When a large number of 'catchup' is requested, the responder can easily publish the data. It may actually publish it fast enough to 'overload' the broker's data queues.
If the requester could pull the data off the broker's queue faster, this would help mitigate this.
Moving logic out of the on_message and to the gen_replica method would move the processing to a different thread and accomplish this goal.
When a large number of 'catchup' is requested, the responder can easily publish the data. It may actually publish it fast enough to 'overload' the broker's data queues.
If the requester could pull the data off the broker's queue faster, this would help mitigate this.
Moving logic out of the
on_messageand to thegen_replicamethod would move the processing to a different thread and accomplish this goal.