Renku administrators to manage access to specific computing resources. The
service exposes an HTTP API for the administrators to interact with. In this
way, Renku administrators can create, update or delete resource pools, and can
add to or remove users from resource pools. Please note that a user interface
for the CRC has not been added yet but it will be added in a future release.
Currently, the only way for administrators to interact with the CRC service is
through the swagger page which can be found at the path
``/swagger/?urls.primaryName=crc%20service`` appended to the base URL of a Renku
deployment. A Renku administrator is any user who has the `renku-admin` realm
role. Assigning users to this role can be performed by the Keycloak
administrator via the Keycloak UI or API.
The CRC service also brings changes to the user interface for launching
sessions, specifically when it comes to selecting compute resources for a
specific session. With this version we have grouped different pre-set
configurations of memory, CPU, RAM and GPU in resource classes. Resource classes
are further grouped in resource pools and users are asked to select the resource
pool and class they wish to use when they launch a session rather than
separately specify memory, CPU, RAM and GPU requirements. The selection for the
amount of disk storage required is also changed but now has more freedom than
before. Users can now select disk storage with a slider that only has a maximum
limit and no pre-set steps. For projects where the users have specified resource
requests in the project settings the UI will provide hints as to which resource
classes are suitable based on the settings. When the quick launch button is used
to start a session the closest equal or greater resource class based on the
project settings will be automatically selected.
Apart from the changes needed to support compute resource access features,
support has also been added for common R file extensions.
This release also includes a hotfix for an issue with the horizontal scaling of
the core-service where users could get redirected to the wrong service instance
and subsequent requests to the core-service would fail due to partial cache
misses.
User-Facing Changes
~~~~~~~~~~~~~~~~~~~
**🌟 New Features**
- 🧑💻 **UI**: Update session start options and project settings to use compute resource pools
(`2484 <https://github.com/SwissDataScienceCenter/renku-ui/issues/2484>`_).
**🐞 Bug Fixes**
- 〽️ **UI**: Support common R file extensions
(`2638 <https://github.com/SwissDataScienceCenter/renku-ui/issues/2638>`_).
- 🛠 **Renku Core Service**: Isolates core-service cache per instance, improves cache cleanup.
(`3555 <https://github.com/SwissDataScienceCenter/renku-python/pull/3555>`_).
Internal Changes
~~~~~~~~~~~~~~~~
**🌟 New Features**
- **Infrastructure**: Add the compute resource control service.
- **Renku Notebooks**: Use the compute resource control (CRC) service.
Individual components
~~~~~~~~~~~~~~~~~~~~~~