Skip to content

YARN Checks

These checks use the YARN Resource Manager REST API — no Hadoop client needed.

The RM URL is resolved in this priority order: 1. config.yarn.rm_url (explicit) 2. Auto-derived from ambari_url (replaces port with 8088)

YarnNodeHealthCheck

Checks that all YARN nodes are in RUNNING state.

API: GET /ws/v1/cluster/nodes

Result Condition
OK All nodes RUNNING
WARNING ≥1 node UNHEALTHY
CRITICAL ≥1 node LOST
UNKNOWN Cannot reach YARN RM

Config: 

# Optional — auto-detected from ambari_url if omitted
yarn:
  rm_url: http://rm-host:8088

YarnQueueCheck

Monitors queue capacity utilization in the YARN scheduler.

API: GET /ws/v1/cluster/scheduler

Works with both CapacityScheduler and FairScheduler queue hierarchies.

Result Condition
OK All queues below warning threshold
WARNING ≥1 queue above usage_warning_pct
CRITICAL ≥1 queue above usage_critical_pct
UNKNOWN Cannot reach YARN RM

Config: 

checks:
  yarn_queues:
    usage_warning_pct: 80
    usage_critical_pct: 90