Before opening a Bug, opening this issue as requested.
According to the docs here:
The kubernetes.container.cpu.usage.node.pct and kubernetes.container.memory.usage.node.pct are supposed to report the node allocatable usage percentages, however they actually report node capacity percentages.
The relevant code is here:
Indeed changing the the code from "Capacity" to "Allocatable" in the line above fixes the problem. I notice the Memory usage calculation is also likewise incorrect on line 632 of the above source.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.