Skip to main content

Diagnose a failed execution

A process reports failed when its body terminates abnormally. Core keeps that run in execution history so the failure can be investigated without reproducing it immediately.

1. Open the execution​

From Processes, locate the failed process and open its latest execution. Check:

  • when the run started and last changed;
  • whether it failed during normal work or while stopping;
  • the worker and host that owned the run;
  • whether earlier executions show the same pattern.

2. Read its output​

Open the execution's logs and begin with the final error lines. Official libraries capture uncaught process exceptions as error output. Look just before the exception for the last successful unit of work, external dependency, or cancellation point.

No captured lines can mean capture is disabled, the process printed nothing, the output occurred outside the execution scope, or the worker has not delivered it on a successful heartbeat yet.

3. Check the action trail​

Open Actions and filter by the process. Confirm the order of the start, heartbeat, stop, and acknowledgement events. This distinguishes a process failure from a command that has not yet reached an inactive worker.

4. Check worker connectivity​

If the worker is inactive, inspect the application's own logs and network path to Core. A Core outage does not fail a user process by itself; the worker continues and reconnects later. An authentication failure appears as 401 on the heartbeat and usually means a missing, replaced, or revoked worker token.

5. Retry deliberately​

Fix the cause or confirm the operation is safe to repeat, then choose Start. A new start creates a new execution; it does not erase the failed one. Compare the two runs to confirm the change worked.

:::tip Make failures useful Print stable milestones and identifiers, not credentials or complete sensitive payloads. The best execution log explains progress without turning Core into a copy of application data. :::