bug: KubeSteller setup scripts do not wait for KubeFlex control plane sync

[MikeSpreitzer]

Describe the bug

A KubeFlex ControlPlane is like a Kubernetes Pod or CustomResourceDefinition --- merely creating the API object is not enough, you generally have to wait for the asynchronous processing to finish. For example, kflex create ... -p SomePCH will complete successfully before any attempt is even made to find or apply the named PCH. See kubestellar/kubeflex#235 .

Steps To Reproduce

This is from code examination.

Expected Behavior

Waiting where completion of asynchronous processing is needed.

Additional Context

No response

[MikeSpreitzer]

/cc @pdettori
/cc @francostellari
/cc @dumb0002
/cc @effi-ofer

[effi-ofer]

The bottom line is that we need to wait both for the SYNCED and the READY to be set to true?

[pdettori]

That's necessary but not sufficient. "SYNCED" will only tell if the templates have been applied without errors, but not if the objects encoded in those templates are up and running. While kflex can tell if a control plane is up checking the status of the API server, it cannot tell if arbitrary objects part of a PCH are up and running. Consider the case of the OCM PCH for example, there is no general way to know that OCM is up and running unless we write use-case specific logic for it.

[pdettori]

Note that even helm (as an example) will not tell you if all objects in the release are in a READY state. It will only tell if all objects have been applied without errors.

[MikeSpreitzer]

The good news here is that our setup-kubestellar.sh script has constant (known to us developers) needs from the PCHs and so can specifically wait for that.

[ezrasilvera]

@effi-ofer Any update on this ? (CC: @pdettori )