环境说明
| 地址 | 端口 | 角色 | 说明 |
|---|---|---|---|
| 127.0.0.1 | 3306 | master | - |
| 127.0.0.1 | 3307 | slave | - |
| 127.0.0.1 | 6032/6033 | ProxySQL | 版本2.0.1 |
| 127.0.0.1 | 10001 | replication-manager | - |
一、MySQL搭建
这里省略，推荐增强半同步
复制账号：repl/repl4slave
二、ProxySQL
2.1安装
[root@izbp12nspj47ypto9t6vyez src]# rpm -ivh proxysql-2.0.1-1-centos7.x86_64.rpm
Preparing... ################################# [100%]
Updating / installing...
1:proxysql-2.0.1-1 warning: group proxysql does not exist - using root
################################# [100%]
2.2用户配置
在master上创建连接账户及监控账户
grant select,update,delete,insert on *.* to xucl@'%' identified by 'xuclxucl';
grant replication client on *.* to monitor@'%' identified by 'monitor';
2.3server配置
w_hg : 10
ro_hg : 11
监控账号
monitor monitor
配置读写组
insert into mysql_replication_hostgroups(writer_hostgroup,reader_hostgroup) values(10,11);
配置MySQL实例
insert into mysql_servers(hostgroup_id,hostname,port,max_connections,max_replication_lag) values(10,'127.0.0.1',3306,100,300),(11,'127.0.0.1',3307,100,300);
配置USERS
insert into mysql_users(username,password,default_hostgroup,default_schema,max_connections) values('xucl','xuclxucl',10,'xucl',90);
配置监控账号
set mysql-monitor_username = 'monitor';
set mysql-monitor_password = 'monitor';
save mysql users to disk;
load mysql users to runtime;
save mysql servers to disk;
load mysql servers to runtime;
save mysql variables to disk;
load mysql variables to runtime;
2.4配置读写分离
INSERT INTO mysql_query_rules (rule_id,active,match_digest,destination_hostgroup,apply)
VALUES
(1,1,'^SELECT.*FOR UPDATE$',10,1),
(2,1,'^SELECT',11,1);
LOAD MYSQL QUERY RULES TO RUNTIME;
SAVE MYSQL QUERY RULES TO DISK;
2.5简单测试读写分离
(xucl:db1@xucl:09:29:57)[(none)]> select @@port;
+--------+
| @@port |
+--------+
| 3307 |
+--------+
1 row in set (0.05 sec)
(xucl:db1@xucl:09:30:00)[(none)]> select @@port for update;
+--------+
| @@port |
+--------+
| 3306 |
+--------+
1 row in set (0.00 sec)
三、replication-manager
3.1安装
rpm -ivh replication-manager-osc-2.0.0_rc2-1.x86_64.rpm
3.2配置文件
[root@izbp12nspj47ypto9t6vyez src]# cat /etc/replication-manager/config.toml
[db3306]
title = "db3306"
db-servers-hosts = "127.0.0.1:3306,127.0.0.1:3307"
db-servers-prefered-master = "127.0.0.1:3306"
db-servers-credential = "xucl:xuclxucl"
replication-credential = "repl:repl4slave"
failover-mode = "manual"
proxysql=true
proxysql-server="127.0.0.1"
proxysql-port=6033
proxysql-admin-port=6032
proxysql-writer-hostgroup=10
proxysql-reader-hostgroup=11
proxysql-user="admin"
proxysql-password="admin"
proxysql-bootstrap=false
[Default]
monitoring-datadir = "/data/replication-manager"
monitoring-sharedir = "/data/share_replication-manager"
log-level=7
log-file = "/var/log/replication-manager.log"
replication-multi-master = false
replication-multi-tier-slave = false
failover-readonly-state = true
http-server = true
http-bind-address = "0.0.0.0"
http-port = "10001"
3.3启动replication-manager
[root@izbp12nspj47ypto9t6vyez src]# service replication-manager restart
Restarting replication-manager (via systemctl): [ OK ]
日志
2019/02/22 09:34:38 [db3306] INFO - Failover in interactive mode
2019/02/22 09:34:38 [db3306] INFO - Loading 1 proxies
2019/02/22 09:34:38 [db3306] INFO - Loading ProxySQL...
