Reviewing CFS metrics from multiple clusters shows iptables-alerter DaemonSet pods being throttled, eg:

$ oc -n openshift-monitoring port-forward prometheus-k8s-0 9090:9090 --address='0.0.0.0' &

$ curl -ks 'http://localhost:9090/api/v1/query' --data-urlencode 'query=sort_desc(rate(container_cpu_cfs_throttled_periods_total[30m]))' |jq -r '.data.result[] | [.value[1], .metric.namespace, .metric.pod] | @tsv' | sed 's/\t/ /g'  | grep ipta
1.8054096033892513 openshift-network-operator iptables-alerter-8l749
1.5280656859472481 openshift-network-operator iptables-alerter-8lk9j
1.5202738858741267 openshift-network-operator iptables-alerter-pfjsz
0.9603678925351364 openshift-network-operator iptables-alerter-vmh25
0.8210337602333986 openshift-network-operator iptables-alerter-lfbm5
0.8030910999316458 openshift-network-operator iptables-alerter-69rlp
0.7933652001688735 openshift-network-operator iptables-alerter-km7wr
0.7764942528735631 openshift-network-operator iptables-alerter-mcccz
0.7553201867306674 openshift-network-operator iptables-alerter-dtsqg
0.7482219855274379 openshift-network-operator iptables-alerter-svvb7
0.7015708424719596 openshift-network-operator iptables-alerter-d9bkg
0.37218384643277824 openshift-network-operator iptables-alerter-sxj6j
0 openshift-network-operator iptables-alerter-24f9v
0 openshift-network-operator iptables-alerter-2xqsq
0 openshift-network-operator iptables-alerter-w2259
0 openshift-network-operator iptables-alerter-mncbc
0 openshift-network-operator iptables-alerter-m9v6v
0 openshift-network-operator iptables-alerter-4rkwl

Seen on 4.18 cluster and 4.20.0-ec.5

    100%

On the 4.20.0-ec.5 cluster
    1. Cluster is configured with workload partitioning enabled
    2. Review prometheus metrics as shown in description
    3. iptables-alerter pods consistently show cfs throttling


    

No throttling of pods