8.9. Data Store Retention Plans

Specification and plans for what the PRRTE data store — the service behind PMIx_Publish, PMIx_Lookup and PMIx_Unpublish — keeps, and for how long.

• 8.9.1. Data Store Retention: Specification
  • 8.9.1.1. Purpose
  • 8.9.1.2. Scope
  • 8.9.1.3. Definitions
  • 8.9.1.4. Retention rules
  • 8.9.1.5. A new persistence
  • 8.9.1.6. Who may remove an item
  • 8.9.1.7. Retention timeout
  • 8.9.1.8. Storage cap
  • 8.9.1.9. Observable status codes
  • 8.9.1.10. MCA parameters
  • 8.9.1.11. Departures from the PMIx Standard
  • 8.9.1.12. Where each rule is enforced
  • 8.9.1.13. Where the implementation stands
  • 8.9.1.14. Decisions taken
• 8.9.2. Data Store Retention: Design
  • 8.9.2.1. Starting position
  • 8.9.2.2. Data structures
  • 8.9.2.3. Publish
  • 8.9.2.4. Lookup
  • 8.9.2.5. Unpublish
  • 8.9.2.6. Purge and the horizons
  • 8.9.2.7. The retention timeout
  • 8.9.2.8. PMIx-side work
  • 8.9.2.9. MCA parameters
  • 8.9.2.10. Testing
  • 8.9.2.11. The one deferred piece
  • 8.9.2.12. Traps
• 8.9.3. Data Store Retention: Implementation Plan
  • 8.9.3.1. Phase 0 — the reported bug
  • 8.9.3.2. Phase 1 — the multi-node purge bug
  • 8.9.3.3. Phase 2 — removal by uid
  • 8.9.3.4. Phase 3 — the horizons
  • 8.9.3.5. Phase 4 — the timeout
  • 8.9.3.6. Phase 5 — the cap
  • 8.9.3.7. Phase 6 — documentation
  • 8.9.3.8. In parallel: PMIx
  • 8.9.3.9. Verification at each phase