My cluster has around 2m docs divided into 5 nodes .
The only thing that is running is an XDCR to Elasticsearch (1.7.5)
All the nodes RAM usage is under 70-80% at rest. Once I start the XDCR a few of the cpus jump to 90-100% and after a few minutes the nodes start to fail.
The logs don’t state anything I can make sense of.