Demystifying MySQL Replication Crash Safety
summary
Covers concrete failure modes and exact configs that break or restore replication crash safety in MySQL 5.5–8.0: 5.5 is unsafe; 5.6 adds relay_log_info_repository=TABLE plus relay_log_recovery=1 to be safe; 5.7 relaxes defaults but GTID needs sync_binlog=1 and still needs relay_log_recovery; 8.0 ships safer defaults but changing params can remove safety. Shows how OS vs server crashes lose binlogs/relaylogs, why multi‑threaded slaves and GTID introduce races, and gives practical recovery workarounds.
AI-written summary · about
chapters (8)
- 0:00Start
- 1:56Overview of Replication
- 2:55How Does Replication Work in My Sequel
- 4:11Replication Crash Safety
- 20:55Global Transaction Id
- 27:12Failure Modes
- 32:13Multi-Threaded Replication
- 37:30Conclusion
related talks
49mMySQL Replication — Advanced Features
14mThe State of the Art of MySQL Replication
22mMySQL Replication: Latest Developments
43mHow MySQL Replication Works
53mMySQL replication setup, benefits, limitations
41mAsynchronous Replication Without Censorship
54mThe Evolution of a MySQL Database, From a Single Instance to HA With Disaster Recovery
29mEverything You Ever Wanted to Know About MariaDB Replication but Were Afraid to Ask
