Kafka Connect: influence assignment of tasks to workers

48 Views Asked by At

We run a small Kafka Connect cluster (or rather, MirrorMaker 2 in distributed mode, which uses Kafka Connect in the background). All tasks in this cluster are configured with tasks.max greater than 1. Some tasks use very little memory and CPU, while others are quite "heavy-weight". We have observed that Kafka Connect sometimes distributes tasks such that one worker gets all the "easy" tasks while another worker gets all the hard ones. For example, something we've seen quite often, is that both tasks of a resource-intensive connector with tasks.max=2 are placed on the same worker. Is there a mechanism through which we could influence task assignment to individual workers?

0

There are 0 best solutions below