Home >Database >Mysql Tutorial >Summary of new features of MySQL8.0 (with code)

Summary of new features of MySQL8.0 (with code)

不言
不言forward
2019-03-19 10:25:533478browse

This article brings you a summary of the new features of MySQL8.0 (with code). It has certain reference value. Friends in need can refer to it. I hope it will be helpful to you. helped.

1. The default character set changes from latin1 to utf8mb4

Before version 8.0, the default character set was latin1, utf8 pointed to utf8mb3, and the default character set in version 8.0 It is utf8mb4, and utf8 also points to utf8mb4 by default.

(Recommended: MySQL Tutorial)

2. Replace all MyISAM system tables with InnoDB tables

Replace all system tables As a transactional innodb table, the default MySQL instance will not contain any MyISAM tables unless you manually create a MyISAM table.

# MySQL 5.7
mysql> select distinct(ENGINE) from information_schema.tables;
+--------------------+
| ENGINE             |
+--------------------+
| MEMORY             |
| InnoDB             |
| MyISAM             |
| CSV                |
| PERFORMANCE_SCHEMA |
| NULL               |
+--------------------+
6 rows in set (0.00 sec)
 
# MySQL 8.0
mysql> select distinct(ENGINE) from information_schema.tables;
+--------------------+
| ENGINE             |
+--------------------+
| NULL               |
| InnoDB             |
| CSV                |
| PERFORMANCE_SCHEMA |
+--------------------+
4 rows in set (0.00 sec)

3. Persistence of auto-increment variables

In versions before 8.0, if the value of auto-increment primary key AUTO_INCREMENT is greater than max (primary key) 1, AUTO_INCREMENT will be reset after MySQL restarts =max(primary key) 1. This phenomenon may lead to business primary key conflicts or other hard-to-find problems in some cases. The problem of auto-incremented primary key restart and reset was discovered very early (https://bugs.mysql.com/bug.ph...) and was not solved until 8.0. Version 8.0 will persist the AUTO_INCREMENT value. This value will not change after MySQL is restarted.

4. DDL atomization

The DDL of the InnoDB table supports transaction integrity, either successful or rolled back, and the DDL operation rollback log is written to the data dictionary data dictionary table mysql.innodb_ddl_log. Used for rollback operations. This table is a hidden table and cannot be seen through show tables. By setting parameters, the ddl operation log can be printed to the mysql error log.

mysql> set global log_error_verbosity=3;
mysql> set global innodb_print_ddl_logs=1;
mysql> create table t1(c int) engine=innodb;
 
# MySQL错误日志:
2018-06-26T11:25:25.817245+08:00 44 [Note] [MY-012473] [InnoDB] InnoDB: DDL log insert : [DDL record: DELETE SPACE, id=41, thread_id=44, space_id=6, old_file_path=./db/t1.ibd]
2018-06-26T11:25:25.817369+08:00 44 [Note] [MY-012478] [InnoDB] InnoDB: DDL log delete : by id 41
2018-06-26T11:25:25.819753+08:00 44 [Note] [MY-012477] [InnoDB] InnoDB: DDL log insert : [DDL record: REMOVE CACHE, id=42, thread_id=44, table_id=1063, new_file_path=db/t1]
2018-06-26T11:25:25.819796+08:00 44 [Note] [MY-012478] [InnoDB] InnoDB: DDL log delete : by id 42
2018-06-26T11:25:25.820556+08:00 44 [Note] [MY-012472] [InnoDB] InnoDB: DDL log insert : [DDL record: FREE, id=43, thread_id=44, space_id=6, index_id=140, page_no=4]
2018-06-26T11:25:25.820594+08:00 44 [Note] [MY-012478] [InnoDB] InnoDB: DDL log delete : by id 43
2018-06-26T11:25:25.825743+08:00 44 [Note] [MY-012485] [InnoDB] InnoDB: DDL log post ddl : begin for thread id : 44
2018-06-26T11:25:25.825784+08:00 44 [Note] [MY-012486] [InnoDB] InnoDB: DDL log post ddl : end for thread id : 44

Let’s look at another example. There is only one t1 table in the library, drop table t1, t2; trying to delete two tables t1 and t2, in 5.7, the execution error is reported, but the t1 table is deleted, in 8.0 An error is reported during execution, but the t1 table is not deleted, which proves the atomicity of 8.0 DDL operations. Either all operations succeed or are rolled back.

