Storm High Availability and Failover - Real-time Analytics with Storm and Cassandra - page 109

cause two of the other available Zookeepers have sufficient resources in terms of
connections that can keep the Storm cluster up and running.
• Test 3 (all components are up and the topology is running): Kill two Zookeeper
nodes and you will notice that the topology will continue to execute normally, be-
cause one of the other two available Zookeepers have sufficient resources in
terms of connections that they can keep the Storm cluster up and running.
• Test 4 (all components are up and the topology is running): Kill supervisor 2;
now we have one of the gray workers on this node. So when this node goes down,
the gray worker dies, and then because the second supervisor is not available it's
spawned again, this time on supervisor 1. So all workers of the topology will be
executing on one single supervisor now, but the system will continue to perform
with limited resources but will not fail.