Fork Order for all Subagents of a SubagentSelection

Current Situation

• With JS-1954 Subagent Clusters are organized for one of following scheduling modes:
  • fixed-priority: always use the first Agent and in case of unavailability only switch to the next Agent
  • round-robin: execute each next job with the next Agent
• With JS-1985 Subagent Clusters support the following scheduling mode:
  • next-priority: use the next available Subagent and continue subsequent job executions caused by the same order on the same Subagent.

Desired Behavior

• New scheduling mode for Subagent Cluster: all-priority
  • An order is split to child orders that will start simultaneously with all Subagents of the Subagent Cluster.
    • The Subagent Cluster is assigned to a ForkList/Join Instruction in a workflow that creates as many child orders as Subagents are configured with the Subagent Cluster.
    • For disabled Subagents no child orders are created.
  • Any subsequent job executions caused by the same child order are executed on the same Subagent (fixed-priority).
• If the Subagent is unavailable or becomes unavailable then the child order remains
  • in a waiting state if unavailability occurs between job executions,
  • in a failed state if unavailability occurs during job execution.
• A waiting child order
  • continues once the Subagent becomes available again,
  • can be suspended/resumed or cancelled by the user.
• A failed child order
  • can be resumed or cancelled by the user.
• User operations to resume and to cancel a child order do not change the child order's assignment to a Subagent.