Sascha Grunert

@aroradaman yes, we have Linux version 5.14.0-427.20.1.el9_4.x86_64

PR adding conntrack got merged, waiting for those jobs to run again to re-build the images: * [periodic-ci-cri-o-cri-o-main-periodics-setup-periodic](https://prow.ci.openshift.org/?job=periodic-ci-cri-o-cri-o-main-periodics-setup-periodic) * [periodic-ci-cri-o-cri-o-main-periodics-setup-fedora-periodic](https://prow.ci.openshift.org/?job=periodic-ci-cri-o-cri-o-main-periodics-setup-fedora-periodic) * [periodic-ci-cri-o-cri-o-main-periodics-evented-pleg-periodic](https://prow.ci.openshift.org/?job=periodic-ci-cri-o-cri-o-main-periodics-evented-pleg-periodic)

@danwinship I was hoping to get the right libraries in when installing conntrack: ``` > podman run -it fedora:41 [root@88acd4e05a73 /]# dnf install conntrack-tools Dependencies resolved. ============================================================================================================================================================================== Package Architecture Version...

The panic comes from an index out of bounds access in ``` k8s.io/kubernetes/test/e2e/framework/metrics.(*KubeProxyMetrics).GetCounterMetricValue ``` kube-proxy only complains that the rules cannot be reconciled.

@aojea here we go: https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/test-platform-results/pr-logs/pull/cri-o_cri-o/8240/pull-ci-cri-o-cri-o-main-ci-crun-e2e/1798980955491274752/artifacts/crun-e2e/cri-o-gather/artifacts/e2e.log

> can you change that to validate that the metric exist before trying to use that index? we may need to fail the test somehow later? Closing the loop: https://github.com/kubernetes/kubernetes/pull/125478

The test still fails: https://prow.ci.openshift.org/view/gs/test-platform-results/pr-logs/pull/cri-o_cri-o/8273/pull-ci-cri-o-cri-o-main-ci-crun-e2e/1801151953510600704 At least the conntrack error log is now gone, but that does not fix the nfacct issue: https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/test-platform-results/pr-logs/pull/cri-o_cri-o/8273/pull-ci-cri-o-cri-o-main-ci-crun-e2e/1801151953510600704/artifacts/crun-e2e/cri-o-gather/artifacts/kube-proxy.log

At least I can spawn a machine now using: ``` gcloud compute instances create --zone europe-west1-b --image-project openshift-node-devel --image-family crio-setup sgrunert ``` And then reproduce it using: ```go package main...

I have a `Vagrantfile` for local reproduction on centos stream 9: ```ruby # -*- mode: ruby -*- # vi: set ft=ruby : # Vagrant box for testing Vagrant.configure("2") do |config|...

> @cri-o/cri-o-maintainers, why does CRI-O even carry such a test? Shouldn't this be something for upstream Kubernetes to care about? We could think about changing the test `skip`/`focus` in https://github.com/cri-o/cri-o/blob/1ba7ce5ee9c96e22cad01c8fb87e8567bcd8fab4/contrib/test/ci/e2e.yml#L10-L30