关于缓存Redis主从配置

16-11-23 来源：[db:作者]

收藏我要投稿

redis主从复制过程：

当配置好slave后，slave与master建立连接，然后发送sync命令。无论是第一次连接还是重新连接，master都会启动一个后台进程，将数据库快照保存到文件中，同时master主进程会开始收集新的写命令并缓存。后台进程完成写文件后，master就发送文件给slave，slave将文件保存到硬盘上，再加载到内存中，接着master就会把缓存的命令转发给slave，后续master将收到的写命令发送给slave。如果master同时收到多个slave发来的同步连接命令，master只会启动一个进程来写数据库镜像，然后发送给所有的slave。master同步数据时是非阻塞式的，可以接收用户的读写请求。然而在slave端是阻塞模式的，slave在同步master数据时，并不能够响应客户端的查询。

可以在master禁用数据持久化，只需要注释掉master 配置文件中的所有save配置，然后只在slave上配置数据持久化

拥有主从服务器的好处（从服务器是只读的，可以一主多从）

1.主服务器进行读写时，会转移到从读，减轻服务器压力

2.热备份主从都可以设置密码，也可以密码不一致

进入/usr/data/redis/slave

创建 master slave1 slave2

1.复制redis.conf到3个目录，修改端口 1000，2000，3000

2.修改pid路径，日志路径

pidfile /usr/data/redis/slave/master/redis.pid

logfile /usr/data/redis/slave/master/redis.log

ps -ef | grep redis            
root     19000     1  0 08:27 ?        00:00:00 redis-server 192.168.1.1:1000
root     19012     1  0 08:27 ?        00:00:00 redis-server 192.168.1.1:2000
root     19016     1  0 08:27 ?        00:00:00 redis-server 192.168.1.1:3000

连接客户端

[root@iZ23pv5rps8Z ~]# redis-cli -h 192.168.1.1 -p 3000

查看权限

192.168.1.1:3000> info

3台服务器都是 # Replication role:master

设置从服务器方式

1.命令方式

# Replication
role:master
connected_slaves:2
slave0:ip=192.168.1.1,port=2000,state=online,offset=113,lag=0
slave1:ip=192.168.1.1,port=3000,state=online,offset=113,lag=0
master_repl_offset:113

# Replication
role:slave
master_host:192.168.1.1
master_port:1000
master_link_status:up

服务器停止，主从就不起作用

2.配置文件

# slaveof  
slaveof 192.168.1.1 1000

服务器停止，主从依然起作用

主从同步，2者密码可以不一致

192.168.1.1:1000> set lyg945 liuyonggang
192.168.1.1:1000> get lyg945
"liuyonggang"

192.168.1.1:2000> get lyg945
"liuyonggang"

192.168.1.1:3000> get lyg945
"liuyonggang"

Redis的主从架构，如果master发现故障了，还得手动将slave切换成master继续服务，手动的方式容易造成失误，导致数据丢失，那Redis有没有一种机制可以在master和slave进行监控，并在master发送故障的时候，能自动将slave切换成master呢？有的，那就是哨兵。

哨兵的作用：
1、监控redis进行状态，包括master和slave
2、当master down机，能自动将slave切换成master

下面配置哨兵监控redis进程，假如我们已经配置好了Master和Slave，具体详细配置参

手动切换master

下面配置哨兵监控redis进程，假如我们已经配置好了Master和Slave，具体详细配置参

手动切换master

  master  SLAVEOF NO ONE   

  slave  SLAVEOF 192.168.1.1 3000

创建哨兵

touch sentinel.conf 内容如下

sentinel monitor 主机名主机ip 主机端口票数n 票数多余n的从机作为主机

sentinel monitor mymaster 192.168.1.1 1000 1

启动哨兵

redis-sentinel sentinel.conf

20526:X 20 Nov 13:24:29.243 # WARNING: The TCP backlog setting of 511 cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of 128.
20526:X 20 Nov 13:24:29.303 # Sentinel ID is 5d351f7edc80148f60036a6c0c2e74510ece4221
20526:X 20 Nov 13:24:29.303 # +monitor master mymaster 192.168.1.1 1000 quorum 1
20526:X 20 Nov 13:24:29.304 * +slave slave 192.168.1.1:2000 192.168.1.1 2000 @ mymaster 192.168.1.1 1000
20526:X 20 Nov 13:24:29.317 * +slave slave 192.168.1.1:3000 192.168.1.1 3000 @ mymaster 192.168.1.1 1000

