Clustered Edition

If you cluster components (in addition to distributing them), it lets you use the following features:

• Application Server clustering (with SonicMQ JMS)

• Mediation server load balancing (with the Application Server round robin process)

[spacer]

This does not support database replication (you must upgrade to HA edition)

Clustered servers can process using a round-robin method so that systems are protected from server failure. Also, with distributed mediation servers (also called mediation agents), the system can handle a larger number and volume of traps and events. This is usually used with even larger systems, or when the server will be processing a lot of traps (for example, 1-20 traps per second, bursting to 100 tps).

Figure 2-1  Mediation Cluster Setup Screen

Installation now includes a setup screen for the Config Server (the primary server in a cluster). The cluster’s multicast address is configured automatically. Fill in these values if you are installing a clustered or high availability system.

The following are some clustering considerations. For additional details consult Clustering.

• Clustering does not require DNS entries for all servers.

• Servers (Application or Mediation) in a cluster need to share the following two pieces of information:

Partition Name - Some unique string for this cluster. You can configure this name one of several ways for clustering:

– Command line: - typical for testing

– installed.properties - the command line overrides this property file.

– pmstartup.dat - settings used by automatic server startup (i.e., Windows Service)

[spacer]

Refer to Complex Installations for a real-world example of such an installation.

Mediation Listeners as Redundant Peers -- For mediation server clusters, you can also elect to either send redundant messages to each server, or to have each server process messages independently. Select the former for a high availability mediation pair.