Journey.Scheduler.Heartbeat (Journey v0.10.44)

Copy Markdown

Sends periodic heartbeat updates for a running computation.

Runs as a linked sibling to the worker process. Uses trap_exit to receive EXIT messages when the worker exits (normal or crash), allowing immediate cleanup without waiting for the next heartbeat interval.

Exit conditions:

  • Worker exits → receives {:EXIT, pid, reason} → exits immediately
  • Worker crashes → heartbeat process receives EXIT, worker is already dead
  • Heartbeat process crashes → worker receives exit signal and dies (not trapping)
  • Computation state changes → update_heartbeat returns 0 rows → checks state
  • Deadline exceeded → marks as abandoned, exits non-normally to kill worker

Summary

Functions

run(execution_id, computation_id, node_name, interval_seconds, timeout_seconds)

Runs the heartbeat loop for a computation.

Functions

run(execution_id, computation_id, node_name, interval_seconds, timeout_seconds)

Runs the heartbeat loop for a computation.

Spawned as a linked sibling to the worker - exits when:

  • Worker exits (link causes termination)
  • Computation state changes (update_heartbeat returns 0 rows)

Arguments

  • computation_id: The ID of the computation to heartbeat
  • interval_seconds: How often to send heartbeats
  • timeout_seconds: The heartbeat timeout (used to calculate new deadline)