[
https://issues.apache.org/jira/browse/IGNITE-9825?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17418890#comment-17418890
]
Stanislav Lukyanov commented on IGNITE-9825:
--------------------------------------------
So, I tried to revisit this and see whether our tests today would catch this.
I've checked the following:
* If you revert 6e0ff06, the problem is reproduced with a simple test (see
below) - logs are full of NPEs
* At the same time, the NPEs seem not to cause any real issues - the data is
intact as confirmed by idle_verify in the test
* RunAll with 6e0ff06 passes - so, we indeed don't have a test that treats the
NPE in the logs as a failure
I'd say that while in ideal world we would've caught this and any other NPE in
the logs in RunAll. But in reality, test logs can be full of (perfectly valid)
exceptions, and it would be hard to catch them. I think the NPE in question
also could be found in the RunAll with the 6e0ff06 reverted even in the current
tests, but we don't scrape test logs for NPEs.
The bottom line is: we can certainly say that our tests may fail to find a new
weird exception in the logs if it doesn't cause any issues other than annoying
log messages; at the same time, I currently don't have any basis to say that a
gross regression would not have been caught by our tests. Doesn't look like we
need to do anything here.
> Need test for cache writes on unstable topology
> -----------------------------------------------
>
> Key: IGNITE-9825
> URL: https://issues.apache.org/jira/browse/IGNITE-9825
> Project: Ignite
> Issue Type: Test
> Reporter: Stanislav Lukyanov
> Assignee: Stanislav Lukyanov
> Priority: Major
> Time Spent: 10m
> Remaining Estimate: 0h
>
> It appears that ignite-core test suites have no tests for a simple case -
> cache writes on unstable topology.
> Need to add a simple test covering this.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
