Feature #15397
Updated by Peter Amstutz 3 months ago
There are several legacy tables "humans" "specimens" and "traits". These were added extremely early on with the best intentions of supporting the PGP use case but as far as I know they have never used for their intended purpose. They should be deprecated and removed to de-clutter the API. We are also dropping the hosting git repository support (the "repositories" table). This means we can delete @arvados/services/githttpd@. We would also like to remove and stop publishing anything related to the jobs API, e.g. "jobs" and "job_tasks", "pipeline_instances", "pipeline_templates", "nodes". We should also get rid of the "keep_disks" table. We should remove some unused fields from "api_client_authorization" response: @default_owner_uuid@, @api_client_id@, @user_id@ (the @_id@ fields may need to remain internally but should not be published by the API because they are not usable with any other API calls). We should remove redundant resource methods from the Discovery Document (the Python and R autogenerated APIs already filter these out, so they are undocumented and nobody should be using them): * "show" (synonym for "get") * "index" (synonym for "list") * "destroy" (synonym for "delete") There's a "Managed" section under SLURM that has options related to the obsolete slurm-on-cloud configuration. # Announce deprecation/removal in future version (already done) # A few integration tests use these APIs because they are generic resources with no business logic. These tests need to be updated to use a different resource or otherwise perform test a different way. # Delete models, controllers, tests, and routes from API server # Delete from documentation These should not appear in the discovery document or the auto-generated python docs. The auto-generated SDKs (e.g. the R SDK) should also be updated. Another thing to get rid of: support for legacy component-specific config files.