Is it possible to use a task assignment strategy other than round-robin for connect? Say I’m running distributed mode across two systems, where one system has 2x the CPU/memory of the other, I’d like to give the larger system 2x the tasks. Is that doable?
I’m not aware of any additional “Assignore” algorithms for connect. Assigning and rebalancing is a fairly complex topic if you’d like to read this KIP which goes into some details on the reasons for the current implementation: https://cwiki.apache.org/confluence/display/KAFKA/KIP-415%3A+Incremental+Cooperative+Rebalancing+in+Kafka+Connect
I realized shortly after saying this that it would be an effective work-around for a simple case like I suggested to just run 2 instances of connect on the larger system and 1 on the smaller, but thanks for the reading!