EJBCLIENT-333 Unable to invoke any EJB of the same module after failu…

[chengfang]

…re of a SFSB in that module.

[fl4via]

@rachmatowicz Since this affects, affinity, can you please review it? thanks!

[rachmatowicz]

@fl4via Yes, but I may not get to it until Monday.

[rachmatowicz]

I'm looking at this example now, the reproducer doesn't build correctly for me, so spending some time with that. Cheng, any tips on what you did to get it running?

[chengfang]

@rachmatowicz I just run "mvn clean install" in tests directory, with java 1.8:

[INFO] testBuild .......................................... SUCCESS [  0.951 s]
[INFO] secondTestEJB3 ..................................... SUCCESS [  2.279 s]
[INFO] secondTestEJB3_EAR ................................. SUCCESS [  0.556 s]
[INFO] testEJB3Client ..................................... SUCCESS [  1.446 s]
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time:  5.561 s
[INFO] Finished at: 2019-07-01T15:33:54-04:00
[INFO] ------------------------------------------------------------------------
/Users/cfang/dev/customer/EJBCLIENT-333/tests > java -version
java version "1.8.0_181"

Then once it's built successfully, follow tests/README.txt to configure 2 nodes, deploy testEJB3Client war to node1 and secondTestEJB3_EAR to node2, and start 2 nodes in the same machine with port offset.

[chengfang]

I found a bit hard to understand the tests at first, so I put a note in the jira comment explaining it a little more. The reproducer is probably adapted from some other case. Other than that, I was able to reproduce it consistently, and verify the fix with the reproducer.

[rachmatowicz]

Maven settings problem. Working now. Thanks.
Finally got the reproducer to work to reproduce the error.

[rachmatowicz]

@chengfang @fl4via Comments on PR on the issue.

[rachmatowicz]

Might be worth adding a comment referring to this issue at the point where you introduce the check for stateful bean, to explain why the condition is needed.

[rachmatowicz]

@chengfang @fl4via Sorry, one other point which comes to mind. Should we be in the habit of writing a test case to validate every bug fixed and every new feature added, in order to expand the test coverage of the component test suites?

[chengfang]

@rachmatowicz I agree we should strive to write tests along with bug fixes and new features, especially unit tests. For more complex use case testing and integration tests (such as this issue), which usually requires complex setup and testing framework, it's best handled by dedicated QE tests.

[xstefank]

@chengfang is this good to be merged or are you still working on a testcase, please?

[chengfang]

@xstefank No, I'm not working on tests for this issue. As I commented earlier, this is one of those issues that are not easily testable by dev tests. So I think this PR is ready for merge.

[xstefank]

@fl4via can this be merged then?