Details
-
Bug
-
Resolution: Not a Bug
-
Normal
-
None
-
4.10.z
-
Moderate
-
False
-
Description
Description of problem:
When deploying a workload and setting cpu limit corresponding to the max used without limit, pod execution is slow and throttling is shown. A fix seems to be released in kernel 5.4+ but seems not to be backported in ocp 4.8.36 kernel 4.18.0-305.40.2.el8_4.x86_64 nor new ocp 4.8 releases. Source: https://access.redhat.com/solutions/5285071
Version-Release number of selected component (if applicable):
4.8.36
How reproducible:
Always
Steps to Reproduce:
1. Start a pod without limit 2. observe his cpu consuption 3. set limit to maximum observed 4. restart the pod
Actual results:
the pod is throttled response time is slow
Expected results:
the pod is not throttles while it remains in his limit.
Additional info:
Can CustomCPUCFSQuotaPeriod released in kube 1.26 help here ? If it can, do we have a roadmap to integrate it ?