Tencent's Database R&D team supports internal services such as WeChat Pay, WeChat Red Packets, Tencent Advertising, and Tencent Music, and provides external support on Tencent Cloud for TencentDB products like CynosDB, CDB, and TDSQL. This public account aims to promote and share professional database knowledge, growing together with database enthusiasts.
Change Buffer in MySQL InnoDB (formerly insert buffer) caches modifications to secondary index pages when they are not in the buffer pool, using a B‑tree structure to batch‑apply inserts, delete‑marks, and deletes while avoiding index page splits, merges, and ensuring purge safety.
The article explains why a seemingly non‑conflicting UPDATE on a secondary index can cause lock‑wait alerts under RC isolation, detailing the missing end‑range condition push‑down, the InnoDB row‑search flow, semi‑consistent read behavior, and differences between primary‑key and secondary‑index locking.
The article details the refactoring of memory monitoring in MySQL 8.0.28, introducing a new PFS metadata structure for precise tracking of memory allocations and deallocations across various components like Buffer Pool and user variables.
This article explains why timestamps in MySQL binary logs can appear out of order, detailing how event times are derived from the thread start time, how long‑running transactions and binlog rotation affect ordering, and provides code examples and an experiment to illustrate the behavior.
This article explains how MySQL 8.0’s clone plugin changes GTID handling by storing GTID information in InnoDB undo logs, detailing the allocation of undo log segments, recording GTIDs in undo log headers, persisting them to the gtid_executed table, and coordinating purge operations.
This article investigates why S‑type GAP locks appear in deadlocks under the READ COMMITTED isolation level in MySQL InnoDB, detailing the transaction workflow, lock acquisition during INSERT operations, lock inheritance after rollbacks, and reproducing the deadlock scenario with code examples and diagrams.
This article provides a comprehensive examination of InnoDB's physical file layout, detailing page components such as the file header, page header, record storage formats, variable‑length and NULL handling, free space, page directory, file trailer, external storage pages, and the differences among row formats and compression techniques, illustrated with concrete hexdump examples and SQL statements.
This article explores memory management concepts from both operating system and MySQL perspectives, covering virtual vs physical memory, process memory usage analysis using tools like top and /proc, and experimental verification of memory allocation behaviors.
This article examines how MySQL slaves recover their replication state in GTID‑based and binary‑log‑file‑position scenarios, identifies unsafe factors that can cause crash‑unsafe behavior, and explains how enabling relay_log_recovery=on mitigates these risks while outlining remaining limitations.
This article explains MySQL GTID (Global Transaction Identifier), covering its composition, lifecycle, how GTIDs are generated during commit, the mechanisms that maintain GTID state, and practical examples that illustrate GTID behavior in replication scenarios.
