Arvados

Unused or redundant columns¶

• 'id' - most tables have a default 'id' column added by rails, despite the fact that 'uuid' is the real primary key
• 'updated_at' - this seems to be redundant with 'modified_at'

Unused or redundant tables¶

• humans
• specimens
• traits

Legacy tables introduced to support PGP use case but never used. Recommended to use 'properties' for metadata and
'links' for standalone records

• keep_disks
• keep_services

Now stored 100% in the config file. Need to keep the "keep_services/accessible" endpoint but can drop the underlying tables and answer queries using only config file content.

• nodes
• jobs
• job_tasks
• pipeline_templates
• pipeline_instances

Support for the legacy "jobs" API and legacy cloud node manager, both of which are obsolete and disabled in new installations.

• api_clients

This still has an effect -- tokens are associated with clients, clients can be "trusted" or "untrusted", which controls whether they can manipulate other tokens. However this is confusing and has been a consistent stumbling point on Arvados installation & configuration.

This seems to have been originally implemented to for the login process of the legacy SSO server, where the "api_client" table represented a OAuth2 client id used by the SSO server (???). Without the SSO server it isn't clear it serves any purpose (the "trusted/untrusted" feature could be made a per-token flag, or eliminated).