Simultaneous startup of all cluster instances might cause monitoring to fail for some instances (6796388)

When a cluster is started using the asadmin start-cluster command or the Start Cluster command in the GlassFish Admin Console, GlassFish starts all the cluster instances simultaneously. In some cases, this simultaneous startup causes some of the instances to start without monitoring available. In Performance Monitor, such instances display "Unavailable" in their Memory and Uptime fields on the cluster view.

This issue has been observed on clusters of 8 instances with JMX over SSL, with the instances distributed on uniform hardware. It might also occur on larger clusters without SSL.

Workaround: Start the instances individually at staggered intervals. For example, using a script that starts the instances 10 to 20 seconds apart prevents this issue from occurring. (The exact delay time is hardware dependent.)