ISPN-2161 InitiatorCrashPessimisticReplTest.testInitiatorNodeCrashesBefo... #1958
Conversation
…eforeCommit fails randomly * Cleanup by replacing all the assert keywords by AssertJUnit.assertEquals() * Problematic test implementation is no-op. See code commentary.
|
thanks! |
| In pessimist caches, we have only one phase (PrepareCommand with onePhaseCommit to true). | ||
| So we only have two scenarios: | ||
| 1. initiator dies before prepare (covered in testInitiatorNodeCrashesBeforePrepare) | ||
| 2. initiator dies after prepare (covered in testInitiatorCrashesBeforeReleasingLock) |
There was a problem hiding this comment.
@pruivo I thought the PrepareCommand is sent during the commit phase?
There was a problem hiding this comment.
@danberindei you are correct. However, this test is invoking TM.commit() that will perform the prepare phase + commit phase
Also the test tries to block the commit in the begin of the interceptor chain (that never happens because the commit is not sent). The tests fails when the prepare is faster than "main" thread and it commits and releases the locks before the assertions.
If the comment is not clear I can try to improve it.
There was a problem hiding this comment.
Yeah, I didn't read the test properly and I thought it was using TransactionXaAdapter.prepare/commit to check the locks before sending the prepare/commit (confusion with InitiatorCrashPessimisticTest, perhaps).
Maybe the comment could be clearer, e.g "before the prepare is received on the coordinator", but I wouldn't ask for a new PR just for that :)
...reCommit fails randomly
https://issues.jboss.org/browse/ISPN-2161