I'm a bit late with my response, but it still may be helpful for people searching to solve a similar problem. This constant rebalancing will happen when two kafka connect clusters with the same group id are connecting with the same kafka cluster. As described above, the problem is solved by choosing a different group id. When this happens, there is probably a different connect cluster (even if it only has 1 replica) running with the same group id. Maybe a first try that you forgot to stop on your docker host?