]
William Burns updated ISPN-5884:
--------------------------------
Status: Resolved (was: Pull Request Sent)
Resolution: Done
ClusteredListenerReplTest.testPrimaryOwnerGoesDownAfterBackupRaisesEvent random failures
----------------------------------------------------------------------------------------
Key: ISPN-5884
URL:
https://issues.jboss.org/browse/ISPN-5884
Project: Infinispan
Issue Type: Bug
Components: Core, Test Suite - Core
Affects Versions: 8.1.0.Alpha2
Reporter: Dan Berindei
Assignee: Dan Berindei
Priority: Blocker
Fix For: 8.1.0.Beta1, 8.0.2.Final
{{ClusteredListenerReplTest}} uses a {{BlockingInterceptor}} to ensure a write command is
blocked on the backup owners while the primary owner is killed.
The {{BlockingInterceptor}} is removed after the primary owner was killed, but only on
one of the backups. This is enough if the node 0 becomes the primary owner of the key,
because the command is retried before the previous invocation is unblocked, and
{{BlockingInterceptor}} only blocks one command at a time. But if node 2 becomes the
primary owner, the retry happens after the initial command was unblocked, and it blocks in
{{BlockingInterceptor}}.
The fix would be to remove the interceptor from both nodes.
I also believe the "block one command at a time" logic should be removed,
because it makes tests too dependent on timing.