Problem Statement
The roadmap lists New okdp-server & okdp-ui versions under the Sandbox section, but the sandbox still deploys the current console, and there is no tracked plan yet for how the new control-plane should land here.
I spent some time running the new control plane server and new control plane UI against the current sandbox (Keycloak, current contexts and releases) to understand what is still needed before they can replace the existing control-plane.
The good news: after a few adjustments, every catalog service deploys end-to-end from the new console: SeaweedFS, Hive Metastore, Polaris, Trino, Superset, JupyterHub, Spark History Server, Airflow and OKDP Examples.
The integration required a handful of changes across the sandbox and the server, so the migration needs a bit of coordination rather than a single PR.
Proposed Solution
Land the integration as a sequence of small PRs. From what the experiment showed, the work falls into four areas:
- Additional metadata in the sandbox contexts consumed by the new server (catalog block, root-level ingress, identity provisioning settings).
- Keycloak configuration for the new UI (client scopes, redirect URIs, roles).
- Alignment of the Releases created by the new server with the KuboCD
defaultContexts mechanism used by the sandbox (server-side change).
- Packaging: new images and KuboCD packages for the server and the UI, then switching the okdp-server / okdp-ui releases to them.
I have all of this working on a fork and I'm happy to split it into small PRs,
starting wherever you prefer. Is there already a preferred sequence or an existing migration plan I should align with?
Alternatives Considered
A single big-bang PR replacing the console at once; I'd rather avoid it: the changes span the sandbox contexts, the server and the packaging, and each area deserves its own review.
Additional Context
Related: the new server's Identity API currently depends on kubauth CRDs, which are not installed on the sandbox; tracked separately in
issue 25 on control plane server.
Its outcome affects step 4 (the new console cannot fully replace the current one while that page targets kubauth).
Problem Statement
The roadmap lists
New okdp-server & okdp-ui versionsunder the Sandbox section, but the sandbox still deploys the current console, and there is no tracked plan yet for how the new control-plane should land here.I spent some time running the new control plane server and new control plane UI against the current sandbox (
Keycloak,current contextsandreleases) to understand what is still needed before they can replace the existing control-plane.The good news: after a few adjustments, every catalog service deploys end-to-end from the new console:
SeaweedFS,Hive Metastore,Polaris,Trino,Superset,JupyterHub,Spark History Server,AirflowandOKDP Examples.The integration required a handful of changes across the sandbox and the server, so the migration needs a bit of coordination rather than a single PR.
Proposed Solution
Land the integration as a sequence of small PRs. From what the experiment showed, the work falls into four areas:
defaultContextsmechanism used by the sandbox (server-side change).I have all of this working on a fork and I'm happy to split it into small PRs,
starting wherever you prefer. Is there already a preferred sequence or an existing migration plan I should align with?
Alternatives Considered
A single big-bang PR replacing the console at once; I'd rather avoid it: the changes span the sandbox contexts, the server and the packaging, and each area deserves its own review.
Additional Context
Related: the new server's Identity API currently depends on kubauth CRDs, which are not installed on the sandbox; tracked separately in
issue 25 on control plane server.
Its outcome affects step 4 (the new console cannot fully replace the current one while that page targets kubauth).