Fault-Tolerant Replication with Pull-Based Consensus in MongoDB
summary
Claims linearizable replication that tolerates any minority failure by adapting Raft into a novel pull-based data synchronization model where any replica can pull from any other. Implementation adds an updatePosition includes term check to prevent Raft-safety violations found during design, reuses the oplog and pla plus model checking, and relies on WiredTiger MVCC to support speculative execution and rollback. Supports cloning for join, up to 50 replicas for local reads, cuts cross‑DC traffic in half in one test, and recovers primaries to commit writes within seconds.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 0:46Replication Protocol
- 1:37Replication in MongoDB
- 2:46Poolbased replication
- 3:57New protocol
- 5:15Example
- 6:13Updated Position
- 6:26Updated Term
- 6:41Key Features
- 8:26Snapshotting
- 9:25Component
- 9:35Extensions
- 10:20Experiments
- 10:45Results
- 11:17Conclusion
related talks
44mHow Raft consensus algorithm will make replication even better in MongoDB 3 2
44mHow Raft consensus algorithm will make replication even better in MongoDB 3 2
14mDistributed Consensus in 15 Minutes!
51mDo Not Reimplement Distributed Consensus Algorithms
47mDistributed Systems: When Democracy and Computer Science Come Together
42mPractical Raft Consensus
32mAutomating high availability MongoDB cluster deployment and management
29mMulti Active-Active and disaster recovery with MongoDB database center