Description of problem: When checking the status of cluster managed daemonsets and deployments, some report that not all pods are available. However, it appears that all pods are actually running without issue. Some of these "out of sync" daemonsets/deployments appear to cause their associated operators to go into a degraded state. This issue came up without any changes or known activity in the cluster. How reproducible: Uncertain Actual results: Some cluster operators reporting as degraded due to the out of sync deployments. Expected results: Deployment state would be accurate.
Looks like this [1] upstream issue, which is still active. 1. https://github.com/kubernetes/kubernetes/issues/53023
Rolling the daemonset seems to mitigate the issue for now. ``` oc rollout restart ds [ds name]. ```
Bug 1804717 might help with this. Or it will at least maximize the benefit of a fix to Kube's Deployment controller.
1804717 works around the problem for a single DaemonSet but the problem still exists for every other DaemonSet. If we are not going to fix it in kubelet then we need to get rid of every DaemonSet in OCP...
There are patches in later releases fixing this issue in 4.2. If this issue is found again in later release, please open a new bug.