如何确定 MariaDB 上元数据锁定信息行 (SELECT * FROM information_schema.metadata_lock_info) 中显示的线程 ID 的 SQL 语句是什么？
Server version: 10.0.15-MariaDB MariaDB Server
所有相关问题都从 MySQL 的角度深入探讨“等待表元数据锁”，但这对 MariaDB 没有帮助，因为它们的内省实现方式与我所知的不同。谷歌搜索并没有找到很多东西。
“显示完整进程列表”给出如下行：
| 57295 | main | localhost | joints | Execute | 50 | Waiting for table metadata lock | select ...
它确实显示了该语句，但也不显示它具有锁。因此，我打开了元数据锁定信息，如此处所述 [0]。这仅提供了锁持有者的线程ID，但不提供语句：
MariaDB [joints]> SELECT * FROM information_schema.metadata_lock_info;
+-----------+--------------------------+-----------------+----------------------+--------------+----------------+
| THREAD_ID | LOCK_MODE | LOCK_DURATION | LOCK_TYPE | TABLE_SCHEMA | TABLE_NAME |
+-----------+--------------------------+-----------------+----------------------+--------------+----------------+
| 57322 | MDL_INTENTION_EXCLUSIVE | MDL_EXPLICIT | Global read lock | | |
| 57322 | MDL_SHARED_NO_READ_WRITE | MDL_EXPLICIT | Table metadata lock | joints | 16_study |
| 57322 | MDL_INTENTION_EXCLUSIVE | MDL_EXPLICIT | Schema metadata lock | joints | |
| 57269 | MDL_SHARED_READ | MDL_TRANSACTION | Table metadata lock | joints | authentication |
| 57301 | MDL_SHARED_READ | MDL_TRANSACTION | Table metadata lock | joints | authentication |
| 57280 | MDL_SHARED_READ | MDL_TRANSACTION | Table metadata lock | joints | authentication |
| 57317 | MDL_SHARED_READ | MDL_TRANSACTION | Table metadata lock | joints | ship |
| 57271 | MDL_SHARED_READ | MDL_TRANSACTION | Table metadata lock | joints | administration |
| 57264 | MDL_SHARED_READ | MDL_TRANSACTION | Table metadata lock | joints | server |
+-----------+--------------------------+-----------------+----------------------+--------------+----------------+
我真正想要的是在锁定发生时看到这两个输出的“连接”。我没有找到连接这两个“表”中的数据的方法，因为前者似乎不是一个表。我想避免得到：
ERROR 1933 (HY000): Target is not running an EXPLAINable command
在尝试实时执行此操作时，由于线程在检查时结束。
THREAD_IDinformation_schema.PROCESSLIST.IDshow
[full] processlist;SELECT * FROM information_schema.METADATA_LOCK_INFO AS mli
JOIN information_schema.PROCESSLIST AS pl ON mli.THREAD_ID = pl.ID
我更喜欢下面这样的东西，这样可以更容易地看到正在发生的事情（尽管换行符不能很好地与 cli 配合使用）：
SELECT
mli.THREAD_ID, mli.LOCK_MODE, mli.LOCK_TYPE,
CAST(GROUP_CONCAT(DISTINCT CONCAT(mli.TABLE_SCHEMA, '.', mli.TABLE_NAME) ORDER BY mli.TABLE_SCHEMA, mli.TABLE_NAME SEPARATOR '\n') AS CHAR) AS locked_tables,
pl.USER, pl.HOST, pl.DB, pl.COMMAND, pl.TIME, pl.STATE, pl.INFO, pl.QUERY_ID, pl.TID
FROM information_schema.METADATA_LOCK_INFO AS mli
JOIN information_schema.PROCESSLIST AS pl ON mli.THREAD_ID = pl.ID
GROUP BY mli.THREAD_ID, mli.LOCK_MODE, mli.LOCK_TYPE
ORDER BY time DESC, pl.ID;
特别有趣的是当
pl.COMMAND = 'Sleep'获取导致元数据锁定的查询并不简单，因为它可能已经完成，但仍然可以将违规线程发出的查询中的信息关联起来并找出来。
首先需要启用通用查询日志（https://mariadb.com/kb/en/general-query-log/）来记录线程 ID 以及在线程上执行的查询。
此外，还需要按照https://mariadb.com/kb/en/metadata-lock-info-plugin/
中所述安装METADATA_LOCK_INFO插件然后重现锁定状态以确认锁定的线程ID，本例中
9SHOW FULL PROCESSLIST;
+--+----+----...+------+-------+----+-------------------------------+----------------------------------
|Id|User|Host...|db |Command|Time|State |Info
+--+----+----...+------+-------+----+-------------------------------+----------------------------------
|3 |root|172....|mydb01|Sleep |3457| |null
|6 |root|172....|mydb01|Sleep |2434| |null
|7 |root|172....|mydb01|Sleep |664 | |null
|8 |root|172....|mydb01|Sleep |3242| |null
|9 |root|172....|mydb01|Query |3242|Waiting for table metadata lock|DROP INDEX mytbl01_idx1 ON mytbl01
...
但是它没有告诉有问题的线程是什么，所以让我们使用以下命令来确定它：
SELECT * FROM information_schema.metadata_lock_info;
+---------+-----------------------+...--------------------+...+------------+----------
|THREAD_ID|LOCK_MODE |...LOCK_TYPE |...|TABLE_SCHEMA|TABLE_NAME
+---------+-----------------------+...--------------------+...+------------+----------
|9 |MDL_BACKUP_ALTER_COPY |...Backup lock |...| |
|9 |MDL_INTENTION_EXCLUSIVE|...Schema metadata lock|...|mydb01 |
|8 |MDL_SHARED_READ |...Table metadata lock |...|mydb01 |mytbl01
|8 |MDL_SHARED_WRITE |...Table metadata lock |...|mydb01 |mytbl01
|9 |MDL_SHARED_UPGRADABLE |...Table metadata lock |...|mydb01 |mytbl01
+---------+-----------------------+...--------------------+...+------------+----------
SELECT
CONCAT('Thread ',P.ID,' executing "',P.INFO,'" IS LOCKED BY Thread ',
M.THREAD_ID) WhoLocksWho
FROM INFORMATION_SCHEMA.PROCESSLIST P,
INFORMATION_SCHEMA.METADATA_LOCK_INFO M
WHERE LOCATE(lcase(LOCK_TYPE), lcase(STATE))>0;
+-----------------------------------------------------------------------------+
|WhoLocksWho |
+-----------------------------------------------------------------------------+
|Thread 9 executing "DROP INDEX mytbl01_idx1 ON mytbl01" IS LOCKED BY Thread 8|
...
这表明有问题的线程是
8现在，有了这些信息，让我们打开常规查询日志并查找锁定的查询 (
DROP INDEX mytbl01_idx1 ON mytbl01mytbl0188 Query select count(*) from mytbl01
...
9 Query DROP INDEX mytbl01_idx1 ON mytbl01
线程
mytbl018select count(*) from mytbl01最后，要了解更多背景知识，请学习 https://dev.mysql.com/doc/refman/8.4/en/metadata-locking.html