gibi | sean-k-mooney: fyi the latest pyroute2 bump broke os-vif so I proposed a revert https://review.opendev.org/c/openstack/requirements/+/948283 | 11:19 |
---|---|---|
sean-k-mooney[m] | didnt rodolfo fix that | 11:19 |
sean-k-mooney[m] | although we have not done a release since | 11:20 |
sean-k-mooney[m] | @gibi https://review.opendev.org/c/openstack/os-vif/+/946977 | 11:21 |
gibi | sean-k-mooney: that is just unit test adaptation | 11:21 |
gibi | what I see is interface detach broken in tempest | 11:21 |
sean-k-mooney[m] | gibi: oh its using asyncio | 11:21 |
gibi | yeah | 11:22 |
gibi | we need to remain on 0.8.1 I guess | 11:22 |
sean-k-mooney[m] | we do do that in tempset | 11:22 |
sean-k-mooney[m] | but i guess the bump was not done at the time | 11:22 |
sean-k-mooney[m] | we could try changing the executor for eventlet to asyncio but i think that also does not work properly yet | 11:23 |
sean-k-mooney[m] | ok well i think a revert for now makes sense | 11:23 |
sean-k-mooney[m] | and then we need to look deeper | 11:23 |
gibi | yeah | 11:24 |
gibi | anyhow I only stopped by for rechecks. I'm moving on :) | 11:25 |
frickler | I approved the revert, but I'd be interested in learning why this wasn't detected by the tempest job we are running on reqs, does it only affect ovs? maybe we should have a second tempest job in reqs, then? | 13:35 |
gibi | frickler: thanks. As far as I see it is only a problem if the interface detach times out due to slow node and retried by nova, the os-vif unplug is retried while I guess in pyroute2 something is still running for the first try. So probably the original tempest run does not hit it as it did not need to re-try the detach | 14:56 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!