I have a cluster with 2 Master and 3 slaves . The NN and RM are configured in HA mode . My 3 Journal Node are in : Master1 , Master2 and slave1 . My Zookeper are in : Master1 , Master2 and Slave1 . Ambari server is installed in Master1 . In master2 , there is some History service (Spark and Yarn) , monitoring tools (Graphana) and zeeplin . This is a tomporary topology because we are waiting for 3 additional edge to better manage master components . Master2 crash , and we are forced to format it . What's the best strategy to replace it ? I didn't find a lot ressource about decomminsioning Master node (Standby). Thanks in advance .
... View more