Comment 5 for bug 1859766

I managed to get one reproduction (in 27 test full functional run) with some extra logging[1]. It seems that a ResourceTracker from an earlier test leaks to the later test as the both the python object id of the ResourceTracker and the uuid of the compute node is the same in both tests.

Now based on that the leaking test case calls nova.test.TestCase.restart_compute_service() I suspect both [2] and [3] needs to be backported to get rid of the problem. But I'm only like 75% sure as the problem is really hard to reproduce.

[1] https://paste.ubuntu.com/p/Jw9bc3yKxW/
[2] https://review.opendev.org/#/c/656709/
[3] https://review.opendev.org/#/c/512552/