2019/02/22 09:34:38 [db3306] DEBUG - Monitoring server loop
2019/02/22 09:34:38 [db3306] DEBUG - Server [0]: URL: 127.0.0.1:3306 State: Suspect PrevState: Suspect
2019/02/22 09:34:38 [db3306] DEBUG - Server [1]: URL: 127.0.0.1:3307 State: Suspect PrevState: Suspect
2019/02/22 09:34:38 [db3306] DEBUG - State unconnected set by non-master rule on server 127.0.0.1:3307
2019/02/22 09:34:38 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:38 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:38 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:38 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:38 [db3306] DEBUG - GetMasterFromReplication server 3307 lookup if server 127.0.0.1:3307 is the one : 3307
2019/02/22 09:34:38 [db3306] DEBUG - Server 127.0.0.1:3306 is configured as a slave
2019/02/22 09:34:38 [db3306] DEBUG - Privilege check on 127.0.0.1:3306
2019/02/22 09:34:38 [db3306] DEBUG - Client connection found on server 127.0.0.1:3306 with IP 127.0.0.1 for host 127.0.0.1
2019/02/22 09:34:38 [db3306] DEBUG - Server 127.0.0.1:3307 was set master as last non slave
2019/02/22 09:34:38 [db3306] DEBUG - Privilege check on 127.0.0.1:3307
2019/02/22 09:34:38 [db3306] DEBUG - Client connection found on server 127.0.0.1:3307 with IP 127.0.0.1 for host 127.0.0.1
2019/02/22 09:34:38 [db3306] DEBUG - GetMasterFromReplication server 3307 lookup if server 127.0.0.1:3307 is the one : 3307
2019/02/22 09:34:38 [db3306] DEBUG - Checking if server 127.0.0.1 is a slave of server 127.0.0.1
2019/02/22 09:34:38 [db3306] DEBUG - GetMasterFromReplication server 3307 lookup if server 127.0.0.1:3307 is the one : 3307
2019/02/22 09:34:38 [db3306] INFO - Init Proxy Type: proxysql Host: 127.0.0.1 Port: 6032
2019/02/22 09:34:38 [db3306] STATE - OPENED WARN0048 : No semisync settings on slave 127.0.0.1:3306
2019/02/22 09:34:38 [db3306] STATE - OPENED WARN0058 : No GTID strict mode on slave 127.0.0.1:3306
2019/02/22 09:34:38 [db3306] STATE - OPENED WARN0060 : No semisync settings on master 127.0.0.1:3307
2019/02/22 09:34:38 [db3306] STATE - OPENED WARN0070 : No GTID strict mode on master 127.0.0.1:3307
2019/02/22 09:34:38 [db3306] STATE - OPENED ERR00021 : All cluster db servers down
2019/02/22 09:34:38 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:38 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:39 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:39 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:39 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:39 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:40 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:40 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:40 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:40 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:40 [db3306] DEBUG - Monitoring server loop
2019/02/22 09:34:40 [db3306] DEBUG - Server [0]: URL: 127.0.0.1:3306 State: Slave PrevState: Slave
2019/02/22 09:34:40 [db3306] DEBUG - Server [1]: URL: 127.0.0.1:3307 State: Master PrevState: StandAlone
2019/02/22 09:34:40 [db3306] DEBUG - Master [ ]: URL: 127.0.0.1:3307 State: Master PrevState: StandAlone
2019/02/22 09:34:40 [db3306] DEBUG - Slave [0]: URL: 127.0.0.1:3306 State: Slave PrevState: Slave
2019/02/22 09:34:40 [db3306] DEBUG - Server 127.0.0.1:3306 is configured as a slave
2019/02/22 09:34:40 [db3306] DEBUG - Privilege check on 127.0.0.1:3306
2019/02/22 09:34:40 [db3306] DEBUG - Client connection found on server 127.0.0.1:3306 with IP 127.0.0.1 for host 127.0.0.1
2019/02/22 09:34:40 [db3306] DEBUG - Server 127.0.0.1:3307 was set master as last non slave
2019/02/22 09:34:40 [db3306] DEBUG - Privilege check on 127.0.0.1:3307
2019/02/22 09:34:40 [db3306] DEBUG - Client connection found on server 127.0.0.1:3307 with IP 127.0.0.1 for host 127.0.0.1
2019/02/22 09:34:40 [db3306] DEBUG - GetMasterFromReplication server 3307 lookup if server 127.0.0.1:3307 is the one : 3307
2019/02/22 09:34:40 [db3306] DEBUG - Checking if server 127.0.0.1 is a slave of server 127.0.0.1
2019/02/22 09:34:40 [db3306] DEBUG - GetMasterFromReplication server 3307 lookup if server 127.0.0.1:3307 is the one : 3307
2019/02/22 09:34:40 [db3306] STATE - RESOLV ERR00021 : All cluster db servers down
2019/02/22 09:34:41 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:41 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:41 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:41 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:42 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:42 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:42 [db3306] DEBUG - Monitoring server loop
2019/02/22 09:34:42 [db3306] DEBUG - Server [0]: URL: 127.0.0.1:3306 State: Slave PrevState: Slave
2019/02/22 09:34:42 [db3306] DEBUG - Server [1]: URL: 127.0.0.1:3307 State: Master PrevState: Master
2019/02/22 09:34:42 [db3306] DEBUG - Master [ ]: URL: 127.0.0.1:3307 State: Master PrevState: Master
2019/02/22 09:34:42 [db3306] DEBUG - Slave [0]: URL: 127.0.0.1:3306 State: Slave PrevState: Slave
2019/02/22 09:34:42 [db3306] DEBUG - Server 127.0.0.1:3306 is configured as a slave
2019/02/22 09:34:42 [db3306] DEBUG - Privilege check on 127.0.0.1:3306
2019/02/22 09:34:42 [db3306] DEBUG - Client connection found on server 127.0.0.1:3306 with IP 127.0.0.1 for host 127.0.0.1
2019/02/22 09:34:42 [db3306] DEBUG - Server 127.0.0.1:3307 was set master as last non slave
2019/02/22 09:34:42 [db3306] DEBUG - Privilege check on 127.0.0.1:3307
2019/02/22 09:34:42 [db3306] DEBUG - Client connection found on server 127.0.0.1:3307 with IP 127.0.0.1 for host 127.0.0.1
2019/02/22 09:34:42 [db3306] DEBUG - GetMasterFromReplication server 3307 lookup if server 127.0.0.1:3307 is the one : 3307
2019/02/22 09:34:42 [db3306] DEBUG - Checking if server 127.0.0.1 is a slave of server 127.0.0.1
2019/02/22 09:34:42 [db3306] DEBUG - GetMasterFromReplication server 3307 lookup if server 127.0.0.1:3307 is the one : 3307
2019/02/22 09:34:43 [db3306] DEBUG - Lookup server 127.0.0.1:3306 if maxscale binlog server: 127.0.0.1:3306
2019/02/22 09:34:43 [db3306] DEBUG - Lookup server 127.0.0.1:3307 if maxscale binlog server: 127.0.0.1:3306
3.4切换
手动点击switch over即可实现切换
切换日志
2019/02/22 09:38:22 [db3306] INFO - Rest API receive Switchover request
2019/02/22 09:38:22 [db3306] INFO - Signaling Switchover...
