API: Audit Process

[tingulfsen]

Audit Process's view of the world

-- see #29 for story ticket

pseudo code to try expressing the inventory service and the methods it will need:

Which Objects Are to Be Audited?

Could be

• by single druid
• by list of druids
  • (all known druids)
  • by online storage root
  • by archive endpoint
  • by timestamp range
  • not audited since xxx
  • by arbitrary query

For Each Object being audited:

Given information about an online Moab object read from Object Inventory Data Store:

• id
• current version
• object size
• expected location? (which storage root?)

Confirm Object EXISTENCE at Online Storage Root

Afterwards, update timestamp in database?

Discrepancy handling? ("trigger/signal recovery process")

Confirm Object SIZE at Online Storage Root

Afterwards, update timestamp in database?

Discrepancy handling? (trigger/signal recovery process?)

Confirm Object's CURRENT VERSION at Online Storage Root

Afterwards, update timestamp in database?

Discrepancy handling? (trigger/signal recovery process?)

Perform FIXITY Checking

aka "fixity workflow"

• current version only
• all versions

Discrepancy handling? Afterwards, update timestamp in database?

What ARCHIVE Copies Exist?

• For each expected archive copy:

  • do we do the same checks as above?
    If so, for anything besides existence and size, would need to unpack archived object

  • Discrepancy handling? (trigger/signal archived object recovery process?)

[tingulfsen]

These are some resources that provide background information:

• Story: Create a Standalone Inventory and Audit Process
• Distilled MVP (see the "Audit & Retrieval" section)
• Object Inventory Store Reads & Writes
• Story: Create an audit process for online Moab objects
• The MVP of SDR-PC