Clusters (multiple-node systems) — Concepts and terms

Use of multiple nodes can provide high system availability and a degree of fault-tolerance, as the failure or offline status of one node does not affect the ability of users connected to another node to continue working.

When a server node is running, the server appears as an instance of the Data-Admin-Nodes class.

You do not need to define instances of this class. The system creates an instance each time that PRPC starts on that server. (You can start and stop the engine software on each node independent of the state of other nodes.)

All the nodes share a single common PegaRULES database. The PegaRULES engine software on each node must be a compatible version — preferably an identical version — of PRPC.

Each node contains an in-memory cache of recently accessed rules that may, in multiple-node operations, contain different instances. (The correct and most recent versions of the rules are always in the database.) Periodic pulse processing synchronizes the contents of the rule cache on a node to the latest copy in the PegaRULES database. Optionally, you can associate a description of each node using Data-Admin-Node instances. The information you enter appears on the System Nodes Detail display.

Node names are limited to 32 characters.

In a multinode PRPC system, ensure that the clocks of every node are synchronized, and the clock on the server hosting the PegaRULES database is synchronized. Most vendor operating systems offer a means to accomplish this. Many calculations and decisions such as time-qualified rules, service level rules, and management reporting depend on clock settings and intervals. See the PDN article 10206 How to synchronize server clocks and database clocks with NTP for correct service level computations ion for more information.

UNIX and Linux-based systems require special configuration settings to ensure that the JVM recognizes Daylight Savings time. If not set, the times recorded in the PRPC log may not match the server clock time. See IBM Tech Note "Incorrect time stamps displayed by an application or in log files" about the TZ environment variable, and additional details for IBM WebSphere.

Other resources

These PDN articles are relevant:

• 21845 How to test whether a Process Commander node is running
• 12247 How to configure a reverse proxy server
• 20001 Load balancing a multi-node cluster
• 19914 Recommendations for load balancing and WebSphere session replication

	AES, cluster, load-balancing, horizontal cluster, node, node ID, pulse, system ID, time-qualified rules, vertical cluster
	About the System data instance About System Node data instances

Concepts