent.py#L835 | 11:50 |
| ralonsoh | and I see right now the logs with the debug message | 11:50 |
| ralonsoh | several times | 11:50 |
| ralonsoh | gibi, just a heads-up | 11:50 |
| gibi | thanks. Then I will dig deeper on the nova side | 11:51 |
| ralonsoh | I had a problem with RPC clients | 11:51 |
| ralonsoh | when using the threading backend | 11:51 |
| ralonsoh | because we were instantiating them before the fork oslo.service is doing | 11:51 |
| gibi | so far I only tried the RPC server side, as I'm working with nova-scheduler | 11:51 |
| ralonsoh | so nothing worked | 11:51 |
| ralonsoh | when are you creating the server? in the init method? | 11:52 |
| gibi | nope, it is in the Service.start | 11:54 |
| gibi | https://github.com/openstack/nova/blob/023be4f5615255d06f61b393605da1308f045220/nova/service.py#L177 | 11:55 |
| gibi | I guess if you do before the fork then the problem is similar to what I found with threadpool started before the fork https://gibizer.github.io/posts/Eventlet-Removal-The-First-Threading-Bug/#threads--osfork--confused-threadpools | 11:55 |
| hberaud[m] | No sorry I didn't personal tests yet with the new backend, Daniel did. | 11:57 |
| hberaud[m] | I'm going to meet with Daniel in ~1h I'll forward your observations to him | 11:58 |
| gibi | hberaud[m]: ack, I will try to debug what happens in nova-scheduler. I see that the thread for the looping call was created for the periodics, I need to look deeper | 11:59 |
| hberaud[m] | thanks, do not hesitate to share links or logs with me, this way I'll forward it to Daniel | 12:00 |
| gibi | hberaud[m]: I think I see the reason but not sure about the solution. I added comment inline https://review.opendev.org/c/openstack/oslo.service/+/945720/comment/35eb6a53_0097ad88/ | 12:09 |
| hberaud[m] | ack, thanks | 12:10 |
| gibi | I think I see the fix now too. Left comments | 12:15 |
| *** iurygregory_ is now known as iurygregory | 13:08 | |
| damani[m] | gibi, i will check now your two last comments | 13:35 |
| gibi | ack | 13:35 |
| gibi | thanks | 13:36 |
| hberaud[m] | damani: thanks for working on that | 13:36 |
| opendevreview | Daniel Bengtsson proposed openstack/oslo.service master: Add threading backend implementation using cotyledon and standard threads https://review.opendev.org/c/openstack/oslo.service/+/945720 | 13:59 |
| damani[m] | gibi, patch updated | 14:00 |
| opendevreview | Daniel Bengtsson proposed openstack/oslo.service master: Add threading backend implementation using cotyledon and standard threads https://review.opendev.org/c/openstack/oslo.service/+/945720 | 14:40 |
| opendevreview | Daniel Bengtsson proposed openstack/oslo.service master: Add threading backend implementation using cotyledon and standard threads https://review.opendev.org/c/openstack/oslo.service/+/945720 | 14:42 |
| opendevreview | Daniel Bengtsson proposed openstack/oslo.service master: Add threading backend implementation using cotyledon and standard threads https://review.opendev.org/c/openstack/oslo.service/+/945720 | 14:54 |
| damani[m] | tkajinam, i have implemented the restart, change the raise message, none by default for the workers | 14:58 |
Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!