Intermitent Newton unit test failures
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Cinder |
Invalid
|
Undecided
|
Unassigned |
Bug Description
It looks like some tests are not deterministic and fails "sometimes". It failed for me twice, and worked only once so far, so I'll be blacklisting these tests in my Debian package. Here's the build log:
FAIL: cinder.
cinder.
-------
_StringException: Empty attachments:
stderr
stdout
pythonlogging:'': {{{
2016-09-14 19:00:54,440 INFO [cinder.
2016-09-14 19:00:54,443 INFO [cinder.rpc] Automatically selected cinder-backup objects version 1.5 as minimum service version.
2016-09-14 19:00:54,446 INFO [cinder.rpc] Automatically selected cinder-backup RPC version 2.0 as minimum service version.
2016-09-14 19:00:54,448 INFO [cinder.rpc] Automatically selected cinder-volume objects version 1.5 as minimum service version.
2016-09-14 19:00:54,451 INFO [cinder.rpc] Automatically selected cinder-volume RPC version 2.2 as minimum service version.
2016-09-14 19:00:54,453 INFO [cinder.rpc] Automatically selected cinder-scheduler objects version 1.5 as minimum service version.
2016-09-14 19:00:54,455 INFO [cinder.rpc] Automatically selected cinder-scheduler RPC version 2.1 as minimum service version.
2016-09-14 19:00:54,491 INFO [cinder.
2016-09-14 19:00:54,494 INFO [cinder.
2016-09-14 19:00:54,495 INFO [cinder.
}}}
Traceback (most recent call last):
File "/usr/lib/
return func(*args, **keywargs)
File "cinder/
'connector': os_brick.
AttributeError: 'module' object has no attribute 'FakeConnector'
The 2nd one is this one, so I'm not pasting it:
https:/
I've previously been able to run all unit tests without a single failure, so that really is non-deterministic.
This was due to some hard coded internal paths in the unit tests that have since been updated.