Tuesday, 2025-04-29

amoringibi amazing!08:21
gibifolks, do you have any tips why would oslo logging start behaving strangely (missing logs like the oslo.config dump at the service startup, but some service logs are OK) when switching to the threading backend?09:04
hberaud[m]well I think we can potentially look at https://opendev.org/openstack/oslo.log/src/branch/master/oslo_log/pipe_mutex.py09:05
hberaud[m]this pipe mutex things is the root cause of several well know bugs in OpenStack, by example, all the problem behind the heartbeat in pthread find their root causes in this pipe mutex stuff09:07
hberaud[m]an other aspect that we should also consider, is the config glue topic that I mentionned yesterday in this comment https://review.opendev.org/c/openstack/oslo.service/+/945720/comment/aed21ec8_eef31ee0/09:10
gibihberaud[m]: thanks for the pointers. I will take a look09:14
hberaud[m]If tell that the logging problem disappear when switching to the threading backend then, that's surely because this pipe mutex left the scene09:22
hberaud[m]*you09:22
gibiI tell that nova-scheduler dumps the oslo.conf config options if run in eventlet mode (as the rest of the nova services) but does not bump it when run in threading mode, and such dumping logically happens after the oslo service backend selection10:55
gibiI will dig into this10:55
gibimore, to get a proper picture10:55
hberaud[m]in this case I think we are closer to cotyledon's oslo config glue stuff https://github.com/sileht/cotyledon/blob/main/cotyledon/oslo_config_glue.py11:18
hberaud[m]I think that adding this oslo.config glue into your DNM patch (https://review.opendev.org/c/openstack/oslo.service/+/948310) would straightly tell you if this is this missing call to this module is the culprit or not 11:20
hberaud[m](By waiting for Daniel's new updates)11:21
gibiack11:33

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!