Update patch set 11

Patch Set 11: Code-Review-1

The py37 timeout issue seems to be a real thing.

I dug into it for a couple of hours today and made some progress, but I'm far from having it resolved, and I can't really give it any more time I'm afraid.

Here's what I've found out so far:

(1) There are a number of tests that are hitting OS_TEST_TIMEOUT, currently configured at 60s (see tox.ini). Here's an example: https://zuul.opendev.org/t/openstack/build/c8f500d7d3204b2e9bd230cbded4b082/log/job-output.txt#839

(2) Despite hitting this timeout, the tests are *passing*. This is bad, but a separate issue.

(3) This is only happening in py37. Those same tests run sub-second in py36 and py27. I have no idea what changed between py36 and py37 that would make this difference.

So I'm guessing the overall test timeout is happening because we're jamming the thread pool with these long-running test cases.

I'm able to recreate the long-running test cases locally, so it should at least be possible to investigate without throwing things at the gate.

The next course of action is probably to instrument or step through the code to narrow down where this time is being spent. At a glance, I couldn't see any major differences between long-running test cases and nearby ones that run quickly.

Good luck :)

Patch-set: 11
Reviewer: Gerrit User 14070 <14070@4a232e18-c5a9-48ee-94c0-e04e7cca6543>
Label: Code-Review=-1
This commit is contained in:
Gerrit User 14070 2019-09-16 20:01:04 +00:00 committed by Gerrit Code Review
parent 808917cacc
commit f2a8b20202

Diff Content Not Available