Now I'm pretty confident that the ResourceTracker created by the test case test_migrate_disk_and_power_off_crash_finish_revert_migration during restart_compute_service() [1] call survives the test case and gets activated later by another test case. (Look at rt object id in the logs, they are the same, and the compute uuid is the same too). This upgrade my 75% confidence to 95% regarding the need of [3].
One more reproduction with more trace (needed around 200 functional test run to reproduce): http:// paste.openstack .org/show/ 788538/
Now I'm pretty confident that the ResourceTracker created by the test case test_migrate_ disk_and_ power_off_ crash_finish_ revert_ migration during restart_ compute_ service( ) [1] call survives the test case and gets activated later by another test case. (Look at rt object id in the logs, they are the same, and the compute uuid is the same too). This upgrade my 75% confidence to 95% regarding the need of [3].
[1] https:/ /github. com/openstack/ nova/blob/ 2fceddc451bd021 f137c6756c5314e 08eb4443de/ nova/test. py#L425 /review. opendev. org/#/c/ 656709/ /review. opendev. org/#/c/ 512552
[2] https:/
[3] https:/