The following pre submit jobs for Local Zones are perm failing since August:
- e2e-aws-ovn-localzones: https://prow.ci.openshift.org/job-history/gs/origin-ci-test/pr-logs/directory/pull-ci-openshift-installer-master-e2e-aws-ovn-localzones?buildId=1716457254460329984
- e2e-aws-ovn-shared-vpc-localzones: https://prow.ci.openshift.org/job-history/gs/origin-ci-test/pr-logs/directory/pull-ci-openshift-installer-master-e2e-aws-ovn-shared-vpc-localzones

Investigating we can see common failures in tests '[sig-network] can collect <poller_name> poller pod logs', leading the most of jobs to not completed correctly for those failures.

Exploring the code I can see it was recently added, near August and matches with when the failures started.

It is required to tolerate the label "node-role.kubernetes.io/edge" to run pods on instances located in Local Zone ("edge nodes"). I am not sure if I am looking in the correct place, but it seems it is tolerating only master labels: https://github.com/openshift/origin/blob/master/pkg/monitortests/network/disruptionpodnetwork/host-network-target-deployment.yaml#L42

4.15.0

always

trigger the job:
1. open a PR on installer
2. run the job
3. check failed tests '[sig-network] can collect <poller_name> poller pod logs' 

Example of 4.15 blocked feature PR (Wavelength Zones): https://github.com/openshift/installer/pull/7369#issuecomment-1783699175

https://prow.ci.openshift.org/view/gs/origin-ci-test/pr-logs/pull/openshift_installer/7590/pull-ci-openshift-installer-master-e2e-aws-ovn-localzones/1715075142427611136
{  1 pods lacked sampler output: [pod-network-to-pod-network-disruption-poller-d94fb55db-9qfpz]}

E1018 22:06:34.773866       1 disruption_backend_sampler.go:496] not finished writing all samples (1 remaining), but we're told to close
E1018 22:06:34.774669       1 disruption_backend_sampler.go:496] not finished writing all samples (1 remaining), but we're told to close

Monitor jobs be scheduled in edge nodes?
How we can track job failures for new monitor tests?

Edge nodes have NoSchedule taints applied by default, to run monitor pods in those nodes you need to tolerate the label "node-role.kubernetes.io/edge"

See the enhancement for more informaation: https://github.com/openshift/enhancements/blob/master/enhancements/installer/aws-custom-edge-machineset-local-zones.md#user-workload-deployments

Looking the must-gather of job 1716457254460329984, you can see the monitor pods not scheduled due the missing tolerations:

$ grep -rni pod-network-to-pod-network-disruption-poller-7c97cd5d7-t2mn2 \
  1716457254460329984-must-gather/09abb0d6fc08ee340563e6e11f5ceafb42fb371e50ab6acee6764031062525b7/namespaces/openshift-kube-scheduler/pods/ \
  | awk -F'] "' '{print$2}' | sort | uniq -c
    215 Unable to schedule pod; no fit; waiting" pod="e2e-pod-network-disruption-test-59s5d/pod-network-to-pod-network-disruption-poller-7c97cd5d7-t2mn2" 
err="0/7 nodes are available: 1 node(s) had untolerated taint {node-role.kubernetes.io/edge: }, 
6 node(s) didn't match pod anti-affinity rules. preemption: 0/7 nodes are available: 
1 Preemption is not helpful for scheduling, 6 No preemption victims found for incoming pod.."