1.将3000 kill掉

redis-cli -h 192.168.1.1 -p 3000
Could not connect to Redis at 192.168.1.1:3000: Connection refused
Could not connect to Redis at 192.168.1.1:3000: Connection refused

192.168.1.1:1000> set zhangsan 1
192.168.1.1:1000> get zhangsan

"1"

192.168.1.1:2000> get zhangsan

"1"

3000重启

192.168.1.1:3000> get zhangsan

"1"

2.将master(1000) kill掉

redis-cli -h 192.168.1.1 -p 1000
Could not connect to Redis at 192.168.1.1:1000: Connection refused

查看哨兵后台打印信息

20526:X 20 Nov 13:27:40.550 # +sdown master mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:40.550 # +odown master mymaster 192.168.1.1 1000 #quorum 1/1
20526:X 20 Nov 13:27:40.550 # +new-epoch 1
20526:X 20 Nov 13:27:40.550 # +try-failover master mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:40.559 # +vote-for-leader 5d351f7edc80148f60036a6c0c2e74510ece4221 1
20526:X 20 Nov 13:27:40.559 # +elected-leader master mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:40.559 # +failover-state-select-slave master mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:40.612 # +selected-slave slave 192.168.1.1:2000 192.168.1.1 2000 @ mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:40.612 * +failover-state-send-slaveof-noone slave 192.168.1.1:2000 192.168.1.1 2000 @ mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:40.688 * +failover-state-wait-promotion slave 192.168.1.1:2000 192.168.1.1 2000 @ mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:40.766 # +promoted-slave slave 192.168.1.1:2000 192.168.1.1 2000 @ mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:40.766 # +failover-state-reconf-slaves master mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:40.818 * +slave-reconf-sent slave 192.168.1.1:3000 192.168.1.1 3000 @ mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:41.813 * +slave-reconf-inprog slave 192.168.1.1:3000 192.168.1.1 3000 @ mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:41.813 * +slave-reconf-done slave 192.168.1.1:3000 192.168.1.1 3000 @ mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:41.896 # +failover-end master mymaster 192.168.1.1 1000
20526:X 20 Nov 13:27:41.896 # +switch-master mymaster 192.168.1.1 1000 192.168.1.1 2000
20526:X 20 Nov 13:27:41.896 * +slave slave 192.168.1.1:3000 192.168.1.1 3000 @ mymaster 192.168.1.1 2000
20526:X 20 Nov 13:27:41.896 * +slave slave 192.168.1.1:1000 192.168.1.1 1000 @ mymaster 192.168.1.1 2000
20526:X 20 Nov 13:28:11.907 # +sdown slave 192.168.1.1:1000 192.168.1.1 1000 @ mymaster 192.168.1.1 2000

192.168.1.1:2000> info 变成了主

# Replication
role:master
connected_slaves:1
slave0:ip=192.168.1.1,port=3000,state=online,offset=1625,lag=0

192.168.1.1:2000> set aa 11
OK
192.168.1.1:2000> get aa
"11"

192.168.1.1:3000> info 还是从

# Replication
role:slave
master_host:192.168.1.1
master_port:2000
master_link_status:up

192.168.1.1:3000> get aa
"11"

重启1000原来的master服务器

查看哨兵

20526:X 20 Nov 13:42:51.554 # -sdown slave 192.168.1.1:1000 192.168.1.1 1000 @ mymaster 192.168.1.1 2000
20526:X 20 Nov 13:43:01.604 * +convert-to-slave slave 192.168.1.1:1000 192.168.1.1 1000 @ mymaster 192.168.1.1 2000

192.168.1.1:1000> info 变成从

# Replication
role:slave
master_host:192.168.1.1
master_port:2000
master_link_status:up

192.168.1.1:1000> get aa
"11"

点击复制链接与好友分享!回本站首页