Heavy CPU utilisation causes nodes failure

We have cases when the couchbase nodes start to consume 100% CPU. Usually it due to compaction but rebalance does it too.
This causes a connection close and server fails (and after few minutes come back a live).
Is there a way to configure couchbase to use less CPU?

We are running a couchbase 3 cluster of 4 r3.large ec2 instances with general purpose EBS attached to it as the data path.


Can you share a little more info about the workload on these instances? Are their views defined? Are there any other processes competing for resources on the instances?

One value eviction bucket with ~50m documents and 200ops. No views.
One value eviction bucket with 2m documents and 20ops. 1 view to find the oldest ttl.
One value eviction bucket with 50k documents and 20ops.
One full eviction bucket with 200 documents that are updated using the increment method with 1 view on it.

No other processes competing with Couchbase.