Restarting a stream once it's entered an error state.

[MeeJosh]

I think the answer here is no, but figured I'd check in case I've missed something obvious.

Currently, when one of our streams ends up in an error state - we rely on a health check to notice and then k8s kills the pod and restarts it.

This works, but we want to know if we're missing a trick.

Is it possible to restart the stream internally via code without killing off the application?

It looks like it's alluded to in this change using REPLACE_THREAD in the Confluent libraries:

https://issues.apache.org/jira/browse/KAFKA-10810

[LGouellec]

@MeeJosh
No indeed, today you can customise your application to continue in case of failure.

What does it mean ?

If an exception is thrown, the thread will catch the exception, close the consumer, abort the current transaction, and restart processing from the last message committed. But the thread is still the same, just closing the consumer, and restart the processing from the last time.

[MeeJosh]

Sorry - to clarify. Is there a built in way in Streamiz to handle this?

[LGouellec]

If you don't want to fail when an exception occurs, you can override the InnerExceptionHandler

config.InnerExceptionHandler += (e) =>
            {
                return ExceptionHandlerResponse.CONTINUE;
            };

[MeeJosh]

Thanks - works a treat!