]
Tristan Tarrant updated ISPN-6235:
----------------------------------
Fix Version/s: 8.1.4.Final
ClusterTopologyManagerImpl join during cluster status recovery
--------------------------------------------------------------
Key: ISPN-6235
URL:
https://issues.jboss.org/browse/ISPN-6235
Project: Infinispan
Issue Type: Bug
Affects Versions: 8.1.3.Final
Reporter: Dan Berindei
Assignee: Dan Berindei
Labels: testsuite_stability
Fix For: 8.2.0.CR1, 8.1.4.Final
If the joiner has the correct view id, but the current status is
RECOVERING_CLUSTER, we should wait for the cluster status recovery to
finish before adding the new member.
We are currently not doing that, so the new member could be erased by the status recovery
process that's in progress. This can happen if the coordinator joiner already had been
a member of the JGroups cluster for some time, and there's no view change when they
actually start their caches (exactly the scenario in {{ConcurrentStartTest}}).