We currently have a way to define the schema that is returned by the Usage Collectors.

They usually end up under stack_stats.kibana.plugins.[]. But we need to find a way to define and keep up to date the structure of the data that we return for every other fields (other stack_stats.[PRODUCT], and the stack_stats.kibana and root keys like version, cluster_stats, etc.).

This way we can have a contract with the remote telemetry service for all the fields that we send.

Items to cover:

• Define "static" definition of the schema
• Add functional tests that validate the payload against the schema to be aware of breaking changes in ES APIs (split definition between OSS and X-Pack)
• Check original issue where ES defines their contracts (Add aggregation list to node info elasticsearch#60074)
• ML specific use case: speak ES-ML devs to know how to best test this.
• Maybe train the devs that constantly change the contract? Include it in the wiki? Emails?

N.B.: Elasticsearch will eventually provide an API for telemetry purposes (elastic/elasticsearch#58198). But it's not happening anytime soon.