LowNodeUtilization not working in k8s 1.9 (GKE non-alpha cluster) about descheduler

[ptagr]

The CPU and Mem percentages reported in the logs are most definitely wrong. The one that kubectl top shows is the correct one.

[aveshagarwal]

@ptagr Thanks for reporting this. I am looking into this. meanwhile could you send "kubectl describe node" for the node gke-asia-northeast1-std--default-pool-36ae422e-32s4?

[aveshagarwal]

"I0508 07:06:40.392782 1 lownodeutilization.go:69] No node is underutilized, nothing to do here, you might tune your thersholds further"

@ptagr the reason, the descheduler did not evict any pod as it did not find any nodes that are under utilized. You should increase your "nodeResourceUtilizationThresholds: thresholds:" to some higher values.

[ptagr]

@aveshagarwal I get it but the descheduler is getting the wrong values for the current CPU and Mem utilization. I have multiple sources that report about 1% CPU usage for the nodes and from the logs, the descheduler says its 46% utilized.

Stark difference between what is reported in the logs vs reported by kubectl top

lownodeutilization.go:144] Node "gke-asia-northeast1-std--default-pool-36ae422e-32s4" is over utilized with usage: api.ResourceThresholds{"cpu":45.972309628697296, "memory":55.872961663211015, "pods":7.2727272727272725}
vs

gke-asia-northeast1-std--default-pool-36ae422e-32s4 199m 1% 12555Mi 12%

[ptagr]

I think I got it completely wrong. CPU Usage is not the actual usage but the (CPU Request) Usage and the same for memory! Let me look over it again.

[aveshagarwal]

@ptagr Descheduler like scheduler does its computation based on resource requests not actual utlization on any node.

[ptagr]

this makes sense. the describe yields:

  (Total limits may be over 100 percent, i.e., overcommitted.)
  CPU Requests  CPU Limits   Memory Requests  Memory Limits
  ------------  ----------   ---------------  -------------
  6305m (39%)   6200m (39%)  41200Mi (42%)    41310Mi (42%)```

got confused in my head 

[ptagr]

Closing this issue.