Workflow operator, support engineer, or developer
Read an execution trace before retrying anything
An execution is the evidence that connects the received event, selected workflow version, evaluated path, attempted actions, approvals, and result. Start there instead of guessing from a notification or provider screen.
Before you start
- At least one seeded or safe test execution.
- Permission to view workflow executions.
#Read the trace in a fixed order
- Confirm workspace, workflow, version, source event, and time.
- Read node status and timing in path order.
- Inspect only the context needed to explain each decision.
- Match connector or notification delivery evidence to the action.
- Identify whether the run completed, paused, failed, or was safely ignored.
#Replay only after checking side effects
Compare the original and replay contexts, node outputs, and branch paths. If the original may have reached an external system, verify its delivery record or idempotency key before replay. A replay is a new auditable run, not an edit to history.
#Keep durable work supervised
Queue workers execute workflows and notifications after web requests end. The scheduler starts scheduled triggers, resumes waits, performs health checks, and runs optional periodic tasks. Monitor backlog, dead letters, repeated connector failures, stuck approvals, and due waits.
#Common mistakes
- Retrying before determining whether the external action succeeded.
- Reading raw context as a substitute for authorization.
- Running queue workers without process supervision.
#Verify the result
- The source event and workflow version are known.
- External delivery status is known.
- Replay duplicate risk is resolved.
- Workers and scheduler are monitored.