Nodes resource mismatch

[shubha21m]

I have install descheduler on my cluster. kubectl top node command gives different output than descheduler.
Descheduler shows All nodes are underutilized. but when used kubectl top node command, its show many off nodes are on 100% utilization.

How to fix this bug. Please guide me.

[a7i]

/kind support

@19bcs2405 Per README:

LowNodeUtilization Spreads pods according to pods resource requests and node resources available

HighNodeUtilization Spreads pods according to pods resource requests and node resources available

As the description suggests, these strategies do not consider actual utilization but rather resource allocation (poor naming). kubectl top looks at actual resource utilization.

There are a few open PRs that look at utilization such as:
#1092
#1087

Given that they behave correctly, I am going to close this issue but feel free to reopen if you think there is a bug.

[a7i]

/remove-kind bug

[shubha21m]

@a7i Is there any way to descheduler get exact Resources like kubectl top node.

[a7i]

Not currently, but that's what these PRs do (as new Plugins):
#1092
#1087