# MySQL 5.7
mysql> show tables;
+---------------+
| Tables_in_db |
+---------------+
| t1            |
+---------------+
1 row in set (0.00 sec)
mysql> drop table t1, t2;
ERROR 1051 (42S02): Unknown table 'db.t2'
mysql> show tables;
Empty set (0.00 sec)
 
# MySQL 8.0
mysql> show tables;
+---------------+
| Tables_in_db |
+---------------+
| t1            |
+---------------+
1 row in set (0.00 sec)
mysql> drop table t1, t2;
ERROR 1051 (42S02): Unknown table 'db.t2'
mysql> show tables;
+---------------+
| Tables_in_db |
+---------------+
| t1            |
+---------------+
1 row in set (0.00 sec)

5. Parameter modification persistence

MySQL 8.0 version supports online modification of global parameters and persistence. By adding the PERSIST keyword, the modified parameters can be persisted to a new configuration file. (mysqld-auto.cnf), when restarting MySQL, you can obtain the latest configuration parameters from this configuration file.
For example, execute:
set PERSIST expire_logs_days=10;
The system will generate a file containing mysqld-auto.cnf in json format in the data directory. After formatting, it will be as follows. When my.cnf and When mysqld-auto.cnf exists at the same time, the latter has higher priority.

{
    "Version": 1,
    "mysql_server": {
        "expire_logs_days": {
            "Value": "10",
            "Metadata": {
                "Timestamp": 1529657078851627,
                "User": "root",
                "Host": "localhost"
            }
        }
    }
}

6. New descending index

MySQL has supported descending indexes in syntax for a long time, but in fact, it still creates an ascending index, as shown in MySQL 5.7 below, the c2 field is in descending order , but from show create table, c2 is still in ascending order. 8.0 can see that the c2 field is in descending order.

# MySQL 5.7
mysql> create table t1(c1 int,c2 int,index idx_c1_c2(c1,c2 desc));
Query OK, 0 rows affected (0.03 sec)
mysql> show create table t1\G
*************************** 1. row ***************************
       Table: t1
Create Table: CREATE TABLE `t1` (
  `c1` int(11) DEFAULT NULL,
  `c2` int(11) DEFAULT NULL,
  KEY `idx_c1_c2` (`c1`,`c2`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4
1 row in set (0.00 sec)
 
# MySQL 8.0
mysql> create table t1(c1 int,c2 int,index idx_c1_c2(c1,c2 desc));
Query OK, 0 rows affected (0.06 sec)
mysql> show create table t1\G
*************************** 1. row ***************************
       Table: t1
Create Table: CREATE TABLE `t1` (
  `c1` int(11) DEFAULT NULL,
  `c2` int(11) DEFAULT NULL,
  KEY `idx_c1_c2` (`c1`,`c2` DESC)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_0900_ai_ci ROW_FORMAT=DYNAMIC
1 row in set (0.00 sec)

Let’s take a look at the performance of the descending index in the execution plan. Insert 100,000 pieces of random data into the t1 table and view the execution plan of select * from t1 order by c1, c2 desc;. It can be seen from the execution plan that the scan number of 100113 in 5.7 is much larger than the 5 lines in 8.0, and filesort is used.

