关于缓存Redis主从配置

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 <masterip> <masterport>
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"