Partitioning Applications on Cluster Nodes

In a My webMethods Server cluster, partitioning is the division of applications among nodes of the cluster. By configuring cluster partitions, you can control which applications run on a node and which do not run. Partitioning is not the same as asymmetrical mode ( Component Deployment in a My webMethods Server Cluster), which causes components to be deployed to a node manually. Partitioning controls the actual execution of components.

There are several practical reasons why you might want to create partitions in your cluster:

Through the use of load balancer rules, you can route one set of users to one node and another set of users to a different node. Users on a partition have access to only the set of applications you choose to provide for them.

Because you run certain applications only within one partition, it is easier to update an application while the rest of the cluster continues to operate normally.

If a particular node experiences slow performance due to heavy traffic, users on other nodes are not affected. In the current Task Engine architecture for example, tasks typically reside in an in-memory cache. The existence of too many task instances in the cache will overrun the available memory.

There are a few things you should know about how partitions are created within a cluster:

By default, a cluster has a single partition. If there is only one partition, all nodes are part of it, regardless of whether you have explicitly listed them as part of the partition.

Each node is part of a partition. Any nodes you have not specifically listed as part of a partition become members of the last partition defined in the configuration files. See Creating and Modifying Partitions.

Each partition is associated with a phaseProvider.xml file, which governs whether portlets are enabled or disabled. If a partition is not associated with a specific phase provider file, it is associated with the default phase provider by default.