HDDS-13310. Intermittent failure in TestContainerOperations #8770
Conversation
| assertEquals(expected, usageInfoList.get(0).getContainerCount()); | ||
| assertThat(usageInfoList.get(0).getContainerCount()) | ||
| .isGreaterThanOrEqualTo(expected); |
There was a problem hiding this comment.
is it due to some datanode having outdated information, which gets updated later before the assertion kicks in? In that case can something like GenericTestUtils.waitFor or some util which triggers datanode to update the info work?
There was a problem hiding this comment.
The test makes two different calls to SCM that each return data about containers for a specific datanode. They both use the same ultimate source of container list. SCM collects info from datanodes in the background (incremental container report and heartbeat). So if a container gets created by a previous test case, it may be added to SCM's list between the two calls.
We could add waitFor, but would need to repeat both calls, since no matter which one we choose as the reference (first call), the second call may return the same number or more.
|
The test was changed in #8677, apparently it passes now on Closing. Will reopen if it happens again. Thanks @ayushtkn for the review. |
2 participants
What changes were proposed in this pull request?
Fix intermittent test failure in
TestContainerOperations. Let the test pass if any datanode usage info has more containers than SCM's view of the same container (which may be slightly outdated).https://issues.apache.org/jira/browse/HDDS-13310
How was this patch tested?
Before: 5/100 failures.
https://github.com/adoroszlai/ozone/actions/runs/16175931905/job/45662413733
After: no failures
https://github.com/adoroszlai/ozone/actions/runs/16174065813