DELIMITER ;;
CREATE PROCEDURE test_insert ()
BEGIN
DECLARE i INT DEFAULT 1;
WHILE i<100000
DO
insert into t1 select rand()*100000, rand()*100000;
SET i=i+1;
END WHILE ;
commit;
END;;
DELIMITER ;
CALL test_insert();
 
# MySQL 5.7
mysql> explain select * from t1 order by c1 , c2 desc limit 5;
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+--------+----------+-----------------------------+
| id | select_type | table | partitions | type  | possible_keys | key       | key_len | ref  | rows   | filtered | Extra                       |
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+--------+----------+-----------------------------+
|  1 | SIMPLE      | t1    | NULL       | index | NULL          | idx_c1_c2 | 10      | NULL | 100113 |   100.00 | Using index; Using filesort |
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+--------+----------+-----------------------------+
1 row in set, 1 warning (0.00 sec)
 
# MySQL 8.0
mysql> explain select * from t1 order by c1 , c2 desc limit 5;
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+------+----------+-------------+
| id | select_type | table | partitions | type  | possible_keys | key       | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | t1    | NULL       | index | NULL          | idx_c1_c2 | 10      | NULL |    5 |   100.00 | Using index |
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+------+----------+-------------+
1 row in set, 1 warning (0.00 sec)

The descending index is only effective for the specific sort order in the query. If used improperly, the query efficiency will be lower. For example, the above query sorting condition is changed to order by c1 desc, c2 desc. In this case, 5.7 The execution plan is significantly better than that of 8.0, as follows:

# MySQL 5.7
mysql> explain select * from t1  order by c1 desc , c2 desc limit 5;
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+------+----------+-------------+
| id | select_type | table | partitions | type  | possible_keys | key       | key_len | ref  | rows | filtered | Extra       |
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+------+----------+-------------+
|  1 | SIMPLE      | t1    | NULL       | index | NULL          | idx_c1_c2 | 10      | NULL |    5 |   100.00 | Using index |
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+------+----------+-------------+
1 row in set, 1 warning (0.01 sec)
 
# MySQL 8.0
mysql> explain select * from t1 order by c1 desc , c2 desc limit 5;
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+--------+----------+-----------------------------+
| id | select_type | table | partitions | type  | possible_keys | key       | key_len | ref  | rows   | filtered | Extra                       |
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+--------+----------+-----------------------------+
|  1 | SIMPLE      | t1    | NULL       | index | NULL          | idx_c1_c2 | 10      | NULL | 100429 |   100.00 | Using index; Using filesort |
+----+-------------+-------+------------+-------+---------------+-----------+---------+------+--------+----------+-----------------------------+
1 row in set, 1 warning (0.01 sec)

7. group by is no longer implicitly sorted

mysql 8.0 is no longer implicitly sorted for group by fields. If sorting is required, The order by clause must be explicitly added.

# 表结构
mysql> show create table tb1\G
*************************** 1. row ***************************
       Table: tb1
Create Table: CREATE TABLE `tb1` (
  `id` int(11) NOT NULL AUTO_INCREMENT,
  `name` varchar(50) DEFAULT NULL,
  `group_own` int(11) DEFAULT '0',
  PRIMARY KEY (`id`)
) ENGINE=InnoDB AUTO_INCREMENT=11 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_0900_ai_ci ROW_FORMAT=DYNAMIC
1 row in set (0.00 sec)
 
# 表数据
mysql> select * from tb1;
+----+------+-----------+
| id | name | group_own |
+----+------+-----------+
|  1 | 1    |         0 |
|  2 | 2    |         0 |
|  3 | 3    |         0 |
|  4 | 4    |         0 |
|  5 | 5    |         5 |
|  8 | 8    |         1 |
| 10 | 10   |         5 |
+----+------+-----------+
7 rows in set (0.00 sec)
 
# MySQL 5.7
mysql> select count(id), group_own from tb1 group by group_own;
+-----------+-----------+
| count(id) | group_own |
+-----------+-----------+
|         4 |         0 |
|         1 |         1 |
|         2 |         5 |
+-----------+-----------+
3 rows in set (0.00 sec)
 
# MySQL 8.0.11
mysql> select count(id), group_own from tb1 group by group_own;
+-----------+-----------+
| count(id) | group_own |
+-----------+-----------+
|         4 |         0 |
|         2 |         5 |
|         1 |         1 |
+-----------+-----------+
3 rows in set (0.00 sec)
 
# MySQL 8.0.11显式地加上order by进行排序
mysql> select count(id), group_own from tb1 group by group_own order by group_own;
+-----------+-----------+
| count(id) | group_own |
+-----------+-----------+
|         4 |         0 |
|         1 |         1 |
|         2 |         5 |
+-----------+-----------+
3 rows in set (0.00 sec)

8. JSON feature enhancement

MySQL 8 has greatly improved its support for JSON, adding the JSON_EXTRACT() function to extract data from JSON fields based on path query parameters, and the The data is combined into JSON arrays and objects using the JSON_ARRAYAGG() and JSON_OBJECTAGG() aggregate functions respectively.

