+++ This bug was initially created as a clone of Bug #1781283 +++ Noticed this in skew tests, but it can also happen when the CVO gets rescheduled onto a different machine during an upgrade. https://prow.svc.ci.openshift.org/view/gcs/origin-ci-test/logs/release-openshift-origin-installer-old-rhcos-e2e-aws-4.4/11 We should be setting resource requests, looking at e2e runs it's ~ 10m CPU and 50Mi mem, roughly 1Mi disk.
Run upgrade from 4.3.0-0.nightly-2019-12-19-185024 to 4.3.0-0.nightly-2019-12-19-204222, checked extract job runs in burstable qos: # oc get pod/version--np5tt-wrpmp -o json|jq .spec.containers[].resources { "requests": { "cpu": "10m", "ephemeral-storage": "2Mi", "memory": "50Mi" } } # oc get pod/version--np5tt-wrpmp -o json|jq .status.qosClass "Burstable" But checked the 4.2-4.3 upgrade ci test result, still failed with [Feature:Platform][Smoke] Managed cluster should ensure control plane pods do not run in best-effort QoS. https://search.svc.ci.openshift.org/?search=pods+found+in+best-effort+QoS&maxAge=168h&context=2&type=all So i think we need backport the fix to 4.2 to let the ci test pass, so clone a bug to 4.2 and verify this bug.
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, 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/RHBA-2020:0062