Description of problem:

When setting up a model to be served through the "model serving" feature, we need to give it a data connection which contains the model being served.
When creating the data connection, a secret is created in the namespace; editing this secret (or editing the data connection through the UI) does not however restart the model serving pod.
This means that changes made to the data connection after the first time it's set up (e.g. changing the bucket name or model path) will not be picked up by the model server pod, unless the pod is restarted manually.
By comparison, modifying the inferenceservice CR will restart the pod; I would expect the same behaviour when modifying the secret.

Prerequisites (if any, like setup, operators/versions):

Latest model serving build (quay.io/anishasthana/rhods-operator-live-catalog:1.20.2-5897)

Steps to Reproduce

1. Install build
2. set up a DSP with data connection
3. set up a model server using the project/data connection created earlier
4. modify the data connection (or related secret in OCP)
5. verify that model serving pod is not restarted and changes in data connection are not picked up

Actual results:

model serving pod not restarted after modifying data connection

Expected results:

model serving pod restarts and picks up the changes to the data connection

Reproducibility (Always/Intermittent/Only Once):

always

Build Details:

Latest model serving build (quay.io/anishasthana/rhods-operator-live-catalog:1.20.2-5897) 

Workaround:

modify InferenceService CR to trigger a pod restart, or restart it manually (e.g. scale the deployment down and wait for it to scale back up)

Additional info: