Skip to content

Lecture 06. Replication, consistency, recovery and observability ​

Week 11 · Syllabus and preparation from the course outline; no study date or personal lecture notes have been recorded.

Business scenario ​

Delivery state must survive a single replica disconnection and become consistent again after recovery.

Topics to cover in this class ​

  1. Primary-backup log replication, sequence numbers and last-applied markers
  2. Synchronous vs asynchronous acknowledgements and failure trade-offs
  3. Replay, missing-entry detection, durable storage and catch-up after reconnection
  4. Stale reads, data convergence, lag and write acknowledgement policies
  5. Correlation IDs, structured logs, p50/p95 latency, failure recovery metrics

Learning objectives ​

  • Replicate delivery events to two backup processes.
  • Verify catch-up and convergence after a bounded outage.
  • Report observed consistency guarantees and recovery measurements.

Live demonstration checklist ​

  • Create state update, watch followers apply seq numbers, stop follower and continue.
  • Reconnect follower, replay gaps, compare final hash/record counts.

Preparation and optional written reflection (ungraded) ​

Reading focus: Watch selected parts of MIT 6.824 Primary-Backup Replication lecture; note write acknowledgement trade-offs. Optional individual reflection: If the primary acknowledges before backups receive a write, what may happen if the primary fails?

Study materials: official docs, websites and YouTube ​

Built with VitePress · Notes © 2026 [Your Name]