In the master-slave replication, the new parameter binlog_row_value_options is added to control the transmission method of JSON data, allowing partial modification of the Json type. Only the modified part is recorded in the binlog, reducing the situation where there are only a few modifications to the json big data. Down, the occupation of resources.

9. redo & undo log encryption

Add the following two parameters to control the encryption of redo and undo logs.
innodb_undo_log_encrypt
innodb_undo_log_encrypt

10. innodb select for update skips lock waiting

select ... for update, select ... for share (8.0 new syntax) Add NOWAIT, SKIP LOCKED syntax, skip lock waiting, or skip locking.
In versions 5.7 and earlier, select...for update, if the lock cannot be obtained, will wait until innodb_lock_wait_timeout times out.

In version 8.0, by adding nowait and skip locked syntax, you can return immediately. If the queried row has been locked, nowait will immediately return an error, and skip locked will also return immediately, but the returned result will not include the locked row.

# session1:
mysql> begin;
mysql> select * from t1 where c1 = 2 for update;
+------+-------+
| c1   | c2    |
+------+-------+
|    2 | 60530 |
|    2 | 24678 |
+------+-------+
2 rows in set (0.00 sec)
 
# session2:
mysql> select * from t1 where c1 = 2 for update nowait;
ERROR 3572 (HY000): Statement aborted because lock(s) could not be acquired  immediately and NOWAIT is set.
mysql> select * from t1 where c1 = 2 for update skip locked;
Empty set (0.00 sec)

11. Add SET_VAR syntax

Add SET_VAR syntax in sql syntax to dynamically adjust some parameters, which will help improve statement performance.

  • select /+ SET_VAR(sort_buffer_size = 16M) / id  from test order id ;
  • insert  /+ SET_VAR(foreign_key_checks=OFF) / into test(name) values(1);

12. 支持不可见索引

使用INVISIBLE关键字在创建表或者进行表变更中设置索引是否可见。索引不可见只是在查询时优化器不使用该索引,即使使用force index,优化器也不会使用该索引,同时优化器也不会报索引不存在的错误,因为索引仍然真实存在,在必要时,也可以快速的恢复成可见。

# 创建不可见索引
create table t2(c1 int,c2 int,index idx_c1_c2(c1,c2 desc) invisible );
# 索引可见
alter table t2 alter index idx_c1_c2 visible;
# 索引不可见
alter table t2 alter index idx_c1_c2 invisible;

13. 支持直方图

优化器会利用column_statistics的数据,判断字段的值的分布,得到更准确的执行计划。

可以使用 ANALYZE TABLE table_name [UPDATE HISTOGRAM on col_name with N BUCKETS |DROP HISTOGRAM ON  clo_name] 来收集或者删除直方图信息。

直方图统计了表中某些字段的数据分布情况,为优化选择高效的执行计划提供参考,直方图与索引有着本质的区别,维护一个索引有代价。每一次的insert、update、delete都会需要更新索引,会对性能有一定的影响。而直方图一次创建永不更新,除非明确去更新它。所以不会影响insert、update、delete的性能。

# 添加/更新直方图
mysql> analyze table t1 update histogram on c1, c2 with 32 buckets;
+--------+-----------+----------+-----------------------------------------------+
| Table  | Op        | Msg_type | Msg_text                                      |
+--------+-----------+----------+-----------------------------------------------+
| db.t1 | histogram | status   | Histogram statistics created for column 'c1'. |
| db.t1 | histogram | status   | Histogram statistics created for column 'c2'. |
+--------+-----------+----------+-----------------------------------------------+
2 rows in set (2.57 sec)
 
