how to suspend services

[fcomte]

As the usage of Onyxia increases, the demand for cleaning services becomes crucial. We are interested in the ability to suspend services rather than completely terminating them, to avoid the risk of losing user data.

The method for detecting when a service is currently not in use falls outside the scope of this conversation, as we anticipate it might not be integrated into Onyxia, at least in the short term.

This functionality is not presently available in the Helm ecosystem, and we must decide a plan for its implementation.

Initially, we must determine:

• Whether the chart itself should be responsible for this functionality, for example, by setting a specific global value such as global.suspend = true | false.
• Whether we should identify all replicas of all stateful sets (STS) and deployments to adjust their count to 0.

We are inclined to explore the first option as it is simpler and assigns the responsibility directly to the chart.

In our SSPCloud platform, by enabling the suspension of services (and allowing users to resume them), we can quickly suspend services without concern for data loss.

[olevitt]

Implementation for solution 1 global.suspend is done API side : InseeFrLab/onyxia-api#413 (needs some refining before merging but mostly works)
Work on UI side has been defined here : #793
This is what it currently looks like to implement in a service :
Add the global.suspend property to the values.schema.json (this is what sets the service as pausable in regards to Onyxia) : e.g InseeFrLab/helm-charts-dev@968c7f9
Use the global.suspend value wherever you think it makes sense : e.g https://github.com/InseeFrLab/helm-charts-dev/blob/c9a9e160777931d6742f24d946f386e92e545129/charts/vscode-python/templates/statefulset.yaml#L10

We will have to document this behaviour for users, admins and catalogs maintainers