Asking to confirm if my reasoning is correct.
My application has 12 input partitions spread across 3 instances of applications.
Normally, I would expect number of threads to be set as 4 given that 4 partitions equals 4 tasks what indicates at most 4 threads.
However, my application maintains state in statestore with num.standby.replicas configured to 2. Given that this setting should result in processing changelog across 12 partitions, would it make sense to change threads to 12?
During restarts/rebalances I get WARN as below:
Unable to assign 1 of 2 standby tasks for task [0_0] There is not enough available capacity. You should increase the number of threads and/or application instances to maintain the requested number of standby replicas.
Is this an indication of sub-optimal configuration which may result in bad performance? (when rebalancing or context switching between tasks to catch up with changelog)