# 删除直方图
mysql> analyze table t1 drop histogram on c1, c2;
+--------+-----------+----------+-----------------------------------------------+
| Table  | Op        | Msg_type | Msg_text                                      |
+--------+-----------+----------+-----------------------------------------------+
| db.t1 | histogram | status   | Histogram statistics removed for column 'c1'. |
| db.t1 | histogram | status   | Histogram statistics removed for column 'c2'. |
+--------+-----------+----------+-----------------------------------------------+
2 rows in set (0.13 sec)

14. 新增innodb_dedicated_server参数

能够让InnoDB根据服务器上检测到的内存大小自动配置innodb_buffer_pool_size,innodb_log_file_size,innodb_flush_method三个参数。

15. 日志分类更详细

在错误信息中添加了错误信息编号[MY-010311]和错误所属子系统[Server]

# MySQL 5.7
2018-06-08T09:07:20.114585+08:00 0 [Warning] 'proxies_priv' entry '@ root@localhost' ignored in --skip-name-resolve mode.
2018-06-08T09:07:20.117848+08:00 0 [Warning] 'tables_priv' entry 'user mysql.session@localhost' ignored in --skip-name-resolve mode.
2018-06-08T09:07:20.117868+08:00 0 [Warning] 'tables_priv' entry 'sys_config mysql.sys@localhost' ignored in --skip-name-resolve mode.
 
 
# MySQL 8.0
2018-06-21T17:53:13.040295+08:00 28 [Warning] [MY-010311] [Server] 'proxies_priv'  entry '@ root@localhost' ignored in --skip-name-resolve mode.
2018-06-21T17:53:13.040520+08:00 28 [Warning] [MY-010330] [Server] 'tables_priv'  entry 'user mysql.session@localhost' ignored in --skip-name-resolve mode.
2018-06-21T17:53:13.040542+08:00 28 [Warning] [MY-010330] [Server] 'tables_priv'  entry 'sys_config mysql.sys@localhost' ignored in --skip-name-resolve mode.

16. undo空间自动回收

  • innodb_undo_log_truncate参数在8.0.2版本默认值由OFF变为ON,默认开启undo日志表空间自动回收。
  • innodb_undo_tablespaces参数在8.0.2版本默认为2,当一个undo表空间被回收时,还有另外一个提供正常服务。
  • innodb_max_undo_log_size参数定义了undo表空间回收的最大值,当undo表空间超过这个值,该表空间被标记为可回收。

17. 增加资源组

MySQL 8.0新增了一个资源组功能,用于调控线程优先级以及绑定CPU核。
MySQL用户需要有 RESOURCE_GROUP_ADMIN权限才能创建、修改、删除资源组。
在Linux环境下,MySQL进程需要有 CAP_SYS_NICE 权限才能使用资源组完整功能。

[root@localhost~]# sudo setcap cap_sys_nice+ep /usr/local/mysql8.0/bin/mysqld
[root@localhost~]# getcap /usr/local/mysql8.0/bin/mysqld
/usr/local/mysql8.0/bin/mysqld = cap_sys_nice+ep

默认提供两个资源组,分别是USR_default,SYS_default

创建资源组:
create resource group test_resouce_group type=USER vcpu=0,1 thread_priority=5;
将当前线程加入资源组:
SET RESOURCE GROUP test_resouce_group;
将某个线程加入资源组:
SET RESOURCE GROUP test_resouce_group FOR thread_id;
查看资源组里有哪些线程:
select * from Performance_Schema.threads where RESOURCE_GROUP='test_resouce_group';
修改资源组:
alter resource group test_resouce_group vcpu = 2,3 THREAD_PRIORITY = 8;
删除资源组 :
drop resource group test_resouce_group;

# 创建资源组
mysql>create resource group test_resouce_group type=USER vcpu=0,1 thread_priority=5;
Query OK, 0 rows affected (0.03 sec)

mysql> select * from RESOURCE_GROUPS;
+---------------------+---------------------+------------------------+----------+-----------------+
| RESOURCE_GROUP_NAME | RESOURCE_GROUP_TYPE | RESOURCE_GROUP_ENABLED | VCPU_IDS |  THREAD_PRIORITY |
+---------------------+---------------------+------------------------+----------+-----------------+
| USR_default         | USER                |                      1 | 0-3      |                0 |
| SYS_default         | SYSTEM              |                      1 | 0-3      |                0 |
| test_resouce_group  | USER                |                      1 | 0-1      |                5 |
+---------------------+---------------------+------------------------+----------+-----------------+
3 rows in set (0.00 sec)

