配置哨兵监控Redis运行情况

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，具体详细配置参考 http://www.cnblogs.com/liuyansheng/p/6530851.html

1、新建配置文件，命名为sentinel.conf，输入下面内容

sentinel monitor mymaster 192.168.137.101 6379 1
sentinel auth-pass myMaster master123 #配置密码

mymaster ：master服务的名称，随便定义 
192.168.137.101 6379：master的ip和端口 
1：在至少1个哨兵实例都认为master down后把master标记为odown（objective down客观down；相对应的存在sdown，subjective down，主观down）状态

sentinel.conf具体配置参考 http://www.cnblogs.com/liuyansheng/p/6531321.html

2、启动哨兵，在redis的src目录下，执行下面命令

 ./redis-sentinel ../../conf/sentinel.conf 

可以看到哨兵已经把master和slave都监控了。

3、进行哨兵测试，将103的slave进行shutdown，哨兵将监控到状态。

6594:X 14 Jun 01:01:26.647 # +sdown slave 192.168.137.102:6379 192.168.137.103 6379 @ myMaster 192.168.137.101 6379

然后将103重启服务，查看哨兵控制台

6594:X 14 Jun 20:13:22.716 * +reboot slave 192.168.137.102:6379 192.168.137.103 6379 @ myMaster 192.168.137.101 6379

已经进行恢复

4、将master kill掉，查看哨兵后台打印信息

6594:X 14 Jun 20:13:50.300 # +sdown master myMaster  192.168.137.101 6379  #说明master服务已经宕机
6594:X 14 Jun 20:13:50.300 # +odown master myMaster  192.168.137.101 6379 #quorum 1/1  
6594:X 14 Jun 20:13:50.300 # +new-epoch 1
6594:X 14 Jun 20:13:50.300 # +try-failover master myMaster 192.168.137.101 6379  #开始恢复
6594:X 14 Jun 20:13:50.304 # +vote-for-leader a496627d72d98fde98b8270e297ab32b330ebac7 1  #投票选举哨兵leader，现在就一个哨兵所以leader就自己
6594:X 14 Jun 20:13:50.304 # +elected-leader master myMaster 192.168.137.101 6379 # 选中leader
6594:X 14 Jun 20:13:50.304 # +failover-state-select-slave master myMaster 192.168.137.101 6379 #选择master
6594:X 14 Jun 20:13:50.357 # +selected-slave slave 192.168.137.102:6379 192.168.137.102 6379 @ myMaster 192.168.137.101 6379  #选中192.168.137.102 6379作为切换
6594:X 14 Jun 20:13:50.357 * +failover-state-send-slaveof-noone slave 192.168.137.102:6379 192.168.137.102 6379 @ myMaster 192.168.137.101 6379  #发送slaveof no one命令
6594:X 14 Jun 20:13:50.420 * +failover-state-wait-promotion slave 192.168.137.102:6379 192.168.137.102 6379 @ myMaster 192.168.137.101 6379   #等待升级master
6594:X 14 Jun 20:13:50.515 # +promoted-slave slave 192.168.137.102:6379 192.168.137.102 6379 @ myMaster 192.168.137.101 6379  #192.168.137.102 6379升级为master
6594:X 14 Jun 20:13:50.515 # +failover-state-reconf-slaves master myMaster 192.168.137.101 6379
6594:X 14 Jun 20:13:50.566 * +slave-reconf-sent slave 192.168.137.103:6379 192.168.137.103 6379 @ myMaster 192.168.137.101 6379
6594:X 14 Jun 20:13:51.333 * +slave-reconf-inprog slave 192.168.137.103:6379 192.168.137.103 6379 @ myMaster 192.168.137.101 6379
6594:X 14 Jun 20:13:52.382 * +slave-reconf-done slave 192.168.137.103:6379 192.168.137.103 6379 @ myMaster 192.168.137.101 6379
6594:X 14 Jun 20:13:52.438 # +failover-end master myMaster 192.168.137.101 6379 #故障恢复完成
6594:X 14 Jun 20:13:52.438 # +switch-master myMaster 192.168.137.101 6379 192.168.137.102 6379  #master切换
6594:X 14 Jun 20:13:52.438 * +slave slave 192.168.137.103:6379 192.168.137.103 6379 @ myMaster 192.168.137.102 6379  #添加master从库
6594:X 14 Jun 20:13:52.438 * +slave slave 192.168.137.101:6379 192.168.137.101 6379 @ myMaster 192.168.137.102 6379  #添加master从库
6594:X 14 Jun 20:17:22.463 # +sdown slave 192.168.137.101:6379 192.168.137.101 6379 @ myMaster 192.168.137.102 6379  #发现192.168.137.101 6379故障

5、配置多个哨兵，修改配置文件

sentinel monitor myMaster1 192.168.137.101 6379 2
sentinel monitor myMaster2 192.168.137.101 6379 2
sentinel monitor myMaster3 192.168.137.101 6379 2

sentinel auth-pass myMaster1 master123
sentinel auth-pass myMaster2 master123
sentinel auth-pass myMaster3 master123

这里是配置3个哨兵，架构如下：