+

Search Tips   |   Advanced Search


When to use a high availability manager


Overview

The amount of resources HA manager consumes, including CPU, heap memory, and sockets, scales nonlinearly as the size of a core group increases. If HA manager provided services are not being used, we can disable the HA manager on appserver processes, and free those resources.

Do not disable the HA manager on administrative processes, such as node agents and the deployment manager, unless the HA manager is disabled on all application server processes in that core group.

If HA manager is disabled on one member of a cluster, disable it on all of the other members of that cluster.


Memory-to-memory replication

Cluster-based service configured at the application server level. If HA manager is enabled on any cluster member, the HA manager must be enabled on all of the members of the cluster.

Automatically enabled if:


Singleton failover

Cluster-based service. HA manager must be enabled on all members of a cluster if:


Workload management routing

Workload management (WLM) uses the HA manager to propagate the following types of routing information:

Typically applies to clustered resources, but can also apply to non-clustered resources, such as standalone messaging engines. Leave the HA manager enabled on any application server that produces or consumes either IIOP or messaging engine routing information.

For example, if the routing information...

When the servlet calls the EJB, the HA manager must be enabled on all servers in both clusters.

WLM can statically build and export route tables to the file system, eliminating the dependency on the HA manager.


On-demand configuration routing

In a WAS ND system, the on-demand configuration is used for IBM WAS proxy server routing. To use on-demand configuration routing in conjunction with Web services, verify that the HA manager is enabled on the proxy server and on all of the servers to which the proxy server routes work.