目的
在上一篇文章MySQL ProxySQL读写分离使用初探里初步介绍了ProxySQL的使用，本文继续介绍它的一些特点和DBProxy的性能差异。深入一些去了解ProxySQL，通过例子来说明ProxySQL的一些特性和用sysbench对其进行测试来对比分析说明。
环境：
Distributor ID： Ubuntu Description ： Ubuntu 14.04.5 LTS Release ： 14.04 Codename ： trusty
MySQL Master ：192.168.200.202 MySQL Slave ：192.168.200.132 APP IP ：192.168.200.25/64
测试
本文测试环境是在上一篇文章的基础上进行的，所以已经做了一主一从的读写分离。在此基础上，若从库挂了，会怎么样？这里先把从库（132）shutdown掉，看看读去了哪里。
tips：如何修改管理接口的用户名密码？除了初始化时候修改配置文件，还有一个方法是在管理端口设置：
admin@127.0.0.1 : (none) 12:52:53>set admin-admin_credentials='zjy:zjy'; Query OK, 1 row affected (0.00 sec)
admin模块修改（select * from global_variables where variable_name like 'admin%';）需要用admin加载：
admin@127.0.0.1 : (none) 12:53:02>load admin variables to runtime; Query OK, 0 rows affected (0.00 sec) admin@127.0.0.1 : (none) 12:53:33>save admin variables to disk; Query OK, 9 rows affected (0.00 sec)
1，从库不可用
① 关闭从库
当前ProxySQL下后端MySQL的运行情况：
admin@127.0.0.1 : (none) 11:25:26>select hostgroup_id,hostname,port,status from runtime_mysql_servers; +--------------+-----------------+------+--------+ | hostgroup_id | hostname | port | status | +--------------+-----------------+------+--------+ | 100 | 192.168.200.202 | 3306 | ONLINE | | 1000 | 192.168.200.132 | 3306 | ONLINE | +--------------+-----------------+------+--------+
shutdown从库（132）后，后端MySQL的运行情况：
admin@127.0.0.1 : (none) 11:33:24>select hostgroup_id,hostname,port,status from runtime_mysql_servers; +--------------+-----------------+------+---------+ | hostgroup_id | hostname | port | status | +--------------+-----------------+------+---------+ | 100 | 192.168.200.202 | 3306 | ONLINE | | 1000 | 192.168.200.132 | 3306 | SHUNNED | +--------------+-----------------+------+---------+
此时读的操作会报超时：
sbuser@192.168.200.24 : sbtest 11:30:40>select * from x; ERROR 9001 (HY000): Max connect timeout reached while reaching hostgroup 1000 after 10000ms
原因是proxysql的核心都在规则，shutdown从之后，proxysql还是想路由到 hostgroup=1000，它不会自动选择默认的100(mysql_users里配置的default_hostgroup) 。
这里解决的办法是：在mysql_servers的hostgroup 1000 里面要插一条主库的记录，然后把weight设小，当读不到从库，回去主库查询。
admin@127.0.0.1 : (none) 11:50:13>insert into mysql_servers(hostgroup_id,hostname,port,weight,max_connections,max_replication_lag,comment) values(1000,'192.168.200.202',3306,1,1000,10,'test proxysql'); Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 11:50:42>update mysql_servers set weight=9 where hostgroup_id=1000 and hostname='192.168.200.132'; Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 11:53:41>select hostgroup_id,hostname,port,weight from mysql_servers; +--------------+-----------------+------+--------+ | hostgroup_id | hostname | port | weight | +--------------+-----------------+------+--------+ | 100 | 192.168.200.202 | 3306 | 1 | | 1000 | 192.168.200.132 | 3306 | 9 | | 1000 | 192.168.200.202 | 3306 | 1 | +--------------+-----------------+------+--------+ admin@127.0.0.1 : (none) 11:54:03>load mysql servers to runtime; Query OK, 0 rows affected (0.01 sec) admin@127.0.0.1 : (none) 11:54:28>save mysql servers to disk; Query OK, 0 rows affected (0.00 sec) admin@127.0.0.1 : (none) 11:54:38>select hostgroup_id,hostname,port,status from runtime_mysql_servers; +--------------+-----------------+------+---------+ | hostgroup_id | hostname | port | status | +--------------+-----------------+------+---------+ | 100 | 192.168.200.202 | 3306 | ONLINE | | 1000 | 192.168.200.132 | 3306 | SHUNNED | | 1000 | 192.168.200.202 | 3306 | ONLINE | +--------------+-----------------+------+---------+
此时读的操作正常：
sbuser@192.168.200.24 : sbtest 11:52:37>select * from x; +------+ | id | +------+ | 123 | | 123 | | 123 | +------+ 3 rows in set (0.01 sec)
说明从关闭了之后读操作确实去主上执行了。当从库恢复之后，以后的读操作主库也可以处理 1/10 的读请求。
② 从库延迟/从库停止复制
在上一篇文章中已经建立了监控账号：proxysql，由于需要执行show slave status的命令来获得延迟时间，所以需要权限SUPER 和 REPLICATION CLIENT。并且需要设置mysql_servers.max_replication_lag的值，由于mysql_servers.max_replication_lag仅适用于从，但也可以将其配置为所有主机，无论是从还是主（不会有任何影响）。
-- 设置监控账号权限 dba@192.168.200.202 : sbtest 10:44:38>GRANT SUPER, REPLICATION CLIENT ON *.* TO 'proxysql'@'192.168.200.24' IDENTIFIED BY PASSWORD '*BF27B4C7AAD278126E228AA8427806E870F64F39'; Query OK, 0 rows affected (0.00 sec) -- 设置延迟的阈值 admin@127.0.0.1 : (none) 11:04:50>UPDATE mysql_servers SET max_replication_lag=5; Query OK, 3 rows affected (0.00 sec) -- 应用配置 admin@127.0.0.1 : (none) 11:04:54>load mysql servers to runtime; Query OK, 0 rows affected (0.01 sec) admin@127.0.0.1 : (none) 11:05:04>save mysql servers to disk; Query OK, 0 rows affected (0.01 sec)
