| gibi | my summary about this weeks achivement of running nova-scheduler in threading mode https://gibizer.github.io/posts/Eventlet-Removal-Scheduler-First-Run/ | 13:12 |
|---|---|---|
| hberaud[m] | thanks | 13:12 |
| hberaud[m] | reading | 13:12 |
| hberaud[m] | really interesting | 13:17 |
| hberaud[m] | I just have one question, "This also shows that the memory consumption and runtime of the threaded code is not terrible." what do you? That using threads isn't that horrible in terms of performance changes, or, that using threads lead you to performances problems. My understanding is more inclined toward the first meanning, using thread is not catastrophic in terms of performance and resources consumption. | 13:20 |
| hberaud[m] | gibi: ^ | 13:21 |
| hberaud[m] | *what do you mean? | 13:21 |
| hberaud[m] | ralonsoh: FYI ^. This feedback could be useful for your Neutron migration, especially for the oslo.service aspects of your migration | 13:23 |
| ralonsoh | hberaud[m], we achieved that before with other Neutron agents (OVN metadata, metadta agent, OVN agent) | 13:35 |
| ralonsoh | I'll read the article, for sure | 13:35 |
| ralonsoh | we didn't have any noticable performance degradation | 13:35 |
| hberaud[m] | ah cool, sorry I missed that point | 13:35 |
| hberaud[m] | nice | 13:36 |
| ralonsoh | and I have a pending article about the socket server proxy for metadata... | 13:36 |
| ralonsoh | but I have no time... | 13:36 |
| hberaud[m] | awesome | 13:36 |
| gibi | hberaud[m]: you undrestanding is correct. It is OK to move forward with threading based on the initial perf | 13:52 |
| hberaud[m] | ack | 13:53 |
| hberaud[m] | thanks for your clarification | 13:53 |
Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!