Unable to start server instance on Txxxx series (Niagara) machines. (6860787)

Description

On T2000 and T5xxx machines there is possibility of JMX timeouts happening and preventing synchronization. The process of binding the JMX/JMX connection to the domain administration server (DAS) is bounded to 5 seconds. If the JMX connection plus the handshake takes more that 5 seconds, the JMX connection might be deemed a failure. This seems to happen more on a T-series machine, especially if the ORB+SSL initialization takes more time. However, this type of failure also happens in other machine/architecture situations.

Solution

1. Increase the default values from 5 seconds. 10 seconds should be acceptable, unless the target DAS/network is extremely loaded. (2-4 seconds is the approximate subsystem setup overhead for the ORB/SSL/RMI connection even on a normal Sun v240.)

2. In Application Server, use the flags (as JVM options) to set the JMX timeout of the synchronization, and other JMX connections if needed.

   • Set the timeout by using DJMXCONNECTOR_TIMEOUT_MILLISEC=<timeout>

   • Turn the timeout off and revert to the old behavior by using DDISABLE_JMXCON_THREAD=true

. As these are JDK options, the synchronization can happen as follows:

• As a node agent starting a Java program using INSTANCE-SYNC-JVM-OPTIONS as the synchronization JVM flag

  See documentation for details on the option. You can set this property with either of the above values. However due to existing bug 6857893, INSTANCE-SYNC-JVM-OPTIONS can currently only take one JVM value

• As a normal JDK option to the instance/cluster configuration