主从复制正常的情况下，后端MySQL的情况：
admin@127.0.0.1 : (none) 11:05:13>select hostgroup_id,hostname,port,status,max_replication_lag from runtime_mysql_servers; +--------------+-----------------+------+--------+---------------------+ | hostgroup_id | hostname | port | status | max_replication_lag | +--------------+-----------------+------+--------+---------------------+ | 1000 | 192.168.200.132 | 3306 | ONLINE | 5 | | 1000 | 192.168.200.202 | 3306 | ONLINE | 5 | | 100 | 192.168.200.202 | 3306 | ONLINE | 5 | +--------------+-----------------+------+--------+---------------------+ 3 rows in set (0.00 sec)
从库执行stop slave之后，后端MySQL的情况：
admin@127.0.0.1 : (none) 11:06:52>select hostgroup_id,hostname,port,status,max_replication_lag from runtime_mysql_servers; +--------------+-----------------+------+---------+---------------------+ | hostgroup_id | hostname | port | status | max_replication_lag | +--------------+-----------------+------+---------+---------------------+ | 1000 | 192.168.200.132 | 3306 | SHUNNED | 5 | | 1000 | 192.168.200.202 | 3306 | ONLINE | 5 | | 100 | 192.168.200.202 | 3306 | ONLINE | 5 | +--------------+-----------------+------+---------+---------------------+ 3 rows in set (0.00 sec)
此时，132从库不可用，读都到了HG 1000的202上去了，可以自行测试。 也可以在日志里看到：
2017-05-11 11:06:43 MySQL_HostGroups_Manager.cpp:934:replication_lag_action(): [WARNING] Shunning server 192.168.200.132:3306 with replication lag of 60 second
日志显示延迟60s，这个是怎么回事？这里需要说明下几个变量：
mysql-monitor_replication_lag_interval：主从延迟检测时间，默认10秒。
mysql-monitor_slave_lag_when_null：当为null时，设置的延迟值，默认为60。
admin@127.0.0.1 : (none) 11:08:35>select * from global_variables where variable_name like 'mysql-monitor%lag%'; +----------------------------------------+----------------+ | variable_name | variable_value | +----------------------------------------+----------------+ | mysql-monitor_replication_lag_interval | 10000 | | mysql-monitor_replication_lag_timeout | 1000 | | mysql-monitor_slave_lag_when_null | 60 | +----------------------------------------+----------------+ 3 rows in set (0.00 sec)
根据mysql_servers.max_replication_lag设置的阈值，这2个参数可以根据自己的情况来设置，比如设置检测时间为1500。延迟的记录也可以通过表来查看：
admin@127.0.0.1 : (none) 11:19:47>select * from mysql_server_replication_lag_log limit 3; +-----------------+------+------------------+-----------------+----------+-------+ | hostname | port | time_start_us | success_time_us | repl_lag | error | +-----------------+------+------------------+-----------------+----------+-------+ | 192.168.200.132 | 3306 | 1494472189886932 | 411 | 0 | NULL | | 192.168.200.202 | 3306 | 1494472189887224 | 372 | NULL | NULL | | 192.168.200.202 | 3306 | 1494472189887640 | 325 | NULL | NULL | +-----------------+------+------------------+-----------------+----------+-------+ 3 rows in set (0.00 sec)
主从延迟的情况和stop slave的情况一样，只是stop slave是把延迟设置成了60s。
小结：通过上面的测试说明ProxySQL可以在从库不可用时进行下线，不需要人为再进行干预，等到恢复正常之后自动上线提供服务。
2，多路由规则
① 根据库路由
在现有基础上再增加一个主从：
M：192.168.200.97 S：192.168.200.245
授权账号：程序和监控账号
dba@192.168.200.97 : proxysql 12:39:39>GRANT SUPER, REPLICATION CLIENT ON *.* TO 'proxysql'@'192.168.200.24' IDENTIFIED BY PASSWORD '*BF27B4C7AAD278126E228AA8427806E870F64F39'; Query OK, 0 rows affected (0.01 sec) dba@192.168.200.97 : proxysql 12:42:50>grant select,insert,update,delete on proxysql.* to proxysql@192.168.200.24 identified by 'proxysql'; Query OK, 0 rows affected (0.00 sec)
配置ProxySQL：
admin@127.0.0.1 : (none) 12:43:35>insert into mysql_servers(hostgroup_id,hostname,port,weight,max_connections,max_replication_lag,comment) values(101,'192.168.200.97',3306,1,1000,10,'test proxysql'); Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 12:45:15>insert into mysql_servers(hostgroup_id,hostname,port,weight,max_connections,max_replication_lag,comment) values(1001,'192.168.200.245',3306,9,1000,10,'test proxysql'); Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 12:45:24>insert into mysql_servers(hostgroup_id,hostname,port,weight,max_connections,max_replication_lag,comment) values(1001,'192.168.200.97',3306,1,1000,10,'test proxysql'); Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 12:45:36>insert into mysql_users(username,password,active,default_hostgroup,transaction_persistent) values('proxysql','proxysql',1,101,1); Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 12:46:55>INSERT INTO mysql_query_rules(active,schemaname,match_pattern,destination_hostgroup,apply) VALUES(1,'proxysql','^SELECT.*FOR UPDATE$',101,1); Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 12:56:47> admin@127.0.0.1 : (none) 12:56:47>INSERT INTO mysql_query_rules(active,schemaname,match_pattern,destination_hostgroup,apply) VALUES(1,'proxysql','^SELECT',1001,1); Query OK, 1 row affected (0.00 sec) -- 应用保存配置 admin@127.0.0.1 : (none) 12:56:55>load mysql servers to runtime; admin@127.0.0.1 : (none) 12:57:00>load mysql users to runtime;
