-
Bug
-
Resolution: Not a Bug
-
Major
-
None
-
4.16.0
-
Important
-
None
-
CNF Compute Sprint 259
-
1
-
Rejected
-
False
-
-
-
Description of problem:
While following the official docs on "Deploying the NUMA-aware secondary pod scheduler" and applying the NUMAResourcesScheduler config, the resulting secondary-scheduler pod is stuck in a CrashLoopBackOff state. Default pod logs report the following error: ``` 2024-08-21T09:16:31.222353645Z Error: strict decoding error: decoding .profiles[0].pluginConfig[0]: strict decoding error: decoding args for plugin NodeResourceTopologyMatch: strict decoding error: unknown field "cache.informerMode" ``` followed by a "Usage" dump
Version-Release number of selected component (if applicable):
OpenShift 4.16.0 cluster managed by ACM numaresources-operator.v4.16.1
How reproducible:
Unknown
Steps to Reproduce:
1. 2. 3.
Actual results:
secondary-scheduler pod fails to start
Expected results:
secondary-scheduler pod starts
Additional info:
PerformanceProfile with ".spec.numa.topologyPolicy: single-numa-node" was already applied to the target machineconfigpool, meaning kubeletconfig and tuned profiles were already generated