Move Dashboard CRDs to odh-dashboard repo

[VaishnaviHire]

ODH Dashboard CRDs are bundled in the operator bundle. Use crd folder from odh-dashboard to deploy required CRDs.

This will allow following features:

• Installing dashboard CRDs on when managementState is Managed for odh-dashboard
• Installing CRDs depending on the dashboard manifests defined in odh-manifests/odh-dashboard repo

[andrewballantyne]

Installing CRDs depending on the dashboard branch.

Would this not be "just read the location based on where you install the dashboard"? I don't think you need to hunt down different branches. Code and Manifests are co-located. Installing a CRD from a feature branch, but code from another branch will leave you in an unhappy state.

[VaishnaviHire]

Installing CRDs depending on the dashboard branch.

Would this not be "just read the location based on where you install the dashboard"? I don't think you need to hunt down different branches. Code and Manifests are co-located. Installing a CRD from a feature branch, but code from another branch will leave you in an unhappy state.

Yep, thats right. I will update the wording there.

[zdtsw]

By having this change, user will not be able to config from operator UI, e.g odhdashboardconfigs CR any longer, instead they need go to API Explorer look for OdhDashboardConfig's instance.

[zdtsw]

By having this change, user will not be able to config from operator UI, e.g odhdashboardconfigs CR any longer, instead they need go to API Explorer look for OdhDashboardConfig's instance.

after discussion, this is not a problem.

[lugi0]

Verified this in the 2.3 RC for RHODS.
Installing on a fresh cluster without enabling the dashboard component results in none of the CRDs getting created (odhquickstarts, odhapplications, odhdashboardconfigs, odhdocuments, acceleratorprofiles).
Once the component is enabled, the CRDs (and CR Instances) are created.
The CRDs/CRs are not shown/tracked under the operator details page in OCP.

The only weird thing I've noticed is that when I installed RHODS and created a DSC with dashboard disabled, it took an unusually long time to "come to life" and deploy all other components before going into Ready state - approx. 5 minutes, compared to the usual <1 minute. Could anyone try to reproduce this in a different cluster?

[lugi0]

Additionally, all CRs are removed once dashboard is disabled, with the exception of odh-dashboard-config - is this ok? @andrewballantyne @VaishnaviHire @zdtsw

[andrewballantyne]

CRs managed by the operator (ISVs, etc) are definitely to be taken away. The CRDs should all remain, so any resource not managed by the operator still will exist.

The OdhDashboardConfig is not managed by the Operator simply because we need to modify it without the Operator's involvement. It should absolutely stay based on OpenShift Console standards RE CRDs.