admin@127.0.0.1 : (none) 12:57:04>load mysql query rules to runtime; admin@127.0.0.1 : (none) 12:57:11>save mysql servers to disk; admin@127.0.0.1 : (none) 12:57:17>save mysql users to disk; admin@127.0.0.1 : (none) 12:57:21>save mysql query rules to disk;
rules、servers、users信息：
admin@127.0.0.1 : (none) 03:28:11>select rule_id,active,username,schemaname,client_addr,destination_hostgroup,match_pattern,flagIN,flagOUT,apply from mysql_query_rules; +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ | rule_id | active | username | schemaname | client_addr | destination_hostgroup | match_pattern | flagIN | flagOUT | apply | +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ | 3 | 1 | NULL | NULL | NULL | 100 | ^SELECT.*FOR UPDATE$ | 0 | NULL | 1 | | 4 | 1 | NULL | NULL | NULL | 1000 | ^SELECT | 0 | NULL | 1 | | 5 | 1 | NULL | proxysql | NULL | 101 | ^SELECT.*FOR UPDATE$ | 0 | NULL | 1 | | 6 | 1 | NULL | proxysql | NULL | 1001 | ^SELECT | 0 | NULL | 1 | +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ 4 rows in set (0.00 sec) admin@127.0.0.1 : (none) 03:29:10>select username,default_hostgroup from mysql_users; +----------+-------------------+ | username | default_hostgroup | +----------+-------------------+ | sbuser | 100 | | proxysql | 101 | +----------+-------------------+ 2 rows in set (0.00 sec) admin@127.0.0.1 : (none) 03:29:28>select hostgroup_id,hostname,port,status from mysql_servers; +--------------+-----------------+------+--------+ | hostgroup_id | hostname | port | status | +--------------+-----------------+------+--------+ | 1000 | 192.168.200.132 | 3306 | ONLINE | | 100 | 192.168.200.202 | 3306 | ONLINE | | 1000 | 192.168.200.202 | 3306 | ONLINE | | 101 | 192.168.200.97 | 3306 | ONLINE | | 1001 | 192.168.200.245 | 3306 | ONLINE | | 1001 | 192.168.200.97 | 3306 | ONLINE | +--------------+-----------------+------+--------+ 6 rows in set (0.00 sec)
模拟app连接：
/Users/jinyizhou [15:32:09] ~$ mysql -uproxysql -pproxysql -h192.168.200.24 -P6033 -A ... proxysql@192.168.200.24 : (none) 03:32:11>show databases; +--------------------+ | Database | +--------------------+ | information_schema | | proxysql | +--------------------+ 2 rows in set (0.00 sec) proxysql@192.168.200.24 : (none) 03:32:13>use proxysql Database changed proxysql@192.168.200.24 : proxysql 03:32:17>show tables; +--------------------+ | Tables_in_proxysql | +--------------------+ | xx | +--------------------+ 1 row in set (0.00 sec) proxysql@192.168.200.24 : proxysql 03:32:24>insert into xx values(999); Query OK, 1 row affected (0.00 sec) proxysql@192.168.200.24 : proxysql 03:35:49>select * from xx; ERROR 1044 (#4200): Access denied for user 'proxysql'@'192.168.200.24' to database 'proxysql'
只有select的时候没有权限，其他insert，update等都是有权限的，为啥呢？原因是这里的路由关系，ProxySQL的读写分离核心就是路由，这里因为select的路由错了，到了HG为1000的主从上了：
admin@127.0.0.1 : (none) 03:32:28>select hostgroup,schemaname,username,digest_text from stats_mysql_query_digest; +-----------+--------------------+----------+----------------------------------+ | hostgroup | schemaname | username | digest_text | +-----------+--------------------+----------+----------------------------------+ | 1000 | proxysql | proxysql | select * from xx | | 101 | proxysql | proxysql | show tables | | 101 | information_schema | proxysql | show databases | | 1000 | information_schema | proxysql | SELECT DATABASE() | | 101 | information_schema | proxysql | select USER() | | 101 | information_schema | proxysql | select @@version_comment limit ? | +-----------+--------------------+----------+----------------------------------+
mysql_query_rules是整个ProxySQL的核心，上篇文章已经对该表进行了说明，在这里对这例子再次讲解下：
rule_id是表的自增主键，路由规则处理是以 rule_id 的顺序进行匹配，若没有找到规则就直接去mysql_users.default_hostgroup字段里找。上面信息中除了select之外的其他操作都找不到规则就直接去users表里取，所以这些操作不会报错。而我们执行的select被rule_id为4的规则匹配上，因为rule_id=4的是匹配所有库并且apply=1表示该正则匹配后，将不再接受其他匹配，直接转发。这样就转发到了HG为1000上面的主机上了，就报没有权限的错误。若apply=0则继续匹配下面，若没有找到路由规则，则返回再看flagOUT是否为NULL，是NULL则直接匹配，否则报错。大致的流程如下：
