Kepler metrics reporting higher values

[vprashar2929]

What happened?

When Kepler deployed using latest image on OCP 4.14(VM) it reports higher metric values.

Attached is a screenshot of a comparison between metric values reported by two Kepler instances on the same OCP cluster

Here kepler with image release-0.7.1 is deployed inside kepler-old ns and kepler with latest is deployed inside kepler-new

What did you expect to happen?

There should be no difference between the metric value produced

How can we reproduce it (as minimally and precisely as possible)?

Deploy Kepler with release-0.7.1 and latest on a VM cluster and compare the metric values produced by each.

Anything else we need to know?

No response

Kepler image tag

latest, release-0.7.1

Kubernetes version

$ kubectl version
# paste output here

Cloud provider or bare metal

VM

OS version

# On Linux:
$ cat /etc/os-release
# paste output here
$ uname -a
# paste output here

# On Windows:
C:\> wmic os get Caption, Version, BuildNumber, OSArchitecture
# paste output here

Install tools

Kepler deployment config

For on kubernetes:

$ KEPLER_NAMESPACE=kepler

# provide kepler configmap
$ kubectl get configmap kepler-cfm -n ${KEPLER_NAMESPACE} 
# paste output here

# provide kepler deployment description
$ kubectl describe deployment kepler-exporter -n ${KEPLER_NAMESPACE} 

For standalone:

put your Kepler command argument here

Container runtime (CRI) and version (if applicable)

Related plugins (CNI, CSI, ...) and versions (if applicable)

[rootfs]

@marceloamaral can you take a look? I think the delta between 0.7.1 and latest is the new build process and the refactoring

[rootfs]

@vprashar2929 can you add 0.7.2 in the same cluster too? thanks

[vprashar2929]

Sure @rootfs Here is the screenshot of when 3 Kepler instances are deployed on OCP 4.14(VM)

Kepler with release-0.7.2 is deployed in kepler-new ns Kepler with release-0.7.1 is deployed in kepler-old ns and Kepler with latest is deployed in kepler-new ns

[vprashar2929]

@rootfs @marceloamaral Just wanted to know if the latest Kepler image addresses this issue?

[marceloamaral]

@sunya-ch was the trained power model for VMs updated? Or some metric changed?

[sunya-ch]

The local weight has no update. Only the metric name and energy source name changed (cpu_time_ms, intel_rapl).

[vprashar2929]

@marceloamaral @sunya-ch @rootfs I could still see higher numbers reported by Kepler when latest release is deployed on the VM. Attached below is the comparison when Kepler with latest, release-0.7.3, and release-0.7.2 is deployed on the VM

[vprashar2929]

Still could see a higher number reported by Kepler. Comparing latest release-0.7.8 and release-0.7.2 on VM