SPIKE: investigate deleting old versions of records from SRS

When SRS records are updated, the previous version is marked as old (and the newest version is marked as actual), but the older versions are not deleted. Over time, many, many previous versions of records will build up in SRS and potentially affect performance.

If we wanted to remove the old records, how complicated would that be, and what might we need to take into consideration?

KS: there are also lots of "trash" data saved to SRS as a result of failed or stopped imports (records linked to a snapshot/jobExecution that is Cancelled or records that don't have 999 ff i UUIDs) - consider ways to clean up that data as well.

• Developers
  • If UI import log related to the previous version of a record has not yet been deleted, what would happen if an SRS record related to the log was deleted. Would it break the log? Can we prevent SRS records from being deleted if they are still connected to an import log in the UI?
  • Should we plan to keep the current and most recent previous? (would that be helpful for when we implement the rollback feature?)
  • Would there be any issues related to the various UUIDs assigned during imports?
  • Would there be any issues related to
    • quickMARC updates?
    • LDP data extracts?
    • Data export
    • OAI-PMH
  • OK for it to apply to all SRS records? all 3 MARC types, EDIFACT invoices?
  • How often to run the cleanup? Make it variable in the MOD or UI settings?
  • How much effort would this be? T-shirt sizes for UI and BE
• SMEs
  • Would this be helpful? Do we have any way to measure or estimate the impact on SRS/Import performance before implementing?
  • Any requirements?
  • Any questions or concerns?

Results of this spike

• Wiki page with design
• All required Jira stories
• T-shirt sizes for UI and MOD
• Decide if this is a separate feature or just a couple of stories