flagIN, flagOUT, apply: 用来定义路由链 chains of rules 首先会检查 flagIN=0 的规则，以rule_id的顺序；如果都没匹配上，则走这个用户的default_hostgroup 当匹配一条规则后，会检查 apply是否为1，是1则直接转发，不是1则继续匹配，匹配到就转发，否则看flagOUT, 如果不为NULL，并且 flagIN != flagOUT ，则进入以flagIN为上一个flagOUT值的新规则链 如果不为NULL，并且 flagIN = flagOUT，则应用这条规则 如果为NULL，则结束，应用这条规则
通过上面的说明，如何读取到正确的HG呢？这里可以设置apply=0
admin@127.0.0.1 : (none) 04:18:45>update mysql_query_rules set apply=0 where rule_id in (3,4); Query OK, 2 rows affected (0.00 sec) admin@127.0.0.1 : (none) 04:18:56>load mysql query rules to runtime; Query OK, 0 rows affected (0.00 sec) admin@127.0.0.1 : (none) 04:18:59>save mysql query rules to disk; Query OK, 0 rows affected (0.00 sec) admin@127.0.0.1 : (none) 04:19:01>select rule_id,active,username,schemaname,client_addr,destination_hostgroup,match_pattern,flagIN,flagOUT,apply from mysql_query_rules; +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ | rule_id | active | username | schemaname | client_addr | destination_hostgroup | match_pattern | flagIN | flagOUT | apply | +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ | 3 | 1 | NULL | NULL | NULL | 100 | ^SELECT.*FOR UPDATE$ | 0 | NULL | 0 | | 4 | 1 | NULL | NULL | NULL | 1000 | ^SELECT | 0 | NULL | 0 | | 5 | 1 | NULL | proxysql | NULL | 101 | ^SELECT.*FOR UPDATE$ | 0 | NULL | 1 | | 6 | 1 | NULL | proxysql | NULL | 1001 | ^SELECT | 0 | NULL | 1 | +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ 4 rows in set (0.00 sec)
和上面一样模拟app连接，得到的信息：发现全部走了正确的路由。
admin@127.0.0.1 : (none) 05:58:55>select hostgroup,schemaname,username,digest_text from stats_mysql_query_digest; +-----------+------------+----------+----------------------------------+ | hostgroup | schemaname | username | digest_text | +-----------+------------+----------+----------------------------------+ | 101 | proxysql | proxysql | insert into xx values(?) | | 1001 | proxysql | proxysql | select * from xx | | 1001 | proxysql | proxysql | SELECT DATABASE() | | 101 | proxysql | proxysql | select USER() | | 101 | proxysql | proxysql | select @@version_comment limit ? | +-----------+------------+----------+----------------------------------+ 5 rows in set (0.00 sec)
查看路由规则的命中情况：
admin@127.0.0.1 : (none) 05:59:19>select * from stats_mysql_query_rules; +---------+------+ | rule_id | hits | +---------+------+ | 3 | 0 | | 4 | 4 | | 5 | 0 | | 6 | 4 | +---------+------+ 4 rows in set (0.00 sec)
从上面看到，apply=0 & falgOUT=null，会继续往下找路由，找到了rule_id=6的，直接进行转发。apply=1 直接转发，flagOUT != null 直接转发。
小结：通过上面的测试说明ProxySQL只要设置好路由规则，可以有多个主库。
② 根据用户名路由
和多主路由一样，区别是写入到路由表的字段不一样：
admin@127.0.0.1 : (none) 06:09:20>INSERT INTO mysql_query_rules(active,username,match_pattern,destination_hostgroup,apply) VALUES(1,'proxysql','^SELECT.*FOR UPDATE$',101,1); Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 06:10:09>INSERT INTO mysql_query_rules(active,username,match_pattern,destination_hostgroup,apply) VALUES(1,'proxysql','^SELECT',1001,1); Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 06:10:32>select rule_id,active,username,schemaname,client_addr,destination_hostgroup,match_pattern,flagIN,flagOUT,apply from mysql_query_rules; +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ | rule_id | active | username | schemaname | client_addr | destination_hostgroup | match_pattern | flagIN | flagOUT | apply | +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ | 3 | 1 | NULL | NULL | NULL | 100 | ^SELECT.*FOR UPDATE$ | 0 | NULL | 1 | | 4 | 1 | NULL | NULL | NULL | 1000 | ^SELECT | 0 | NULL | 1 | | 1405 | 1 | proxysql | NULL | NULL | 101 | ^SELECT.*FOR UPDATE$ | 0 | NULL | 1 | | 1406 | 1 | proxysql | NULL | NULL | 1001 | ^SELECT | 0 | NULL | 1 | +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ 4 rows in set (0.00 sec)
3，flagIN/flahOUT规则链实现多实例（推荐）
和2中的条件一样，先配置ProxySQL的servers，users：
admin@127.0.0.1 : (none) 01:09:52>insert into mysql_servers(hostgroup_id,hostname,port,weight,max_replication_lag,comment) -> values -> (100, '192.168.200.202', 3306, 1, 10, 'ReadWrite'), -> (1000, '192.168.200.202', 3306, 1, 10, 'ReadWrite'), -> (1000, '192.168.200.132', 3306, 9, 10, 'ReadOnly'); Query OK, 3 rows affected (0.00 sec) admin@127.0.0.1 : (none) 01:09:54>insert into mysql_servers(hostgroup_id,hostname,port,weight,max_replication_lag,comment) -> values -> (101, '192.168.200.97', 3306, 1, 10, 'ReadWrite'), -> (1001, '192.168.200.97', 3306, 1, 10, 'ReadWrite'), -> (1001, '192.168.200.245', 3306, 9, 10, 'ReadOnly'); Query OK, 3 rows affected (0.01 sec) admin@127.0.0.1 : (none) 01:11:01>insert into mysql_users(username, password,active,default_hostgroup,transaction_persistent) -> values -> ('sbuser', 