Description of problem: Ther Performance Addon Operator index pod is created in the openshift-marketplace namespace without needed Workload Partitioning annotation. apiVersion: v1 kind: Pod metadata: annotations: k8s.v1.cni.cncf.io/network-status: |- <snip> k8s.v1.cni.cncf.io/networks-status: |- <snip> openshift.io/scc: anyuid creationTimestamp: "2021-06-24T20:59:17Z" generateName: performance-addon-operator- labels: olm.catalogSource: performance-addon-operator managedFields: <snip> name: performance-addon-operator-99cwx namespace: openshift-marketplace ownerReferences: - apiVersion: operators.coreos.com/v1alpha1 blockOwnerDeletion: false controller: false kind: CatalogSource name: performance-addon-operator uid: 0ac2cca5-e34a-49b3-bb6e-2f390dc3a7dd resourceVersion: "726656" uid: 56d493d8-1f05-493e-a20b-039493217f8f When workload partitioning is enabled this pod is not moved to the management cores and takes/requires CPU requests from user workload capacity. Version-Release number of selected component (if applicable): 4.8 How reproducible: always Steps to Reproduce: 1. Install PAO in SNO cluster with workload partitioning enabled 2. oc describe node shows 10m CPU request for pao pod in openshift-marketplace Actual results: openshift-marketplace performance-addon-operator-h9mk8 10m (0%) 0 (0%) 50Mi (0%) 0 (0%) 25h Expected results: openshift-marketplace performance-addon-operator-h9mk8 0 (0%) 0 (0%) 50Mi (0%) 0 (0%) 25h Additional info:
CatalogSource CR needs to be annotated for workload partitioning: apiVersion: operators.coreos.com/v1alpha1 metadata: name: performance-addon-operator namespace: openshift-marketplace annotations: target.workload.openshift.io/management: '{"effect": "PreferredDuringScheduling"}' kind: CatalogSource spec: displayName: Openshift Performance Addon Operator icon: base64data: "" mediatype: "" image: quay.io/openshift-kni/performance-addon-operator-index:4.8-snapshot publisher: Red Hat sourceType: grpc Testing underway with this change
test on 4.8.0-0.nightly-2021-09-21-220555, I try to install PAO in SNO cluster by 2 ways: 1 using default CatalogSource : redhat-operators , FYI:https://docs.openshift.com/container-platform/4.8/scalability_and_performance/cnf-performance-addon-operator-for-low-latency-nodes.html#install-operator-cli_cnf-master 2 using custom CatalogSource like Comment 1 But both don't get expected results: openshift-marketplace performance-addon-operator-xjxps 10m (0%) 0 (0%) 50Mi (0%) 0 (0%) 36m
Hi @minmli, Was this cluster deployed with Workload Partitioning enabled? If Workload Partitioning is not enabled on the cluster then the CPU requests field will not be mutated (converted over to "cores" and request set to 0). Thanks! Ian (In reply to MinLi from comment #5) > test on 4.8.0-0.nightly-2021-09-21-220555, I try to install PAO in SNO > cluster by 2 ways: > 1 using default CatalogSource : redhat-operators , > FYI:https://docs.openshift.com/container-platform/4.8/ > scalability_and_performance/cnf-performance-addon-operator-for-low-latency- > nodes.html#install-operator-cli_cnf-master > 2 using custom CatalogSource like Comment 1 > But both don't get expected results: > > openshift-marketplace > performance-addon-operator-xjxps 10m (0%) 0 > (0%) 50Mi (0%) 0 (0%) 36m
I don't know how to enable Workload Partitioning, so I think qe from Performance Addon Operator team should re-test this fix. @Gowrishankar Rajaiyan, please, thanks!
Verified on 4.8.13. [yliu1@yliu1 ~]$ oc version Client Version: 4.9.0-rc.0 Server Version: 4.8.13 Kubernetes Version: v1.21.1+a620f50 [yliu1@yliu1 ~]$ oc describe nodes | grep perfo openshift-marketplace performance-addon-operator-42n86 0 (0%) 0 (0%) 50Mi (0%) 0 (0%) 37h openshift-performance-addon-operator performance-operator-79fc9c6947-685qg 0 (0%) 0 (0%) 25Mi (0%) 0 (0%) 37h [yliu1@yliu1 ~]$ oc get pods -n openshift-marketplace performance-addon-operator-42n86 -o yaml |grep workload resources.workload.openshift.io/registry-server: '{"cpushares": 10}' target.workload.openshift.io/management: '{"effect":"PreferredDuringScheduling"}' management.workload.openshift.io/cores: "10" management.workload.openshift.io/cores: "10"
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory (Moderate: OpenShift Container Platform 4.9.0 bug fix and security update), and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHSA-2021:3759