CA incorrectly reports 0% utilisation for nodes that have running pods

[aaronfern]

Which component are you using?:

/area cluster-autoscaler

What version of the component are you using?:

Component version: v1.32

What environment is this in?:

Our own custom cloud-provider, but this error is part of the core CA logic

What did you expect to happen?:
When CA calculates resource utilisation to determine underutilised nodes, it should reliably be able to calculate this utilisation value.

What happened instead?:
We see that at random time periods that CA calculates "0% utilisation". See logs below

2025-11-04 17:12:00 | {"log":"Node <node_name_1> is underutilized: memory requested (0% of allocatable) is below the scale-down utilization threshold","pid":"1","severity":"INFO","source":"eligibility.go:168"}
2025-11-04 17:12:00 | {"log":"Node <node_name_2> is underutilized: memory requested (0% of allocatable) is below the scale-down utilization threshold","pid":"1","severity":"INFO","source":"eligibility.go:168"}
2025-11-04 17:11:04 | {"log":"Node <node_name_3> is underutilized: memory requested (0% of allocatable) is below the scale-down utilization threshold","pid":"1","severity":"INFO","source":"eligibility.go:168"}

I've added diagnostic logs to the CalculateUtilizationOfResource() function here and it's clear that when this happens, nodeInfo.Pods() does not return anything which means that podsRequest ends up being 0 which is returned.
I've verified that the nodes in question do in fact contain running pods (which are not daemonset/mirror pods nor are they terminating)

I've noticed that this "0% utilisation" happens randomly during that day and does not seem to follow any pattern. Moreover, I also did not notice any error logs around the time this occurs

How to reproduce it (as minimally and precisely as possible):

I'm not sure this is relevant. I see this randomly and have not found a way to reproduce this

Anything else we need to know?:

Any pointers to what else I can look at would be helpful