Description of problem: Some CI failures caused by the `Undiagnosed panic detected in pod` origin test failing have been reportedly caused by prometheus-adapter. The latest report can be found in [1]. This is already the second time that this is reported since the rebase on Kubernetes 1.20. It was first though that these panics would be fixed by bumping all Kubernetes depedencies in prometheus-adapter to 1.20.0 but it seems that this panics are still occuring. Although, the panics that were first reported in [2] are quite different and seems to have been fixed by https://github.com/openshift/k8s-prometheus-adapter/pull/41 as they don't seem to occur anymore. Also, it's worth noting that this is a not just a one time flake in a thousand runs. For the past week, this CI job is responsible for 6% of all the failures and out of all this particular job failures, prometheus-adapter seems to be responsible for 71 of them out of 442 according to [3]. [1] https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/origin-ci-test/pr-logs/pull/openshift_cluster-kube-controller-manager-operator/491/pull-ci-openshift-cluster-kube-controller-manager-operator-master-e2e-upgrade/1347197449646641152/artifacts/e2e-upgrade/gather-extra/pods/openshift-monitoring_prometheus-adapter-7956dd46cf-h4d5t_prometheus-adapter.log [2] https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/origin-ci-test/pr-logs/pull/openshift_kubernetes/471/pull-ci-openshift-kubernetes-master-e2e-aws-selfupgrade/1338383564957290496/artifacts/e2e-aws-selfupgrade/gather-extra/pods/openshift-monitoring_prometheus-adapter-66d5b468c5-f6kmf_prometheus-adapter.log [3] https://search.ci.openshift.org/?search=Undiagnosed+panic+detected+in+pod&maxAge=168h&context=1&type=bug%2Bjunit&name=&maxMatches=5&maxBytes=20971520&groupBy=job One of which can be observed Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
The panic seems to have been fixed in upstream Kubernetes as part of this PR: https://github.com/kubernetes/kubernetes/pull/97820 There is currently a backport to 1.20 open https://github.com/kubernetes/kubernetes/pull/97862. Once merged, we will need to upgrade client-go in upstream k8s-prometheus-adapter to bring the fix downstream.
tested with 4.8.0-0.nightly-2021-02-09-221546, prometheus-adapter version is v0.8.3 now
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 4.8.2 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-2021:2438