TiDB系列(II)-一次字符串比较发现的 TiDB 排序规则“时空裂缝”

Mon, 15 Dec 2025 00:00:00 +0000

一、问题是如何被发现的

事情的起点并不复杂, 用户反馈在两个 TiDB 集群上执行了同一条 INSERT：

INSERT INTO job_status
(process_time, name, message, run_id, status, create_time)
VALUES
('2025-09-17 13:21:55', 'test', '', '2025-09-17', 'START ', '2025-09-17 13:21:55');

# 对应的表结构如下:
CREATE TABLE job_status (
  process_time datetime NOT NULL,
  name varchar(255) NOT NULL,
  message text DEFAULT NULL,
  run_id varchar(255) DEFAULT NULL,
  status enum('START','IN_PROGRESS','SUCCESS','FAILURE') NOT NULL,
  create_time datetime DEFAULT CURRENT_TIMESTAMP,
  PRIMARY KEY (process_time,name,status) /*T![clustered_index] NONCLUSTERED */
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin
PARTITION BY RANGE (MONTH(process_time))
(PARTITION JAN VALUES LESS THAN (2),
 PARTITION FEB VALUES LESS THAN (3),
 PARTITION MAR VALUES LESS THAN (4),
 PARTITION APR VALUES LESS THAN (5),
 PARTITION MAY VALUES LESS THAN (6),
 PARTITION JUN VALUES LESS THAN (7),
 PARTITION JUL VALUES LESS THAN (8),
 PARTITION AUG VALUES LESS THAN (9),
 PARTITION SEP VALUES LESS THAN (10),
 PARTITION OCT VALUES LESS THAN (11),
 PARTITION NOV VALUES LESS THAN (12),
 PARTITION DEC VALUES LESS THAN (MAXVALUE));

然而得到的结果却完全不同：

TiDB系列(I)-从一段SQL来看TiDB和MySQL的差异

Thu, 04 Sep 2025 00:00:00 +0000

从一段SQL开始

让我们以 TiDB v8.5.0和 MySQL 8.0 作为测试环境，先初始化测试数据

# 首先初始化库和表以及测试需要的数据
CREATE DATABASE test;
CREATE TABLE test.student (
    id INT AUTO_INCREMENT PRIMARY KEY,
    name VARCHAR(255) NOT NULL
);
INSERT INTO test.student (name) VALUES ('John');
INSERT INTO test.student (name) VALUES ('Alice');
INSERT INTO test.student (name) VALUES ('Bob');
INSERT INTO test.student (name) VALUES ('Emma');

然后我们分别在两个环境按照顺序执行下列SQL

TiDB on 行间

TiDB系列(II)-一次字符串比较发现的 TiDB 排序规则“时空裂缝”

一、问题是如何被发现的

TiDB系列(I)-从一段SQL来看TiDB和MySQL的差异

从一段SQL开始