How do I fix an unhealthy node in YARN NodeManager?
Symptoms
When disk utilization of a core node exceeds 90%, NodeManager will set it as "Unhealthy".
Solution
- We recommend that you use Cloud Monitor and set the alarm threshold for ERM CVM disk utilization to 80-85%, so that the node disk utilization won't exceed 90%.
Configure the EMR disk utilization threshold in Cloud Monitor at the following address:
https://console.tencentcloud.com/monitor/policyTemplate
- 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.
Was this page helpful?