tested with 4.9.0-0.nightly-2021-07-07-030430,node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate is changed to node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate now
      - expr: |
          sum by (cluster, namespace, pod, container) (
            irate(container_cpu_usage_seconds_total{job="kubelet", metrics_path="/metrics/cadvisor", image!=""}[5m])
          ) * on (cluster, namespace, pod) group_left(node) topk by (cluster, namespace, pod) (
            1, max by(cluster, namespace, pod, node) (kube_pod_info{node!=""})
          )
        record: node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate

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