An A-MQ installation on OpenShift has the following deployment configuration in the attached dc.ym. During a simple redeployment (oc rollout latest), various broker pods are created and others are destroyed. The configuration ensures that at least one pod is running at all times. The configuration uses split data directories with network connectors established by discovery.

When the new pods are starting, some of them, some of the time, fail to establish network connectors to the other pods in the mesh. The following errors are seen in the logs of some of the brokers:

2018-10-11 08:26:18,048 |  | INFO  | Establishing network connection from vm://amq-broker-98-vrgcs to tcp://10.94.27.219:61616 | org.apache.activemq.network.DiscoveryNetworkConnector | OpenShift discovery agent Scheduler: amq-broker-tcp
2018-10-11 08:26:18,048 |  | INFO  | Stopping vm://amq-broker-98-vrgcs#10 because Failed with SecurityException: User name [null] or password is invalid. | org.apache.activemq.broker.TransportConnection | ActiveMQ BrokerService[amq-broker-98-vrgcs] Task-8