'sbuser', 1, 100, 1), -> ('proxysql', 'proxysql', 1, 101, 1); Query OK, 2 rows affected (0.00 sec) admin@127.0.0.1 : (none) 01:19:44>set mysql-monitor_username='proxysql'; Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 01:19:44>set mysql-monitor_password='proxysql'; Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 01:56:09>select hostgroup_id,hostname,port,status from mysql_servers; +--------------+-----------------+------+--------+ | hostgroup_id | hostname | port | status | +--------------+-----------------+------+--------+ | 100 | 192.168.200.202 | 3306 | ONLINE | | 1000 | 192.168.200.202 | 3306 | ONLINE | | 1000 | 192.168.200.132 | 3306 | ONLINE | | 101 | 192.168.200.97 | 3306 | ONLINE | | 1001 | 192.168.200.97 | 3306 | ONLINE | | 1001 | 192.168.200.245 | 3306 | ONLINE | +--------------+-----------------+------+--------+ 6 rows in set (0.00 sec) admin@127.0.0.1 : (none) 01:58:18>select username,default_hostgroup from mysql_users; +----------+-------------------+ | username | default_hostgroup | +----------+-------------------+ | sbuser | 100 | | proxysql | 101 | +----------+-------------------+
再配置flagOUT/flagIN，flag20是读，flag21是写：
admin@127.0.0.1 : (none) 01:21:34>INSERT INTO mysql_query_rules(rule_id,active,match_pattern,apply,flagOUT) VALUES(49,1,'^SELECT.*FOR UPDATE$',0,21); Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 01:27:18>INSERT INTO mysql_query_rules(rule_id,active,match_pattern,apply,flagOUT) VALUES(50,1,'^SELECT',0,20); Query OK, 1 row affected (0.00 sec) admin@127.0.0.1 : (none) 01:32:11>insert into mysql_query_rules(active,schemaname,destination_hostgroup,apply,flagIN,flagOUT) values -> (1,'sbtest',100,1,21,21), -> (1,'proxysql',101,1,21,21); Query OK, 2 rows affected (0.00 sec) admin@127.0.0.1 : (none) 01:32:53>insert into mysql_query_rules(active,schemaname,destination_hostgroup,apply,flagIN,flagOUT) values -> (1,'sbtest',1000,1,20,20), -> (1,'proxysql',1001,1,20,20); Query OK, 2 rows affected (0.00 sec) admin@127.0.0.1 : (none) 01:58:28>select rule_id,active,username,schemaname,client_addr,destination_hostgroup,match_pattern,flagIN,flagOUT,apply from mysql_query_rules; +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ | rule_id | active | username | schemaname | client_addr | destination_hostgroup | match_pattern | flagIN | flagOUT | apply | +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+ | 49 | 1 | NULL | NULL | NULL | NULL | ^SELECT.*FOR UPDATE$ | 0 | 21 | 0 | | 50 | 1 | NULL | NULL | NULL | NULL | ^SELECT | 0 | 20 | 0 | | 51 | 1 | NULL | sbtest | NULL | 100 | NULL | 21 | 21 | 1 | | 52 | 1 | NULL | proxysql | NULL | 101 | NULL | 21 | 21 | 1 | | 53 | 1 | NULL | sbtest | NULL | 1000 | NULL | 20 | 20 | 1 | | 54 | 1 | NULL | proxysql | NULL | 1001 | NULL | 20 | 20 | 1 | +---------+--------+----------+------------+-------------+-----------------------+----------------------+--------+---------+-------+
最后保存上线：
-- 应用 load mysql users to runtime; load mysql servers to runtime; load mysql variables to runtime; LOAD MYSQL QUERY RULES TO RUN; -- 保存到磁盘 save mysql users to disk; save mysql servers to disk; save mysql variables to disk; SAVE MYSQL QUERY RULES TO DISK; save mysql users to mem; -- 可以屏蔽看到的明文密码
app连接测试：
1）连接实例202
[zhoujy@localhost ~]$ mysql -usbuser -psbuser -h192.168.200.24 -P6033 ... sbuser@192.168.200.24 : (none) 02:19:41>show databases; +--------------------+ | Database | +--------------------+ | information_schema | | sbtest | +--------------------+ 2 rows in set (0.00 sec) sbuser@192.168.200.24 : (none) 02:19:44>use sbtest Database changed sbuser@192.168.200.24 : sbtest 02:19:48>show tables; ... sbuser@192.168.200.24 : sbtest 02:19:57>insert into x values(10000); ... sbuser@192.168.200.24 : sbtest 02:20:10>select * from x; ...
相关信息：路由的信息都是正确的
admin@127.0.0.1 : (none) 02:24:15>select hostgroup,schemaname,username,digest_text,count_star from stats_mysql_query_digest; +-----------+--------------------+----------+----------------------------------+------------+ | hostgroup | schemaname | username | digest_text | count_star | +-----------+--------------------+----------+----------------------------------+------------+ | 1000 | sbtest | sbuser | select * from x | 5 | | 100 | sbtest | sbuser | insert into x values(?) | 5 | | 100 | sbtest | sbuser | show tables | 2 | | 100 | sbtest | sbuser | show databases | 1 | | 100 | information_schema | sbuser | SELECT DATABASE() | 1 | | 100 | information_schema | sbuser | show databases | 1 | | 100 | information_schema | sbuser | select USER() | 1 | | 100 | information_schema | sbuser | select @@version_comment limit ? | 1 | +-----------+--------------------+----------+----------------------------------+------------+ 8 rows in set (0.00 sec)
