Implementing agents
An agent is an internal background process operating on the server that runs activities on a periodic basis. In a multinode cluster, an agent can run on multiple nodes.
Agents are implemented through two facilities:
- Agents rules – Instances of Rule-Agent-Queue that specify the agents defined for one ruleset, the activity that each agent runs, how the agent handles the queue, and the interval at which the agent runs.
- Agent schedule data instances – Instances of Data-Agent-Queue that determine whether an agent runs.
You can modify the behavior of an agent listed in an agents rule in a locked ruleset (for example, any of the standard Pega Platform agents rules) by editing one or more of the generated agent rules or agent schedule data instances.
- New Rule-Agent-Queue – When the Master Agent detects a new
Rule-Agent-Queue, it creates the new agent schedule instance. You can
modify the following fields in the agent schedule without unlocking the ruleset:
- Pattern
- Interval
- Access Group
Restart the agent to enable the changes.
- Updated Rule-Agent-Queue – When the Master Agent detects a change in a
Rule-Agent-Queue, it updates the agent schedule instance. Existing
agents that belong to the ruleset are updated, removed, or remain unchanged.
If the Master Agent does not update the agent schedule data instance from the changes made to the Rule-Agent-Queue for that ruleset name and version, the corresponding agent schedule data instance must be deleted. The next run of the Master Agent detects the change and creates a new agent schedule data instance for that ruleset.
Use Admin Studio to manage agents and enable email notification on agent failures.
- Agents rules
An agents rule provides configuration information for the agents listed in the rule, including the trigger or wake-up interval or schedule, the activities that the agents execute, and the way that the agent queue is handled. Each agent in the list performs one request (for example, sending generated correspondence or processing assignments) by running the associated agent activity.
- Agent schedule data instances
An agent schedule data instance determines the node type an agent runs on.
- Master agents
In on-premise environments, you can configure the following master agents by updating settings in the prconfig.xml file setting.
- Associating agents with node types
To effectively use node classification, you must associate agents to node types so that only nodes with this node type assignment process the agent work requests.
- Associating an agent with a node type in a locked ruleset
Use Admin Studio to add a new association with a node type to an agent in a locked ruleset.
- Avoiding race conditions for advanced agents
To avoid race conditions associated with custom actions for advanced agents, make the actions thread-safe by ensuring that the agent runs on one node at a time.
Previous topic Managing agents Next topic Agents rules