Expected behaviour

mas operator channel should not even be available in OCP but we know there is work being done to change that. We need a verified solution to migrate installations where the mas operator is installed but all the images currently running in the deployments are On-premise.

Current behaviour

2.11 images running in all pods whilst the 3scale-operator pod is using latest mas operator image (0.11.8 bundle).

This is slightly different to the issue reported in THREESCALE-10595 which concerns environments that have OCP + mas operator + mas images and this issue is for OCP + mas operator + on-premise images.

We are currently applying the same workaround to migrate out of this broken state as that which is documented here. This is primarily due to the fact we have a version of 3scale with no operator version available on the current OCP version nor is the next version available hence the manual application of the images.

Example

mas operator 0.11.8
3scale images 2.11 (amp2 on-premise)
OCP 4.13.21

To migrate from this the nexrt logical step is to go to 2.12 but it's not available in OLM on OCP 4.12 and so the 2.13 operator is installed and the APIManager CR modified to reference the 2.12 images. Thus it "forces" 3scale through the supported upgrade path, once that completes then those image references can be removed and as 2.13 is available on 4.12 the standard upgrade process will be resumed by the operator.