Restart a connector

You can restart a newly created connector and all of its tasks, or selectively restart only the failed tasks, while preserving configuration and committed offsets. This is useful when a connector has failed because of problems such as misconfigured parameters or temporary network connectivity issues. You can also restart connectors that have no failed tasks to recover from transient issues or to pick up changes in external systems or dependencies.

The restart operation is asynchronous. When you submit a restart request, the service returns a connector operation ARN that you can use to track the operation's progress.

Before you restart a connector, make sure that the following prerequisites are met:

• The connector must be in the RUNNING state.

• No other lifecycle operation can be in progress on the connector.

Restarting a connector using the AWS Management Console

1. Open the Amazon MSK console at https://console.aws.amazon.com/msk/.

2. In the left pane, under MSK Connect, choose Connectors.

3. Select the connector that you want to restart.

4. Choose Actions, then choose Restart connector.

5. (Optional) Select Restart only failed tasks to restart only the tasks that are in a FAILED state. This option is available only for connectors using Apache Kafka Connect version 3.7 or later.

6. Choose Restart.

7. You can then monitor the current state of the operation in the Operations tab of the connector.

Restarting a connector using the AWS CLI

To restart a connector and all of its tasks, run the restart-connector command with the connector's ARN:

aws kafkaconnect restart-connector \
   --connector-arn <connector_arn>

To restart only the tasks that are in a FAILED state and leave healthy tasks running, add the --only-failed-tasks option:

aws kafkaconnect restart-connector \
   --connector-arn <connector_arn> \
   --only-failed-tasks

The preceding commands return the connector ARN and a connector operation ARN. To track the restart operation, use the returned connectorOperationArn with the describe-connector-operation command to check the operation status:

aws kafkaconnect describe-connector-operation \
   --connector-operation-arn <operation_arn>

A restart operation can be in one of the following states.

State	Description
RESTART_IN_PROGRESS	The restart operation has been accepted and is executing.
RESTART_COMPLETE	All affected tasks have been successfully restarted.
RESTART_FAILED	The operation was terminated before completion.

Note

A restart operation can complete successfully (RESTART_COMPLETE) even if a task transitions to a FAILED state after restarting. The operation state reflects whether the restart was executed, not whether the task is healthy after the restart. Use the connector's state and the following CloudWatch metrics to monitor post-restart task health: RunningTaskCount, ErroredTaskCount, TaskStartupSuccessPercentage, SinkRecordReadRate (sink) or SourceRecordPollRate (source), and RebalanceTimeSinceLast.

Considerations

• Restart behavior depends on the connector's Apache Kafka Connect version. On Apache Kafka Connect 3.7 and later, restart operations apply to the connector and its tasks by default. To restart only the tasks that are in a FAILED state, use the --only-failed-tasks option. You cannot select specific individual tasks to restart. On Apache Kafka Connect 2.7.1, restart operations only restart the connector instance and not its tasks. If you specify --only-failed-tasks, the request is rejected with a 400 error.

• The restart preserves all connector configuration and committed offsets.

• While a restart is in progress, you cannot perform other lifecycle operations (update, delete) on the connector.

• Use DescribeConnectorOperation with the returned connectorOperationArn to track whether the operation is in progress, complete, or failed.

To use the MSK Connect API to restart a connector, see RestartConnector.