After switching a MySQL master’s IP, the author observed unusually high traffic between master and slave, investigated the role of GTID, semi‑sync and the CHANGE MASTER command, and demonstrated that executing CHANGE MASTER on a lagging replica can clear the relay‑log and cause a temporary traffic surge as the missing binlog entries are re‑fetched.
This article examines common MySQL high‑availability designs—including single master‑slave, dual‑master with multiple slaves, Heartbeat + DRBD, MySQL Cluster, semi‑synchronous replication, and GTID—explaining their benefits, drawbacks, and practical considerations for reliable failover and scalability.
In a MySQL 5.7 GTID setup with dual masters, a network glitch caused the slave to switch masters, leading to the ER_MASTER_HAS_PURGED_REQUIRED_GTIDS error because the new master had purged GTIDs the slave still needed, and the article explains the root cause and step‑by‑step fix.
After switching a MySQL master’s IP, the author observed extreme master‑to‑slave traffic and investigated whether using CHANGE MASTER with auto_position=1 forces the replica to re‑fetch already executed binlogs, especially when the replica is lagging.
This article analyzes why redirecting MySQLdump output can crash the client, explores a GTID_PURGED bug that breaks AUTO_POSITION replication, examines GTID gaps caused by replicate‑do‑db filtering, and explains why TokuDB's OPTIMIZE TABLE appears to increase table size.
This article presents a series of MySQL backend improvements—including Group Commit redesign, DDL fast‑fail handling, GTID allocation refinements, persistent InnoDB auto‑increment values, replication parallelism, and key‑partition hash algorithm changes—detailing motivations, implementation steps, and performance results.
This article explores practical MySQL high‑availability solutions, covering asynchronous and semi‑synchronous replication, monitoring with keepalived or Zookeeper, failover decision criteria, GTID and pseudo‑GTID techniques, and lessons learned from real‑world deployments.
This article details MySQL performance enhancements—including Group Commit redesign, DDL fast‑fail lock handling, GTID allocation refinements, persistent auto‑increment storage, and replication parallelism improvements—illustrating the underlying mechanisms, implementation steps, and benchmark results that demonstrate measurable throughput gains.
This article examines several MySQL issues—including key partition hash algorithm changes causing data loss after upgrades, a client crash triggered by misdirected dump output, GTID_PURGED interactions that break AUTO_POSITION replication, and replicate‑do‑db filtering effects on GTID continuity—and provides concrete compatibility fixes and troubleshooting steps.
This article details MySQL performance enhancements covering Group Commit redesign, XA recovery handling, DDL fast‑fail implementation, GTID allocation optimizations, and a fix for InnoDB auto‑increment duplication, each supported by concrete test data and code changes.
This article examines how differing key‑hash algorithms in MySQL 5.1 versus 5.5/5.6 cause data loss in partitioned tables, describes compatibility fixes with ALTER TABLE … ALGORITHM=1, analyzes a MySQL client crash caused by mis‑redirected dump output, explores GTID_PURGED replication issues and replicate‑do‑db filtering gaps, and evaluates the behavior of OPTIMIZE TABLE on TokuDB tables.
This article explains MySQL's Group Commit mechanism, DDL fast‑fail enhancements, GTID allocation optimizations, fixes for InnoDB auto‑increment duplication after restart, and the evolution of replication performance from 5.5 to 5.7, including detailed code snippets and benchmark results.
This article examines why MySQL 5.5/5.6 fails to retrieve non‑null partitioned rows after a physical upgrade, how the ALTER TABLE … PARTITION BY KEY ALGORITHM=1 fix works, the cause of MySQL client crashes when dump output mixes warnings, GTID_PURGED replication bugs, replicate‑filter side effects, and the behavior of TokuDB’s OPTIMIZE TABLE.
This article explains MySQL's evolution from serialized binlog writes to group commit, introduces DDL fast‑fail and GTID allocation improvements, analyzes InnoDB auto‑increment duplication after restart, and details replication performance gains from multi‑threaded and seqno‑based parallel execution, providing code examples and benchmark results.
This article discusses MySQL high‑availability solutions, covering asynchronous and semi‑synchronous replication, monitoring with keepalived, Zookeeper, and custom agents, failover procedures using binlog positions, GTID and pseudo‑GTID techniques, and the author's practical experiences and future plans.
