We typically install Kafka on its own cluster, although it can use the existing ZK (if its not too busy). Otherwise Kafka can suffer from disk contention with the slaves and CPU/memory contention with the masters.
If this is just test cluster, installing Kafka on the slaves is probably fine. Gwen On Wed, Oct 1, 2014 at 6:54 PM, Ranjan Sen <ranjan_...@hotmail.com> wrote: > I have a 5 node cluster with one Jab tracker and another namenode; zookeeper > on job tracker, namenode and a slave (of the three). > Can you advise me on what nodes should I use for installing kafka? I am > looking for at least 2 replications. > Ranjan
