Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
Discovery Base 1.0.2
-
None
Description
AbstractClusterTest.testConnectorSwitching4139 has a part where it has 5 instances, then crashes 1 and wants to make sure that its crash is properly propagated through the topology. The problem is, that instance has a timeout of 10sec. And the loop does at least 20x500ms and ensures that the crashed instance is properly noticed within its cluster by its peer instance (only). That doesn't mean though, that it also propagated to the others via connectors - as that would just yet happen with the next heartbeat.
So in short: those 20x500ms looping are too close to the 10sec timeout.