A batch workflow containing a Kafka Collection Agent will remain in RUNNING mode even if there is no data to consume, and await more data. This is a major difference from how batch workflows behave for other collection agents; they will STOP when the data from the last start cycle has been consumed. Note also that a batch will not be completed if the configured Batch Size has not been reached, or a hard coded timeout of five (5) seconds has not been reached.
A Kafka collection workflow must be stopped by other means as it is never stopped automatically. It can only stop automatically if it is aborted due to errors. Stopping a workflow can be done through:
The user interface.
Command line interface mzsh.
Operational REST interface.
A workflow stop can be triggered autromatically by evaluating MIMs like Workflow Throughput and Estimated Lag - no lag means that there is no data left to process.
Note that the 5 second timeout of waiting to complete a batch also applies when stopping a Kafka batch collection workflow.