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
In the operator-controller upgrade-e2e, we discovered an issue that causes the catalogd controller to leave stale conditions and fields set on a ClusterCatalog after an upgrade.
If a ClusterCatalog is serving and up-to-date prior to an upgrade, and then a catalogd upgrade occurs, the new catalogd sees that "everything looks good", and then immediately returns from reconcile.
Unfortunately everything is not good. In a transition from 0.29.0 to 0.30.0, one example of a thing that we fail to do is change the Unpacked condition type to Serving.
In order to fix this, we need to always be updating the status based on the known state of the catalog, no matter which code path we take through our reconcile function.
The text was updated successfully, but these errors were encountered:
Taking a look at this now. I'd like to fix this, make a new release, and then bump operator-controller to that version (skipping 0.30.0) to make sure that the upgrade goes more smoothly.
In the operator-controller upgrade-e2e, we discovered an issue that causes the catalogd controller to leave stale conditions and fields set on a
ClusterCatalog
after an upgrade.If a
ClusterCatalog
is serving and up-to-date prior to an upgrade, and then a catalogd upgrade occurs, the new catalogd sees that "everything looks good", and then immediately returns from reconcile.Unfortunately everything is not good. In a transition from 0.29.0 to 0.30.0, one example of a thing that we fail to do is change the
Unpacked
condition type toServing
.In order to fix this, we need to always be updating the status based on the known state of the catalog, no matter which code path we take through our reconcile function.
The text was updated successfully, but these errors were encountered: