Triage a Failed Connector

Triage a Failed Connector

Diagnose a connector failure down to the task and restart it.

1. List the connectors on the worker with their status expanded, giving the caller the fleet-wide picture before drilling into one connector.
2. Read the status of the target connector. The connector-level state tells the caller whether the connector itself failed or whether it is running with failed tasks underneath it.
3. List the tasks belonging to the connector along with their per-task configuration, so the caller can see how the work was divided.
4. Read the status of the individual task, capturing the state, the worker that owns it, and the stack trace that explains the failure.
5. Restart just the failed task, which is the least disruptive recovery action because the connector's other tasks keep running.
6. Restart the connector and all of its tasks, restarting only the failed ones so healthy tasks are not needlessly interrupted.
7. Re-read the connector status to confirm the restart actually recovered the connector rather than assuming the restart call was enough.

What you'll need to have ready

• connectorName: The name of the connector to triage.
• taskId: The task id to inspect in detail. Defaults to the first task, which is task 0 on a single-task connector.

For your developer

Arazzo workflow apache-kafka:triage-failed-connector, 7 operations: listConnectors → getConnectorStatus → listConnectorTasks → getTaskStatus → restartTask → restartConnector → getConnectorStatus. Specs on APIs.io.