I've heard the suggestion that marvel should write its data to a different cluster than its monitoring. Makes sense because if your production cluster goes south so does your monitoring.
Can anyone point me in the direction of where the steps are to do this?
Thanks -Doug
Each node from your monitored cluster should have this in the configuration:
marvel.agent.exporter.es.hosts
set to a list of nodes from the monitoring cluster.
Each node from the monitoring cluster should have this in its configuration:
marvel.agent.enabled: false
Details about the two settings here.