--路由命中 admin@127.0.0.1 : (none) 02:25:13>admin@1* from stats_mysql_query_rules; +---------+------+ | rule_id | hits | +---------+------+ | 49 | 0 | | 50 | 6 | | 51 | 0 | | 52 | 0 | | 53 | 5 | | 54 | 0 | +---------+------+ 6 rows in set (0.00 sec)
结论：通过实例202的账号访问ProxySQL，首先会检查flagIN=0，在其上面进行匹配（Proxysql入口都是flagIN =0，顺序往下）， 匹配到之后检查flagOUT，发现 flagOUT不为NULL且flagIN !（0）= flagOUT （20），则进入以flagIN为上一个flagOUT值的新规则链，即20。再去flagIN=20里匹配，最终匹配到了rule_id=53的规则，最后转发。
2）连接实例97
相关情况和上面一样，最终通过rule_id=54进行转发。
建议：若要用ProxySQL来控制多主从实例的读写分离，推荐使用flagIN/flahOUT规则链实现多实例。
4，flagIN/flahOUT规则链实现分库
目的：客户端应用连接上 proxysql 的ip:port，连接时指定分库db名，自动路由到对应的实例、对应的库。
① ：环境
APP：192.168.200.25、192.168.200.64 M1： IP：192.168.200.202 Port：3306 DB：M1、M2、M3 S1： IP：192.168.200.132 Port：3306 DB：M1、M2、M3 M2： IP：192.168.200.97 Port：3306 DB：M4、M5、M6 S2： IP：192.168.200.245 Port：3306 DB：M4、M5、M6 ProxySQL：192.168.200.24
② 搭建
和之前一样先在后端数据库创建程序和检测账号：
--程序账号 GRANT SELECT,INSERT,UPDATE,DELETE ON `mtest%`.* TO 'mtest'@'192.168.200.24' identified by 'mtest'; --健康检测账号 GRANT SUPER, REPLICATION CLIENT ON *.* TO 'proxysql'@'192.168.200.24' IDENTIFIED BY 'proxysql';
配置ProxySQL：
--插入后端用户信息 insert into mysql_users(username, password,active,transaction_persistent) values('mtest','mtest',1,1); --插入后端数据库信息 insert into mysql_servers(hostgroup_id,hostname,port,weight,max_replication_lag,comment) values(100,'192.168.200.202',3306,1,10,'test proxysql'); insert into mysql_servers(hostgroup_id,hostname,port,weight,max_replication_lag,comment) values(1000,'192.168.200.132',3306,9,10,'test proxysql'); insert into mysql_servers(hostgroup_id,hostname,port,weight,max_replication_lag,comment) values(1000,'192.168.200.202',3306,1,10,'test proxysql'); insert into mysql_servers(hostgroup_id,hostname,port,weight,max_replication_lag,comment) values(101,'192.168.200.97',3306,1,10,'test proxysql'); insert into mysql_servers(hostgroup_id,hostname,port,weight,max_replication_lag,comment) values(1001,'192.168.200.245',3306,9,10,'test proxysql'); insert into mysql_servers(hostgroup_id,hostname,port,weight,max_replication_lag,comment) values(1001,'192.168.200.97',3306,1,10,'test proxysql'); --配置健康检测信息 set mysql-monitor_username='proxysql'; set mysql-monitor_password='proxysql';
应用保存配置：
-- 应用 load mysql users to runtime; load mysql servers to runtime; load mysql variables to runtime; -- 保存到磁盘 save mysql users to disk; save mysql servers to disk; save mysql variables to disk; save mysql users to mem; -- 可以屏蔽看到的明文密码
配置路由规则：
----添加读写分离的路由 --写:写的入口 flagIN=0 INSERT INTO mysql_query_rules(rule_id,active,match_pattern,apply,flagOUT) VALUES(49,1,'^SELECT.*FOR UPDATE$',0,21); --读：读的入口 flagIN=0 INSERT INTO mysql_query_rules(rule_id,active,match_pattern,apply,flagOUT) VALUES(50,1,'^SELECT',0,20); --反向匹配，相当于对 match_digest/match_pattern 的匹配取反，非select，即写 INSERT INTO mysql_query_rules(rule_id, active,match_pattern,negate_match_pattern,apply,flagOUT) values(60, 1,'^SELECT',1,0,21); ----为后端服务器配置路由 --读 insert into mysql_query_rules(active,schemaname,destination_hostgroup,apply,flagIN,flagOUT) values(1,'M1',1000,1,20,20),(1,'M2',1000,1,20,20),(1,'M3',1000,1,20,20); insert into mysql_query_rules(active,schemaname,destination_hostgroup,apply,flagIN,flagOUT) values(1,'M4',1001,1,20,20),(1,'M5',1001,1,20,20),(1,'M6',1001,1,20,20); --写 insert into mysql_query_rules(active,schemaname,destination_hostgroup,apply,flagIN,flagOUT) values(1,'M1',100,1,21,21),(1,'M2',100,1,21,21),(1,'M3',100,1,21,21); insert into mysql_query_rules(active,schemaname,destination_hostgroup,apply,flagIN,flagOUT) values(1,'M4',101,1,21,21),(1,'M5',101,1,21,21),(1,'M6',101,1,21,21); --连接时，若没有指定数据库，则进行show databases/tables、use 等会超时出错，连接时，默认的数据库是在information_schema，所以写一条根据information_schema库的路由,直接返回错误信息。 insert into mysql_query_rules(rule_id,active,schemaname,apply,flagOUT) values(20,1,'information_schema',0,302); insert into mysql_query_rules(rule_id,active,apply, flagIN,flagOUT,error_msg) values(9999,1,1, 302,302,'No query rules matched (by ProxySQL)'); --连接时，若没有指定数据库，则可以使用 schemaname.tablename 的形式匹配数据。 insert into mysql_query_rules(rule_id,active,match_pattern,destination_hostgroup,apply,flagIN,flagOUT) values(1000,1,'([s\`])M(1|2|3)([.\`])',100,1,302,302); insert into mysql_query_rules(rule_id,active,match_pattern,destination_hostgroup,apply,flagIN,flagOUT) values(1001,1,'([s\`])M(4|5|6)([.\`])',101,1,302,302);
