文档目录

SHOW STATS_HEALTHY

SHOW STATS_HEALTHY 语句可以预估统计信息的准确度,也就是健康度。健康度低的表可能会生成次优查询执行计划。

可以通过执行 ANALYZE 表命令来改善表的健康度。当表的健康度下降到低于 tidb_auto_analyze_ratio 时,则会自动执行 ANALYZE 命令。

语法图

ShowStmt

ShowStmt

ShowTargetFilterable

ShowTargetFilterable

ShowLikeOrWhereOpt

ShowLikeOrWhereOpt

示例

加载示例数据并运行 ANALYZE 命令:

CREATE TABLE t1 (
 id INT NOT NULL PRIMARY KEY auto_increment,
 b INT NOT NULL,
 pad VARBINARY(255),
 INDEX(b)
);
INSERT INTO t1 SELECT NULL, FLOOR(RAND()*1000), RANDOM_BYTES(255) FROM dual;
INSERT INTO t1 SELECT NULL, FLOOR(RAND()*1000), RANDOM_BYTES(255) FROM t1 a JOIN t1 b JOIN t1 c LIMIT 100000;
INSERT INTO t1 SELECT NULL, FLOOR(RAND()*1000), RANDOM_BYTES(255) FROM t1 a JOIN t1 b JOIN t1 c LIMIT 100000;
INSERT INTO t1 SELECT NULL, FLOOR(RAND()*1000), RANDOM_BYTES(255) FROM t1 a JOIN t1 b JOIN t1 c LIMIT 100000;
INSERT INTO t1 SELECT NULL, FLOOR(RAND()*1000), RANDOM_BYTES(255) FROM t1 a JOIN t1 b JOIN t1 c LIMIT 100000;
INSERT INTO t1 SELECT NULL, FLOOR(RAND()*1000), RANDOM_BYTES(255) FROM t1 a JOIN t1 b JOIN t1 c LIMIT 100000;
INSERT INTO t1 SELECT NULL, FLOOR(RAND()*1000), RANDOM_BYTES(255) FROM t1 a JOIN t1 b JOIN t1 c LIMIT 100000;
SELECT SLEEP(1);
ANALYZE TABLE t1;
SHOW STATS_HEALTHY; # should be 100% healthy
SHOW STATS_HEALTHY;
+---------+------------+----------------+---------+
| Db_name | Table_name | Partition_name | Healthy |
+---------+------------+----------------+---------+
| test    | t1         |                |     100 |
+---------+------------+----------------+---------+
1 row in set (0.00 sec)

执行批量更新来删除大约 30% 的记录,然后检查统计信息的健康度:

DELETE FROM t1 WHERE id BETWEEN 101010 AND 201010; # delete about 30% of records
SHOW STATS_HEALTHY;
SHOW STATS_HEALTHY;
+---------+------------+----------------+---------+
| Db_name | Table_name | Partition_name | Healthy |
+---------+------------+----------------+---------+
| test    | t1         |                |      50 |
+---------+------------+----------------+---------+
1 row in set (0.00 sec)

MySQL 兼容性

SHOW STATS_HEALTHY 语句是 TiDB 对 MySQL 语法的扩展。

另请参阅