Put Harvest Records to DB after Harvest Runner Compare

[btylerburton]

User Story

In order to create an accurate baseline for our catalog records, datagovteam wants to put all harvested records that should be created, updated, and deleted into the harvest_records DB table.

Depends upon:

• Identify approach for deriving latest set of records for a harvest source #4744

Acceptance Criteria

[ACs should be clearly demoable/verifiable whenever possible. Try specifying them using BDD.]

• GIVEN we have a harvest source in our harvest runner
  WHEN we have completed the compare step
  AND we have a list of records that should be created, updated, or deleted
  THEN those records should be posted to the harvest_records db table in order to establish a baseline for comparison against incoming records in a future harvest.

Background

Our Harvest DB should be the source of truth.

Security Considerations (required)

[Any security concerns that might be implicated in the change. "None" is OK, just be explicit here!] None

Sketch

• Wire in database interface into harvest runner
• Once compare process has run, create new DB records for all record objects that should be created/updated/deleted
  • For records to delete, a slim record containing the identifier and UUID should suffice
• Save newly created records to DB

[rshewitt]

wanted to put this somewhere

reading harvest records

• Option 1
  • Complicated query to derive the most recent successful harvest records from db
• Option 2
  • Add unchanged records to the db
• Option 3
  • Use db/solr query

[btylerburton]

open question on whether we should record unchanged records

[btylerburton]

updated ticket with dependency on #4744

@GSA/data-gov-dev-team i'm dropping this at top of Harvester 2.0 backlog, but please do review AC/Sketch for completeness

[rshewitt]

status may need to be updated to include another value in the enum. something like "pending" because we write the compare results prior to writing them on ckan. a status of "success" doesn't really make sense if the record isn't on ckan yet right? and "error" doesn't make sense in the case where no error has occurred. my thinking is that status represents the sync status.

NVM it's a nullable field

[jbrown-xentity]

Correct. Status should be one of three things (I would think):

• success
• failure
• pending
  Eventually we'll have something to clean up any items that are stuck in pending indefinitely, but that's a cleanup job to be defined at a later date.

[btylerburton]

Should we just make status a nullable value? So we wouldn't post any status until we get a success or failure.