应用规则：
LOAD MYSQL QUERY RULES TO RUN; SAVE MYSQL QUERY RULES TO DISK;
最终的路由规则如下：
select rule_id,schemaname,destination_hostgroup,match_pattern,negate_match_pattern,flagIN,flagOUT,apply,error_msg from mysql_query_rules;
app连接测试：
~$ mysql -umtest -pmtest -h192.168.200.24 -P6033 -A ... mtest@192.168.200.24 : (none) 11:27:29>show databases; --触发了定义的路由 ERROR 1148 (42000): No query rules matched (by ProxySQL) mtest@192.168.200.24 : (none) 11:27:34>select * from M5.mtest5; --可以直接用schema.tables 访问 +------+ | id | +------+ | 5 | | 55 | +------+ 2 rows in set (0.00 sec) mtest@192.168.200.24 : (none) 11:27:47>use M1 --切换数据库 Database changed mtest@192.168.200.24 : M1 11:27:52>show tables; --可以show了 +--------------+ | Tables_in_M1 | +--------------+ | mtest1 | +--------------+ 1 row in set (0.00 sec) mtest@192.168.200.24 : M1 11:27:56>select * from mtest1; +------+ | id | +------+ | 1 | | 11 | | 111 | | 1111 | +------+ 4 rows in set (0.00 sec) mtest@192.168.200.24 : M1 11:28:02>insert into mtest1 values(11111); Query OK, 1 row affected (0.00 sec) mtest@192.168.200.24 : M1 11:28:11>select * from mtest1; +-------+ | id | +-------+ | 1 | | 11 | | 111 | | 1111 | | 11111 | +-------+ 5 rows in set (0.00 sec) mtest@192.168.200.24 : M1 11:28:12>show databases; --可以show了 +--------------------+ | Database | +--------------------+ | information_schema | | M1 | | M2 | | M3 | +--------------------+ 4 rows in set (0.00 sec) mtest@192.168.200.24 : M1 11:28:20>use M5 --切换到另一个实例的db Database changed mtest@192.168.200.24 : M5 11:28:52>show databases; +--------------------+ | Database | +--------------------+ | information_schema | | M4 | | M5 | | M6 | +--------------------+ 4 rows in set (0.00 sec) mtest@192.168.200.24 : M5 11:28:55>select * from mtest5; +------+ | id | +------+ | 5 | | 55 | +------+ 2 rows in set (0.00 sec) mtest@192.168.200.24 : M5 11:29:03>insert into mtest5 values(555); Query OK, 1 row affected (0.01 sec) mtest@192.168.200.24 : M5 11:29:12>select * from mtest5; +------+ | id | +------+ | 5 | | 55 | | 555 | +------+ 3 rows in set (0.00 sec)
查看路由命中率：
select active,hits, mysql_query_rules.rule_id, schemaname,match_pattern,destination_hostgroup hostgroup,flagIn,flagOUT FROM mysql_query_rules NATURAL JOIN stats.stats_mysql_query_rules ORDER BY mysql_query_rules.rule_id;
查看SQL统计信息：
admin@127.0.0.1 : (none) 11:36:46>select hostgroup,schemaname,username,substr(digest_text,120,-120),count_star from stats_mysql_query_digest; +-----------+--------------------+----------+----------------------------------+------------+ | hostgroup | schemaname | username | substr(digest_text,120,-120) | count_star | +-----------+--------------------+----------+----------------------------------+------------+ | 101 | M5 | mtest | show databases | 1 | | 1000 | M1 | mtest | SELECT DATABASE() | 1 | | 101 | M5 | mtest | insert into mtest5 values(?) | 1 | | 100 | M1 | mtest | show databases | 1 | | 100 | M1 | mtest | insert into mtest1 values(?) | 1 | | 1000 | M1 | mtest | select * from mtest1 | 2 | | 1001 | M5 | mtest | select * from mtest5 | 2 | | 100 | M1 | mtest | show tables | 1 | | 101 | information_schema | mtest | select * from M5.mtest5 | 1 | | 0 | information_schema | mtest | show databases | 1 | | 0 | information_schema | mtest | SELECT DATABASE() | 1 | | 0 | information_schema | mtest | select USER() | 1 | | 0 | information_schema | mtest | select @@version_comment limit ? | 1 | +-----------+--------------------+----------+----------------------------------+------------+
具体的说明可以看ProxySQL之读写分离与分库路由演示，到此读写分离的测试介绍完毕，
5，查询重写
查询重写这种对于线上环境SQL问题引起的紧急故障处理还是很有用处的。如果定位到了问题所在，必须修改SQL，时间紧急，这时查询重写这个东西就非常有用了。类似于MySQL5.7的查询重写插件。这里做下相关的说明：
ProxySQL的核心就是路由，查询重写也只是添加一条路由而已，在4的基础上进行测试：
select * from mtest1 order by id 重写成 select * from mtest1
添加路由：
--查询的路由，flagIN=0，当匹配上规则后进行重写，并且不应用，而通过flagOUT下去继续查询 INSERT INTO mysql_query_rules (rule_id,active,match_pattern,replace_pattern,apply,flagOUT) VALUES (48,1,'(.*)order by id','1',0,20);
其实查询重写的实现在proxysql中也实现为正则匹配替换，表示当proxysql匹配到<若干字符>order by id这个模式后，就将这个模式的order by id去掉。那么1是什么意思呢，就是sed的向前引用。
加载load和save完rules之后，查看是否重写成功：
