C0urante commented on PR #13530: URL: https://github.com/apache/kafka/pull/13530#issuecomment-1576963027
Hmm... I think the significant divergence in behavior between the two modes is whether we retry or not. Modifying the status parts of the REST API seems less like introducing further divergence and more like honestly reporting that existing divergence to users. There's also a basic argument to be made about intuition: I can't imagine many users will be thrilled if they try to create a connector via the REST API, get back a 500 error, but then see that the connector is in the RUNNING state but hasn't generated any task configs. We don't have the same problem in distributed mode because we don't wait for the connector to start and instead, only wait for our write to the config topic to land. And it's not quite as bad if we fail to generate task configs in distributed mode since we have automatic retry logic (though this is still far from perfect). Regardless, you're correct that we can't really alter the infinite-retry logic in distributed mode without a KIP. I do think a long-term fix for this problem would involve aligning how the two modes handle this kind of failure. Would love to see that if you have the time 😄 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