2019/02/22 09:38:22 [db3306] INFO - --------------------------
2019/02/22 09:38:22 [db3306] INFO - Starting master switchover
2019/02/22 09:38:22 [db3306] INFO - --------------------------
2019/02/22 09:38:22 [db3306] INFO - Checking long running updates on master 10
2019/02/22 09:38:22 [db3306] INFO - Flushing tables on master 127.0.0.1:3307
2019/02/22 09:38:22 [db3306] INFO - Electing a new master
2019/02/22 09:38:22 [db3306] INFO - Slave 127.0.0.1:3306 has been elected as a new master
2019/02/22 09:38:22 [db3306] INFO - Terminating all threads on 127.0.0.1:3307
2019/02/22 09:38:22 [db3306] INFO - Rejecting updates on 127.0.0.1:3307 (old master)
2019/02/22 09:38:22 [db3306] INFO - Waiting for candidate master to apply relay log
2019/02/22 09:38:22 [db3306] INFO - Reading all relay logs on 127.0.0.1:3306
2019/02/22 09:38:22 [db3306] INFO - Stopping slave thread on new master
2019/02/22 09:38:22 [db3306] INFO - Failover Proxy Type: proxysql Host: 127.0.0.1 Port: 6032
2019/02/22 09:38:22 [db3306] INFO - Resetting slave on new master and set read/write mode on
2019/02/22 09:38:22 [db3306] ERROR - Could not set new master as read-write
2019/02/22 09:38:22 [db3306] INFO - Inject fake transaction on new master 127.0.0.1:3306
2019/02/22 09:38:22 [db3306] INFO - Switching old master as a slave
2019/02/22 09:38:22 [db3306] INFO - Doing MySQL GTID switch of the old master
2019/02/22 09:38:22 [db3306] INFO - Switching other slaves to the new master
2019/02/22 09:38:22 [db3306] INFO - Master switch on 127.0.0.1:3306 complete
2019/02/22 09:38:22 [db3306] STATE - OPENED ERR00041 : Skip slave in election 127.0.0.1:3307 has more than 30 seconds of replication delay (35484)
2019/02/22 09:38:22 [db3306] STATE - OPENED ERR00039 : Skip slave in election 127.0.0.1:3307 repl not electable
2019/02/22 09:38:22 [db3306] STATE - OPENED ERR00032 : No candidates found in slaves list
2019/02/22 09:38:34 [db3306] STATE - RESOLV ERR00041 : Skip slave in election 127.0.0.1:3307 has more than 30 seconds of replication delay (35481)
2019/02/22 09:38:34 [db3306] STATE - RESOLV ERR00039 : Skip slave in election 127.0.0.1:3307 repl not electable
2019/02/22 09:38:34 [db3306] STATE - RESOLV ERR00032 : No candidates found in slaves list
四、replication切换逻辑
在线切换逻辑
failover切换逻辑
总结
-
总得来说这是一套简单又非常高效的MySQL高可用方案，并且是可上生产的方案，并且replication manager又自带web界面，操作起来非常简单
-
这个方案前面有ProxySQL，可以实现读写分离，应用程序连接ProxySQL即可
-
replication manager负责数据库高可用切换，切换完成后，ProxySQL通过read only感知读写节点，对应用程序来说可能会有短暂的影响，确保程序重连机制
-
看到replication manager的在线切换和failover切换逻辑，涉及到多个步骤的freeze，具体整个实现逻辑与过程将在下篇文章进行剖析
- https://www.modb.pro/db/28674
- https://blog.csdn.net/zengxuewen2045/article/details/77712083
- https://www.yisu.com/zixun/27090.html
- http://t.zoukankan.com/allenhu320-p-11339057.html
- https://blog.csdn.net/aohexi1719/article/details/101973393