--初始 #查询路由命中信息 admin@127.0.0.1 : (none) 02:44:52>select * from stats_mysql_query_rules; +---------+------+ | rule_id | hits | +---------+------+ | 20 | 0 | | 48 | 0 | | 49 | 0 | | 50 | 0 | | 60 | 0 | | 61 | 0 | | 62 | 0 | | 63 | 0 | | 64 | 0 | | 65 | 0 | | 66 | 0 | | 67 | 0 | | 68 | 0 | | 69 | 0 | | 70 | 0 | | 71 | 0 | | 72 | 0 | | 1000 | 0 | | 1001 | 0 | | 9999 | 0 | +---------+------+ 20 rows in set (0.00 sec) #查询统计信息 admin@127.0.0.1 : (none) 02:45:09>select * from stats_mysql_query_digest; Empty set (0.00 sec) --操作 ~$ mysql -umtest -pmtest -h192.168.200.24 -P6033 -A ... mtest@192.168.200.24 : (none) 02:45:27>use M1 Database changed mtest@192.168.200.24 : M1 02:45:31>show tables; +--------------+ | Tables_in_M1 | +--------------+ | mtest1 | +--------------+ 1 row in set (0.00 sec) mtest@192.168.200.24 : M1 02:45:33>select * from mtest1; +-------+ | id | +-------+ | 1 | | 11 | | 111 | | 1111 | | 11111 | +-------+ 5 rows in set (0.00 sec) mtest@192.168.200.24 : M1 02:45:37>select * from mtest1 order by id; +-------+ | id | +-------+ | 1 | | 11 | | 111 | | 1111 | | 11111 | +-------+ 5 rows in set (0.00 sec) mtest@192.168.200.24 : M1 02:45:46>select * from mtest1 order by id; +-------+ | id | +-------+ | 1 | | 11 | | 111 | | 1111 | | 11111 | +-------+ 5 rows in set (0.01 sec) ----以上执行了2次order by id和1此没有order by id的查询，去查询统计应该得到的值是3次没有order by id的查询。 --验证 #查询统计信息，查看没有order by id的SQL出现了3次，没有出现有order by id的SQL admin@127.0.0.1 : (none) 02:49:49>select hostgroup,schemaname,digest_text,count_star from stats_mysql_query_digest; +-----------+--------------------+----------------------------------+------------+ | hostgroup | schemaname | digest_text | count_star | +-----------+--------------------+----------------------------------+------------+ | 1000 | M1 | select * from mtest1 | 3 | | 100 | M1 | show tables | 1 | | 0 | information_schema | SELECT DATABASE() | 1 | | 0 | information_schema | select USER() | 1 | | 0 | information_schema | select @@version_comment limit ? | 1 | +-----------+--------------------+----------------------------------+------------+ 5 rows in set (0.00 sec) #重写查询的路由命中了2次 admin@127.0.0.1 : (none) 02:50:12>select * from stats_mysql_query_rules; +---------+------+ | rule_id | hits | +---------+------+ | 20 | 1 | | 48 | 2 | | 49 | 0 | | 50 | 1 | | 60 | 1 | | 61 | 3 | | 62 | 0 | | 63 | 0 | | 64 | 0 | | 65 | 0 | | 66 | 0 | | 67 | 1 | | 68 | 0 | | 69 | 0 | | 70 | 0 | | 71 | 0 | | 72 | 0 | | 1000 | 0 | | 1001 | 0 | | 9999 | 1 | +---------+------+ 20 rows in set (0.00 sec)
从上面的结果看，查询重写已经测试通过。到此，关于ProxySQL的相关测试已经结束，下面分析下和DBProxy的特性差别和性能差异。
性能测试
环境：
ProxySQL：192.168.200.24 DBProxy ：192.168.200.24 M： IP：192.168.200.202 Port：3306 DB：sbtest S： IP：192.168.200.132 Port：3306 DB：sbtest
读写混合(oltp_read_write.lua)测试对比：
直连数据库：
./bin/sysbench --test=./share/sysbench/oltp_read_write.lua --mysql-host=192.168.200.202 --mysql-port=3306 --mysql-user=sbuser --mysql-password=sbuser --mysql-db=sbtest --report-interval=10 --max-requests=0 --time=120 --threads=1 --tables=3 --table-size=1000000 prepare/run/cleanup
ProxySQL连接数据库：
./bin/sysbench --test=./share/sysbench/oltp_read_write.lua --mysql-host=192.168.200.24 --mysql-port=6033 --mysql-user=sbuser --mysql-password=sbuser --mysql-db=sbtest --report-interval=10 --max-requests=0 --time=120 --threads=1 --tables=3 --table-size=1000000 --skip-trx=on --db-ps-mode=disable --mysql-ignore-errors=1062 prepare/run/cleanup
DBProxy连接数据库
./bin/sysbench --test=./share/sysbench/oltp_read_write.lua --mysql-host=192.168.200.24 --mysql-port=3308 --mysql-user=sbuser --mysql-password=sbuser --mysql-db=sbtest --report-interval=10 --max-requests=0 --time=120 --threads=1 --tables=3 --table-size=1000000 --skip-trx=on --db-ps-mode=disable --mysql-ignore-errors=1062 prepare/run/cleanup
测试的线程：1、2、4、8、16、32、64、128
把上面数据以曲线图的形式表现：
TPS：
QPS：
测试小结：
在读写混合的模式下：线程越少差距越大，测试结果和美团点评DBProxy读写分离使用说明里的测试报告基本吻合，这里主要对比ProxySQL和DBProxy的性能情况，从上图看到二者性能差不多，不过DBProxy的CPU消耗是ProxySQL的1到1.5倍。
总结：
通过上面的一些基本介绍，大致了解了ProxySQL读写分离功能的使用，关于ProxySQL的其他功能内容在手册里有了详尽的介绍，具体的情况请参考手册说明。现在大致整理下ProxySQL和DBproxy的差别:
①：连接池，是 multiplexing。
②：强大的正则路由，可以自己干预读写路由算法。
③：从库不可用自动下线，不需要人为干预，支持多主库。
④：支持重写SQL。
⑤：足够轻量，配置简单。
但是在安全配置上面，DBProxy比ProxySQL要强，ProxySQL前后端账号未分离，可以通过mysql_users查看，前后端公用一个账号，但是在runting_mysql_users里面前后端账号是分离的（backend、frontend）。其他的相关安全可以参考美团点评DBProxy读写分离使用说明。最后根据情况选择到底使用哪个proxy，要是使用的是MySQL Server 5.7，因DBProxy没有对5.7进行测试，所以推荐使用ProxySQL。至于如何防止ProxySQL的单点问题，也可以用lvs来解决，具体的说可以看LVS+Keepalived实现DBProxy的高可用。
参考文档
https://github.com/sysown/proxysql/wiki
http://proxysql.blogspot.jp/2015/09/proxysql-tutorial-setup-in-mysql.html
https://severalnines.com/blog/how-set-read-write-split-galera-cluster-using-proxysql