Catalogsource pods are currently run as bare pods on the cluster and managed by OLM directly. The fact that catalogsource pods are not managed by a built-in controller causes several issues when nodes running the catalogsource pods are drained/terminated and the pods either block the node from draining or are not brought back up successfully:

• https://github.com/operator-framework/operator-lifecycle-manager/issues/2709
• https://github.com/operator-framework/operator-lifecycle-manager/issues/2666
• https://github.com/operator-framework/operator-lifecycle-manager/issues/1514

The customer would benefit from having the catalog pod be backed by a built-in controller as it would help with managing the node lifecycle successfully and not run into issues with catalogsource pods.

Having the catalog source pod backed by a built-in controller was considered several times but never prioritized to be implemented. Looking at the number of users affected by this issue it may make sense to reprioritize and include this in the OLM roadmap.

Affected components:

• OLM (Catalog operator)
• Catalogs (Catalog source pods)