It would be nice if the field spec.template.spec.nodeSelector in the external-dns Deployment could be configurable by exposing this configuration via the ExternalDNS custom resource.
I am using the external-dns-operator in a cluster provisioned via hypershift. In this scenario the scheduler is unable to run external-dns pods because there are no master nodes available, so the pods stay in pending forever. The node selector is hardcoded here.
The workaround I'm appying is to remove the node-role.kubernetes.io/master: '' label from the nodeSelector in the Deployment object, as the operator seems to not reconcile this.

Link to original issue: https://github.com/openshift/external-dns-operator/issues/214