The CPU and RAM suggestions return question mark or none about krr HOT 9 CLOSED

Hey, @lindaorny!
This seems to be a common issue for some prometheus installations and we are trying to fix it here: #12
Could you check if that branch is working for you?

Also, none is expected, as simple strategy proposes not to place limits on CPU by default

from krr.

@LeaveMyYard I have changed branch and the CPU metrics worked for me! However, the Memory Requests and Limits are still shown as question mark.

from krr.

I used the fix-CPU-metric branch and still show the same CPU Requests --> none.
My kubernetes version is 1.24.6 in on-premise environment.

from krr.

I used the fix-CPU-metric branch and still show the same CPU Requests --> none. My kubernetes version is 1.24.6 in on-premise environment.

Sorry just check my prometheus query and shows the labels in our workload is different from the cpu_metric.py and memory_metric.py.

from krr.

@muhammad-asn can you share your labels? Perhaps it's possible to support?

from krr.

@aantn I think it's invalid case from my side. I try with native cadvisor (not robusta) installation and by default it uses (example: container_label_io_kubernetes_pod_name) as the labels so need to do relabeling from the prometheus side.

Based on this label:
google/cadvisor#3162

Sorry for the previous wrong link.

from krr.

Hey, @lindaorny! This seems to be a common issue for some prometheus installations and we are trying to fix it here: #12 Could you check if that branch is working for you?

Also, none is expected, as simple strategy proposes not to place limits on CPU by default

Seems the memory requests and limits are still returning none.

from krr.

Hey, @lindaorny! This seems to be a common issue for some prometheus installations and we are trying to fix it here: #12 Could you check if that branch is working for you?
Also, none is expected, as simple strategy proposes not to place limits on CPU by default

Seems the memory requests and limits are still returning none.

Can show the memory limit and request now! Thank you so much!

from krr.

@aantn I think it's invalid case from my side. I try with native cadvisor (not robusta) installation and by default it uses (example: container_label_io_kubernetes_pod_name) as the labels so need to do relabeling from the prometheus side.

Got it, thanks. If you find an easy solution, please share here so we can document it for others.

from krr.