This issue was likely caused by installing all master components (Named Node, RM, Hive Server 2, Druid, etc) on a single node. Redeploying a cluster with the master components distributed across 3 nodes resulted in a more stable cluster -- no more out of memory errors.
... View more
