Bug 1711373

Description of problem:

The DNS operator was observed failing to integrate with metrics and reporting status sync errors, and the only recovery was to restart the operator. This is related to an earlier report for the ingress operator (https://bugzilla.redhat.com/show_bug.cgi?id=1687640). It seems the earlier fixes were not enough.

The DNS operator behavior based shared with the ingress operator (https://github.com/openshift/cluster-ingress-operator/pull/166) appears to have an issue that leads to the current bug. When the code suspects the Kube client discovery info is stale, the client is re-created in a possibly broken way.

It seems likely the ingress controller also suffers from the bug. Both may have gone undetected by existing e2e tests (which is another problem).

Version-Release number of selected component (if applicable):

Client Version: version.Info{Major:"4", Minor:"1+", GitVersion:"v4.1.0+d21f2fc-138-dirty", GitCommit:"d21f2fc", GitTreeState:"dirty", BuildDate:"2019-05-15T17:30:49Z", GoVersion:"go1.11.8", Compiler:"gc", Platform:"darwin/amd64"}
Server Version: version.Info{Major:"1", Minor:"13+", GitVersion:"v1.13.4+d21f2fc", GitCommit:"d21f2fc", GitTreeState:"clean", BuildDate:"2019-05-15T17:27:21Z", GoVersion:"go1.11.8", Compiler:"gc", Platform:"linux/amd64"}

How reproducible:

Create a cluster with the installer in AWS.


Actual results:

$ oc get -n openshift-dns servicemonitor
No resources found.

time="2019-05-16T16:18:27Z" level=error msg="failed to reconcile request /default: failed to ensure dns default: failed to integrate metrics with openshift-monitoring for dns default: failed to ensure servicemonitor for default: no matches for kind \"ServiceMonitor\" in version \"monitoring.coreos.com/v1\""

time="2019-05-16T21:00:22Z" level=info msg="reconciling request: /default"
time="2019-05-16T21:00:22Z" level=error msg="failed to reconcile request /default: [failed to ensure dns default: failed to get cluster IP from network config: failed to get network 'cluster': no matches for 
kind \"Network\" in version \"config.openshift.io/v1\", failed to sync operator status: failed to get clusteroperator dns: no matches for kind \"ClusterOperator\" in version \"config.openshift.io/v1\"]"

The operator never seems to recover on its own — the service monitor retries get throttled and eventually fail permanently.

Expected results:

Functioning DNS operator integrated with metrics and successfully publishing status.

Additional info: