作者：杨涛涛
资深数据库专家，专研 MySQL 十余年。善于 MySQL、PostgreSQL、MongoDB 等开源数据库相干的备份复原、SQL 调优、监控运维、高可用架构设计等。目前任职于爱可生，为各大运营商及银行金融企业提供 MySQL 相干技术支持、MySQL 相干课程培训等工作。
本文起源：原创投稿
* 爱可生开源社区出品，原创内容未经受权不得随便应用，转载请分割小编并注明起源。
ProxySQL 是一个应用十分宽泛并且较稳固的中间件，有很多性能点。比方查问缓存，查问重写，读写拆散，数据分片等等。
本篇要介绍的是 ProxySQL 和 MySQL Replication 以及 MySQL MGR 的初步联合，初步读写拆散以及 failover 性能的体验。
在本机装置一个 ProxySQL 实例，六个 MySQL 实例；ProxySQL 和 MySQL 版本均是最新版。
ProxySQL: 治理端口 6032，流量端口 6033。
MySQL Replication：流量端口别离为：3340、3341、3342
MySQL MGR：流量端口别离为：3343、3344、3345
第一，ProxySQL 以及六个 MySQL 部署。
ProxySQL 装置比较简单，官网 apt/yum，或者本人下载安装。装好六个 MySQL 实例，并且配置好 MySQL 主从以及组复制环境。
第二，ProxySQL 记录 MySQL 实例相干信息。
进入 ProxySQL 治理端，把以上六个 MySQL 实例信息顺次插入到表 mysql_servers：主从实例的 hostgroup_id 对立设置为 1，为了不毁坏后续 failover 相干 hostgroup_id 的连续性，组复制实例的 hostgroup_id 对立设置为 3。
Admin> select hostgroup_id, hostname,port,status from mysql_servers;
+--------------+-----------+------+--------+
| hostgroup_id | hostname | port | status |
+--------------+-----------+------+--------+
| 1 | 127.0.0.1 | 3340 | ONLINE |
| 1 | 127.0.0.1 | 3341 | ONLINE |
| 1 | 127.0.0.1 | 3342 | ONLINE |
| 3 | 127.0.0.1 | 3343 | ONLINE |
| 3 | 127.0.0.1 | 3344 | ONLINE |
| 3 | 127.0.0.1 | 3345 | ONLINE |
+--------------+-----------+------+--------+
6 rows in set (0.00 sec)第三，MySQL 端创立 ProxySQL 所需的两类用户
ProxySQL 对 MySQL 来讲，有须要两类用户，这两类用户须要同时在主从和组复制环境创立。
1，监控用户：为了免去前面反复设置监控用户变量的工作，两种架构用户名和明码保持一致，用户名和明码都是 proxysql_monitor。proxysql_monitor 须要 以下权限：
client,session_variables_admin,system_variables_admin,select
在 MySQL 主从以及组复制环境里别离执行上面 SQL：
MySQL localhost:3343 ssl SQL > create user proxysql_monitor@'127.0.0.1' identified by 'proxysql_monitor';
Query OK, 0 rows affected (0.0596 sec)
MySQL localhost:3343 ssl SQL > grant replication client,session_variables_admin,system_variables_admin,select on *.* to proxysql_monitor@'127.0.0.1';
Query OK, 0 rows affected (0.0103 sec)进入 ProxySQL 治理端，设置监控用户：
Admin> set mysql-monitor_username='proxysql_monitor';
Query OK, 1 row affected (0.00 sec)
Admin> set mysql-monitor_password='proxysql_monitor';
Query OK, 1 row affected (0.00 sec)
2，开发用户：对于主从和组复制环境，别离创立此类用户。
主从环境用户创立：用户名 dev_user
MySQL localhost:3343 ssl SQL > create user dev_user@'127.0.0.1' identified by 'dev_user';
Query OK, 0 rows affected (0.1221 sec)
MySQL localhost:3343 ssl SQL > grant insert,delete,update,select,create on ytt.* to dev_user@'127.0.0.1';
Query OK, 0 rows affected (0.0359 sec)组复制环境用户创立：用户名 dev_user_mgr
MySQL localhost:3343 ssl SQL > create user dev_user_mgr@'127.0.0.1' identified by 'dev_user_mgr';
Query OK, 0 rows affected (0.1221 sec)
MySQL localhost:3343 ssl SQL > grant insert,delete,update,select,create on ytt.* to dev_user_mgr@'127.0.0.1';
Query OK, 0 rows affected (0.0359 sec)进入 ProxySQL 治理端，别离插入主从以及组复制对应的开发用户到表 mysql_users。字段 transaction_persistent 为 1 代表事务不拆分，对立去主库检索。
Admin> insert into mysql_users(username,password,active,default_hostgroup,transaction_persistent)
values
('dev_user','dev_user',1,1,1),
('dev_user_mgr','dev_user_mgr',1,3,1);
Query OK, 1 row affected (0.00 sec)
Admin> select username,active,default_hostgroup from mysql_users;
+--------------+--------+-------------------+
| username | active | default_hostgroup |
+--------------+--------+-------------------+
| dev_user | 1 | 1 |
| dev_user_mgr | 1 | 3 |
+--------------+--------+-------------------+
2 rows in set (0.00 sec)第四，配置读写拆散
进入 ProxySQL 治理端，插入之前创立的两个开发用户到表 mysql_query_rules，定义最根本的读写拆散策略，只有是 select 结尾的语句都分流到从库。
Admin> INSERT INTO mysql_query_rules(username,schemaname,active,match_pattern,destination_hostgroup,apply) VALUES
('dev_user','ytt',1,'^select',2,1),
('dev_user_mgr','ytt',1,'^select',4,1);
Query OK, 2 rows affected (0.00 sec)
Admin> select username,schemaname,active,match_pattern,destination_hostgroup,apply from mysql_query_rules;
+--------------+------------+--------+---------------+-----------------------+-------+
| username | schemaname | active | match_pattern | destination_hostgroup | apply |
+--------------+------------+--------+---------------+-----------------------+-------+
| dev_user | ytt | 1 | ^select | 2 | 1 |
| dev_user_mgr | ytt | 1 | ^select | 4 | 1 |
+--------------+------------+--------+---------------+-----------------------+-------+
2 rows in set (0.00 sec)
设置好相干信息后把以上所有更改加载到内存，并且长久化到磁盘。
Admin> load mysql servers to runtime;
Query OK, 0 rows affected (0.01 sec)
Admin> load mysql users to runtime;
Query OK, 0 rows affected (0.00 sec)
Admin> load mysql variables to runtime;
Query OK, 0 rows affected (0.00 sec)
Admin> load mysql query rules to runtime;
Query OK, 0 rows affected (0.00 sec)
Admin> save mysql servers to disk;
Query OK, 0 rows affected (0.18 sec)
Admin> save mysql users to disk;
Query OK, 0 rows affected (0.13 sec)
Admin> save mysql variables to disk;
Query OK, 140 rows affected (0.17 sec)
Admin> save mysql query rules to disk;
Query OK, 0 rows affected (0.11 sec)别离测试下主从和组复制两种架构的读写拆散成果：开发用户 dev_user/dev_user_mgr 连贯端口 6033，创立一张表 t1，插入一条记录，并且简略查问一次。
-- 主从环境：root@ytt-ubuntu:/home/ytt/scripts# mysql -u dev_user -p -h 127.0.0.1 -P 6033 -e " \
> use ytt;
> create table t1 (id int primary key,str1 varchar(100));
> insert t1 values (1,'replication');
> select * from t1;
> ";
Enter password:
-- 组复制环境把用户 dev_user 和明码替换为 dev_user_mgr 和对应明码反复执行一次。进入 ProxySQL 治理端，检索审计表 stats_mysql_query_digest：写入申请和读取申请依据不同的用户被胜利散发到 mysql_query_rules 表里对应的 hostgroup 上。
Admin> select hostgroup,username,digest_text,count_star from stats_mysql_query_digest where schemaname = 'ytt';
+-----------+--------------+------------------------------------------------------+------------+
| hostgroup | username | digest_text | count_star |
+-----------+--------------+------------------------------------------------------+------------+
| 4 | dev_user_mgr | select * from t1 | 1 |
| 3 | dev_user_mgr | insert t1 values (?,?) | 1 |
| 3 | dev_user_mgr | create table t1 (id int primary key,str1 varchar(?)) | 1 |
| 2 | dev_user | select * from t1 | 1 |
| 1 | dev_user | insert t1 values (?,?) | 1 |
| 1 | dev_user | create table t1 (id int primary key,str1 varchar(?)) | 1 |
+-----------+--------------+------------------------------------------------------+------------+
6 rows in set (0.00 sec)第五，配置主从主动 failover 性能：
进入 RroxySQL 治理端，把主从相干实例信息插入到表 mysql_replication_hostgroups 即可。
ProxySQL 通过实时监控 MySQL 零碎变量（’read_only’,’innodb_read_only’,’super_read_only’）开关与否来探测对应的 MySQL 实例是主库还是从库，完了自动更新 mysql_server 表主库对应的 IP 和端口来达到 failover 的目标。
Admin> insert into mysql_replication_hostgroups (writer_hostgroup,reader_hostgroup,check_type,comment)values(1,2,'super_read_only','MySQL Replication fa
ailover test');
Query OK, 1 row affected (0.00 sec)
Admin> select * from mysql_replication_hostgroups;
+------------------+------------------+-----------------+---------------------------------+
| writer_hostgroup | reader_hostgroup | check_type | comment |
+------------------+------------------+-----------------+---------------------------------+
| 1 | 2 | super_read_only | MySQL Replication failover test |
+------------------+------------------+-----------------+---------------------------------+
1 row in set (0.01 sec)
Admin> load mysql servers to runtime;
Query OK, 0 rows affected (0.01 sec)
Admin> save mysql servers to disk;
Query OK, 0 rows affected (0.18 sec)
我用 MySQL Shell 操作正本集来手动进行主从切换，设置主库为端口 3342。
MySQL localhost:3340 ssl Py > rs = dba.get_replica_set()
You are connected to a member of replicaset 'rs1'.
MySQL localhost:3340 ssl Py > rs.set_primary_instance('root@localhost:3342')
127.0.0.1:3342 will be promoted to PRIMARY of 'rs1'.
The current PRIMARY is 127.0.0.1:3340.
...
127.0.0.1:3342 was promoted to PRIMARY.查看 ProxySQL 日志，曾经感知到主从切换，新的主库主动变为 127.0.0.1:3342
2021-12-15 16:02:08 [INFO] Regenerating read_only_set1 with 1 servers
2021-12-15 16:02:08 [INFO] read_only_action() detected RO=0 on server 127.0.0.1:3342 for the first time after commit(), but no need to reconfigure也能够进入 ProxySQL 治理端来查问表 mysql_servers 的字段 hostgroup_id=1 的匹配记录是否曾经变更为新主库：
Admin> select hostname,port from mysql_servers where hostgroup_id = 1;
+-----------+------+
| hostname | port |
+-----------+------+
| 127.0.0.1 | 3342 |
+-----------+------+
1 row in set (0.00 sec)第六，配置组复制主动 failover 性能：
和主从配置相似，把组复制实例相干信息插入到表 mysql_replication_hostgroups 即可。
这里和主从有点不一样的中央：writer_hostgroup,backup_writer_hostgroup, reader_hostgroup, offline_hostgroup 这四个字段代表不同职责的 Hostgroup，最好是设置不一样。
Admin> insert into mysql_group_replication_hostgroups (writer_hostgroup,backup_writer_hostgroup, reader_hostgroup, offline_hostgroup,active,max_writers,writer_is_also_reader,max_transactions_behind) values (3,4,5,6,1,1,1,1000);
Query OK, 1 row affected (0.00 sec)
Admin> load mysql servers to runtime;
Query OK, 0 rows affected (0.01 sec)
Admin> save mysql servers to disk;
Query OK, 0 rows affected (0.19 sec)进入 ProxySQL 治理端，查问组复制日志表 mysql_server_group_replication_log，能够看到以后的组复制对应的实例数据，其中主库为：127.0.0.1:3343。
Admin> select hostname, port,viable_candidate,read_only, transactions_behind, error from mysql_server_group_replication_log where port in (3343,3344,3345) order by time_start_us desc limit 3;
+-----------+------+------------------+-----------+---------------------+-------+
| hostname | port | viable_candidate | read_only | transactions_behind | error |
+-----------+------+------------------+-----------+---------------------+-------+
| 127.0.0.1 | 3345 | YES | YES | 0 | NULL |
| 127.0.0.1 | 3344 | YES | YES | 0 | NULL |
| 127.0.0.1 | 3343 | YES | NO | 0 | NULL |
+-----------+------+------------------+-----------+---------------------+-------+
3 rows in set (0.00 sec)同样用 MySQL Shell 来验证下组复制的主备角色切换后 ProxySQL 是否会主动感知：把实例 127.0.0.1：3344 晋升为主库。
MySQL localhost:3343 ssl sys Py > rc = dba.get_cluster();
MySQL localhost:3343 ssl sys Py > rc.set_primary_instance('root@localhost:3344');
Setting instance 'localhost:3344' as the primary instance of cluster 'ytt_mgr'...
Instance '127.0.0.1:3343' was switched from PRIMARY to SECONDARY.
Instance '127.0.0.1:3344' was switched from SECONDARY to PRIMARY.
Instance '127.0.0.1:3345' remains SECONDARY.
WARNING: The cluster internal session is not the primary member anymore. For cluster management operations please obtain a fresh cluster handle using dba.get_cluster().
The instance 'localhost:3344' was successfully elected as primary.
MySQL localhost:3343 ssl sys Py > 进入 ProxySQL 治理端，查问组复制日志：127.0.0.1：3344 主动变更为主库。
Admin> select hostname, port,viable_candidate,read_only, transactions_behind, error from mysql_server_group_replication_log where port in (3343,3344,3345) order by time_start_us desc limit 3;
+-----------+------+------------------+-----------+---------------------+-------+
| hostname | port | viable_candidate | read_only | transactions_behind | error |
+-----------+------+------------------+-----------+---------------------+-------+
| 127.0.0.1 | 3345 | YES | YES | 0 | NULL |
| 127.0.0.1 | 3344 | YES | NO | 0 | NULL |
| 127.0.0.1 | 3343 | YES | YES | 0 | NULL |
+-----------+------+------------------+-----------+---------------------+-------+
3 rows in set (0.00 sec)或者查问 mysql_server 表，查找 hostgroup_id 为 3 的记录是否变更为新主库。
Admin> select port from mysql_servers where hostgroup_id = 3;
+------+
| port |
+------+
| 3344 |
+------+
1 row in set (0.00 sec)总结
本篇简略介绍 ProxySQL 配置 MySQL HA 的相干配置与验证测试，更具体的配置与验证策略请关注后续。