When we can't find an operator's catalog source, we show the message "CatalogSource was removed" in a few places like the "Installed Operators" table. This message can be misleading, however, since it can appear during operator install and perhaps other times when no one has deleted a catalog source. We should replace this with a more general message like "CatalogSource not found."
I'm wondering whether the logic we used to get catalogSource correct?
When user create a global wide catalogsource and subscribe one operator into user namespace, the message `CatalogSource not found` is always there during installation early phase ~~ however I doubt if `CatalogSource not found` message is correct since the operators are coming from a global wide catalogsource and the catalogsource does exist
For example, I subscribe prometheusoperator.0.47.0 which is coming from CatalogSource 'Community Operators' in openshift-marketplace namespace, it will first show Unknown, then 'CatalogSource not found' message together with 'Cannot update'
$ oc get catalogsource -n openshift-marketplace
NAME DISPLAY TYPE PUBLISHER AGE
certified-operators Certified Operators grpc Red Hat 9h
community-operators Community Operators grpc Red Hat 9h
custom-console-catalogsource-yanpzhan Custom Console Testing Yanping grpc OpenShift QE yanpzhan 33m
redhat-marketplace Red Hat Marketplace grpc Red Hat 9h
redhat-operators Red Hat Operators grpc Red Hat 9h
$ oc get packagemanifest | grep prometheus
prometheus-exporter-operator Community Operators 9h
prometheus Community Operators 9h
Sam, could you help check my comment above?
Assigning back for confirmation
FWIW, IHAC trying to update rhacs-operator via the OpenShift Web Console and they're getting "Cannot Update - CatalogSource was removed" (slightly different wording, perhaps the same issue?). They're on OCP 4.7.
Here's what we know:
* By comparison, openshift-gitops was successfully updated just before trying to update rhacs-operator
* The rhacs-operator subscription shows a CatalogSource reference to redhat-operators
* The redhat-operators CatalogSource exists in openshift-marketplace
* The image referenced by redhat-operators LGTM
* Although registry.redhat.io was recently down for about 30 min, this was occurring before then
* They've never tried updating rhacs-operator before
* They are not using RHACM nor its policies on this cluster
A ticket is being filed. Meanwhile I want to add this info here in case it is of help during this BZ's investigation.
Did not get a chance to fix this sprint, will address in a future sprint.
OpenShift has moved to Jira for its defect tracking! This bug can now be found in the OCPBUGS project in Jira.