How do I fix an unhealthy node in YARN NodeManager?
When disk utilization of a core node exceeds 90%, NodeManager will set it as "Unhealthy".
- You are recommended to use Cloud Monitor and set the alarm threshold for EMR node disk utilization to 80–85% in order to prevent the node from being flagged as unhealthy in NodeManager when the disk utilization exceeds 90%.
Configure the EMR disk utilization threshold in Cloud Monitor at the following address:
- If the disk capacity is insufficient, you can add core nodes and perform load balancing to reduce the load of HDFS storage on the core nodes.
- Clean up the disk regularly, especially the following parts:
- Storage space of the core nodes.
- Entire storage space of HDFS.