|
|
IBM WAS ND V6 supports a mirrored backup cluster that can failover EJB requests from a primary cluster in one cell to the mirrored backup cluster in another cell when the primary cluster fails. Fail back is automatic as soon as the primary cluster becomes available.
There are certain situations in which you would use backup cluster support instead of other high availability options. Any scenario must deal with IIOP traffic which could be from servlets, Java applications, the client container or other traffic. If your configuration contains two cells with mirrored applications, you might want to enable backup clusters for the case of catastrophic failure when an entire cell goes down.
Before you can enable backup cluster support, create the backup cluster in another cell with the same cluster name as the primary cluster and deploy the same applications into both the primary cluster and the mirrored backup cluster, with the same resources in the backup cluster as in the primary cluster. The primary cluster and the backup cluster must reside in separate cells. The backup cluster bootstrap host and port determine which cell contains the backup cluster.
The basic behavior for backup clusters requires that the Deployment Managers are running in order to perform the failover and possible failback. In WebSphere V6, backup cluster communication no longer only flows through the Deployment Manager. Although the Deployment Manager is still required for the failover (as it stores the location of the backup cluster), core groups and core group bridges allow backup cluster communication to be dispersed over multiple bridges.
The core group bridge service can be configured to allow communication between the core groups. The core group bridge service handles the communication between a primary and backup cluster. You can configure core group bridges between the Deployment Managers, Node Agents and servers to prevent a single point of communication failure. Each cluster can only have one backup cluster. The primary and backup clusters are configured to point to each other to allow fail back.
