Partitions are by default highly available. A partition will only be hosted
on a single cluster member at a time. They are made highly available using
the high availability (HA) manager component. If a cluster member fails either
because of a Java Virtual Machine (JVM) shutdown, a JVM panic or the box hosting
the cluster member fails/gets powered down then the HA manager moves all partitions
which were running on the failed cluster members to the surviving cluster
members.
The following are the state transitions a partition will encounter as member
of a HA group.
- IDLE
- Partition is currently deactivated and waiting for an activation command,
a partition is in this state at the Bean start during server startup after
the createPartitionDefinition method until the PartitionLoadEvent API is called
by the HA manager coordinator.
- ACTIVATING
- The HA manager is attempting to activate the partition but has not been
acknowledged yet, this is in effect during the PSSB PartitionLoadEvent method
execution.
- ACTIVATED
- This means the partition currently activate and working. The PartitionLoadEvent
event has completed. This state implies that the HA manager has picked a target
application server for this partition, it is active and ready to process methods,
and the workload management layer has enabled for client to route requests
to the particular application server associated with the specific partition.
- DISABLED
- This means the partition cannot become an active member while still part
of the group. This is a failure state.
- DEACTIVATING
- The partition received a deactivate signal and is still deactivating.
This occurs while the PartitionUnloadEvent method executes in the partitioned
stateless session bean (PSSB) for this partition. The partition then returns
to IDLE and the HA manager signals the activate coordinator managing that
group and provides a copy of the current local state.
