Description of problem: We have added some taints to nodes with the effect "NoExecute" which evicts any pod without a matching tolerance. The node-exporter daemonset is not able to be updated with new tolerances because the cluster-monitoring-operator reverts it back to its out of the box config. The current tolerance applied to node-exporter does not prevent the pod from being evicted if any taints are applied. Version-Release number of selected component (if applicable): OCP 3.11.59 How reproducible: Always, part of default configuration. Expected results: node-exporter daemonset should come configured with a tolerance that prevents taints on a node to evict them. The sdn daemonsets accomplish this by having the following tolerance. tolerations: - operator: Exists Additional info:
The upstream pull request has been opened to tolerate any taint: https://github.com/coreos/prometheus-operator/pull/2429. Once merged this will need to trickle down to the releases. It's unclear whether we can backport this to 3.11, but this will definitely land in 4.0.
https://github.com/openshift/cluster-monitoring-operator/pull/269 pulls the required changes into cluster-monitoring to fix this.
The changes have been merged. Moving to modified.
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-2019:0758