We are running a 3-master, 1-client, 10-slave elasticsearch cluster with 5 shards and 1 replica each. We've been monitoring it with marvel for a while now and I've seen the cpu usage steadily increasing over a period of months, and at this point I typically see our cpu ranging from 70%-99% on our slave nodes.
Adding more slave nodes SHOULD reduce the load across each node, correct?? I'm just uncertain because I've added more but I'm not sure if its had any effect so far. I'm trying to figure out the best way to guarantee that we have an efficiently-sized cluster.