You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This selector can intersect with other deployments, for instance the CodeFlare operator, which leads to issues where selection based on this selector returns more than expected.
To Reproduce
Steps to reproduce the behavior:
Install the Open Data Hub operator
Install the CodeFlare operator
Go to the Open Data Hub operator deployment details page > Pods tab
Expected behavior
Pods other than that of the ODH operator are listed.
Screenshots
Additional context
This also affect the metrics service pod selections:
thanks! got confused when checking the description of "how to reproduce".
Ah yes, the description is out-dated now with v2, and we fixed it on the CodeFlare operator side, still it'd be valuable the ODH operator fixes it, as chances are the default label selector can intersect with other deployments / pods.
Describe the bug
The operator deployment is configured with the default label selector, as generated by the kubebuilder scaffolding, i.e.:
This selector can intersect with other deployments, for instance the CodeFlare operator, which leads to issues where selection based on this selector returns more than expected.
To Reproduce
Steps to reproduce the behavior:
Expected behavior
Pods other than that of the ODH operator are listed.
Screenshots
Additional context
This also affect the metrics service pod selections:
The selector should be changed, so it's disjunct from any other deployments, ideally following Kubernetes recommended labels: https://kubernetes.io/docs/concepts/overview/working-with-objects/common-labels/
The main challenge is that deployment label selector is immutable once created, which will break the rollout deployment performed during the next OLM upgrade: https://kubernetes.io/docs/concepts/workloads/controllers/deployment/#label-selector-updates. With some more background available at kubernetes/kubernetes#50808.
A possible work-around is to renamed the operator deployment, according to OLM recommendation: operator-framework/operator-lifecycle-manager#952 (comment).
The text was updated successfully, but these errors were encountered: