poseidon
poseidon copied to clipboard
Handle request timeout in Producer
If required_acks is > 1 and the leader is unable to get acks from replicas quick enough it will return a timeout error. We should add a setting that allows a user to indicate that they want to retry when this happens. We don't want to do this by default because retrying timeout errors is not idempotent.
Prior to implementing this investigate what other conditions could cause a timeout error. If required_acks is 0 will a leader issue a timeout if it's under high-load and the write to disk takes too long?
This would also be a good time to fix #31