# 把线程id为60的线程加入到资源组test_resouce_group中,线程id可通过Performance_Schema.threads获取
mysql> SET RESOURCE GROUP test_resouce_group FOR 60;
Query OK, 0 rows affected (0.00 sec)

# 资源组里有线程时,删除资源组报错
mysql> drop resource group test_resouce_group;
ERROR 3656 (HY000): Resource group test_resouce_group is busy.

# 修改资源组
mysql> alter resource group test_resouce_group vcpu = 2,3 THREAD_PRIORITY = 8;
Query OK, 0 rows affected (0.10 sec)
mysql> select * from RESOURCE_GROUPS;
+---------------------+---------------------+------------------------+----------+-----------------+
| RESOURCE_GROUP_NAME | RESOURCE_GROUP_TYPE | RESOURCE_GROUP_ENABLED | VCPU_IDS | THREAD_PRIORITY |
+---------------------+---------------------+------------------------+----------+-----------------+
| USR_default         | USER                |                      1 | 0-3      |               0 |
| SYS_default         | SYSTEM              |                      1 | 0-3      |               0 |
| test_resouce_group  | USER                |                      1 | 2-3      |               8 |
+---------------------+---------------------+------------------------+----------+-----------------+
3 rows in set (0.00 sec)

# 把资源组里的线程移出到默认资源组USR_default
mysql> SET RESOURCE GROUP USR_default FOR 60;
Query OK, 0 rows affected (0.00 sec)

# 删除资源组
mysql> drop resource group test_resouce_group;
Query OK, 0 rows affected (0.04 sec)

18. 增加角色管理

角色可以认为是一些权限的集合,为用户赋予统一的角色,权限的修改直接通过角色来进行,无需为每个用户单独授权。

# 创建角色
mysql> create role role_test;
Query OK, 0 rows affected (0.03 sec)
 
# 给角色授予权限
mysql> grant select on db.* to 'role_test';
Query OK, 0 rows affected (0.10 sec)
 
# 创建用户
mysql> create user 'read_user'@'%' identified by '123456';
Query OK, 0 rows affected (0.09 sec)
 
# 给用户赋予角色
mysql> grant 'role_test' to 'read_user'@'%';
Query OK, 0 rows affected (0.02 sec)
 
# 给角色role_test增加insert权限
mysql> grant insert on db.* to 'role_test';
Query OK, 0 rows affected (0.08 sec)
 
# 给角色role_test删除insert权限
mysql> revoke insert on db.* from 'role_test';
Query OK, 0 rows affected (0.10 sec)
 
# 查看默认角色信息
mysql> select * from mysql.default_roles;
+------+-----------+-------------------+-------------------+
| HOST | USER      | DEFAULT_ROLE_HOST | DEFAULT_ROLE_USER |
+------+-----------+-------------------+-------------------+
| %    | read_user | %                 | role_test         |
+------+-----------+-------------------+-------------------+
1 row in set (0.00 sec)
 
# 查看角色与用户关系
mysql> select * from mysql.role_edges;
+-----------+-----------+---------+-----------+-------------------+
| FROM_HOST | FROM_USER | TO_HOST | TO_USER   | WITH_ADMIN_OPTION |
+-----------+-----------+---------+-----------+-------------------+
| %         | role_test | %       | read_user | N                 |
+-----------+-----------+---------+-----------+-------------------+
1 row in set (0.00 sec)
 
# 删除角色
mysql> drop role role_test;
Query OK, 0 rows affected (0.06 sec)


The above is the detailed content of Summary of new features of MySQL8.0 (with code). For more information, please follow other related articles on the PHP Chinese website!

Statement:
This article is reproduced at:segmentfault.com. If there is any infringement, please contact admin@php.cn delete