This ticket revisits the current SQuaSH InfluxDB data model as implemented in DM-16300, summarizes ongoing changes and converge to a final proposal.

The mapping between lsst.verify concepts to InfluxDB concepts:

• lsst.verify package -> InfluxDB measurement
• lsst.verify metadata -> InfluxDB tag
• lsst.verify metric value - > InfluxDB field
• Jenkins pipeline runtime -> InfluxDB timestamp

Comments:

• It is not possible to combine metric values (i.e. do math operations with metric values) that are in different measurements.
• It seems more appropriate to add metadata like ci_id and squash_id as InfluxDB fields and not tags
• It seems useful to have the timestamp as an InfluxDB field as well because it is not possible to do math with timestamps in InfluxDB. See this InfluxDB community post.
• Define an appropriate time precision for the SQuaSH database in InfluxDB
• The InlfuxDB schema is flat. It means that job metadata and environment metadata will be equally stored as InfluxDB tags.
• Different execution environments (CI, LDF) can store metadata in SQuaSH preserving the original keys. At InfluxDB we can have a common data model so that we present an unified view of the different execution environment to the users (e.g ci_id and {run_id}} mapped to run_id).