Operator Deletion

This page describes how to delete an operator and troubleshoot certain failure scenarios.

General

If the operator is managed by a Syn component, remove the component (instance) from your cluster.

Follow the [Deleting Operators from a cluster using the CLI](docs.openshift.com/container-platform/4.9/operators/admin/olm-deleting-operators-from-cluster.html#olm-deleting-operator-from-a-cluster-using-cli_olm-deleting-operators-from-a-cluster) guide from Red Hat.

Troubleshooting

If for some reason the deletion fails and you still have an operators.operators.coreos.com object which get’s recreated after deletion you can try the following (elasticsearch-operator.openshift-logging used as an example):

For all resources from github.com/operator-framework/operator-lifecycle-manager/blob/608d6bf1991cdc671a94848e884290a4bc4d9804/pkg/controller/operators/components.go get all objects with the label operators.coreos.com/<operator-object-name>='':

kubectl --as=cluster-admin get \
deployment, \
service, \
namespace, \
apiservicelist, \
apiservice, \
subscription, \
csv, \
installplan, \
operatorcondition, \
sa, \
role, \
rolebinding, \
clusterrole, \
clusterrolebinding \
-Al operators.coreos.com/elasticsearch-operator.openshift-logging=''

For each object, check if it’s still required (the case if another operators.coreos.com/… label exists):

kubectl --as=cluster-admin -n openshift-logging get sa --show-labels

If an object is still required, unlabel it. Otherwise delete it:

kubectl --as=cluster-admin -n openshift-logging delete sa elasticsearch-operator

Finally, delete the operator object:

kubectl --as=cluster-admin scale deployment -n openshift-operator-lifecycle-manager olm-operator --replicas=0
kubectl --as=cluster-admin delete operators.operators.coreos.com elasticsearch-operator.openshift-logging
kubectl --as=cluster-admin scale deployment -n openshift-operator-lifecycle-manager olm-operator --replicas=1