Inference agents are configured partly in TIBCO Designer using a BAR resource, and partly using engine property (TRA) files. You can use a supplementary property file for each agent to make command line startup easier.
Inference agents do the decisioning work. Each agent has a Rete network, for processing incoming events. You can think of an inference agent as a rule session, attached to the cluster.
Agent group An agent group is formed by deploying one agent multiple times, generally on different machines. Use of agent groups requires the multi-engine features to be enabled (see
Configure Cache Cluster Properties).
Fault tolerance and load balancing are provided automatically when you use agent groups and multi-engine features. (Fault tolerance is also available without multi-engine). You can configure some agents to be inactive, available only for fault tolerance. However, all active agents load balance and if one agent fails, the others handle the load, so you may not need to keep an agent inactive.
To make an agent inactive, set the Agent.AgentGroupName.maxActive property to a lower number than the number of agents in the group. All deployed agents in addition to that number are inactive—which ones are inactive is specified by a priority number.
Agent group key (and scorecards) It is a good idea to identify each agent in a group uniquely. To do so you define an agent group key property and give each agent a unique value. For example, this property is required if your project uses scorecards. Scorecard values are not shared between agents in a group. When recovering from failure, the system needs to identify which scorecard instance belongs to which agent so that the correct value is restored.
Local storage For performance reasons, it is recommended that you disable local storage of cache data on inference agents and query agents. Instead use dedicated cache server nodes for storage of cache data. To disable local storage, set the following property set to false (it is true by default):