基于docker实现哨兵集群部署

简单dockerfile文件，用于演示sentinel哨兵故障转移
FROM centos:latest
MAINTAINER BIXIAOYU
RUN groupadd -r redis &&useradd -r -g redis redis
RUN yum -y update &&  yum -y install epel-release && yum -y install redis && yum -y install net-tools
EXPOSE 6379

# docker build --no-cache -t redis .

启动docker容器实例

docker run -itd --name redis-master --net=mynetwork -p 6383:6379 --ip 172.60.0.3 redis

docker run -itd --name redis-slave1 --net=mynetwork -p 6384:6379 --ip 172.60.0.4 redis

docker run -itd --name redis-slave2 --net=mynetwork -p 6385:6379 --ip 172.60.0.5 redis

docker run -itd --name redis-sentinel1 --net=mynetwork -p : --ip 172.60.0.5 redis

docker run -itd --name redis-sentinel2 --net=mynetwork -p : --ip 172.60.0.6 redis

docker run -itd --name redis-sentinel3 --net=mynetwork -p : --ip 172.60.0.7 redis

 #docker network create --subnet=172.60.0.0/16 mynetwork  

【配置】

先完成主从同步，主从同步的配置可参考https://www.cnblogs.com/bixiaoyu/p/10706811.html这篇文章，在此就不必细说了，重点是sentinel的配置

bind 0.0.0.0            #设置允许访问的IP，这里仍然设置0.0.0.0
protected-mode no         #允许在连接时，密码为空
port
dir "/tmp"
sentinel myid 6d0d4099c13cdeab018e1f2005455be6f1cd6f6b
sentinel monitor mymaster 172.60.0.3  2        #设置监听
sentinel config-epoch mymaster
sentinel leader-epoch mymaster
logfile "/var/log/redis/sentinel.log"
sentinel known-slave mymaster 172.60.0.2
sentinel known-slave mymaster 172.60.0.4
sentinel known-sentinel mymaster 172.60.0.7  dfce433e021aa3e82276974aa12fa0684fb0b4f0
sentinel known-sentinel mymaster 172.60.0.6  ceb363cf84103950cfa2a785816c4e8a36c02143
sentinel current-epoch 

 选择一个sentinel节点，查看日志信息，发现默认主节点为172.60.0.2

【测试故障转移】

先通过pkill模拟故障master故障宕机，随后观察哨兵任意节点的情况

我们可以从下面sentinel节点上看到172.0.0.3被选举成为新主节点

[root@ef77b61448fc /]# tail /var/log/redis/sentinel.log
:X  Apr ::26.472 * +sentinel sentinel ceb363cf84103950cfa2a785816c4e8a36c02143 172.60.0.6  @ mymaster 172.60.0.2
:X  Apr ::24.154 * +sentinel sentinel dfce433e021aa3e82276974aa12fa0684fb0b4f0 172.60.0.7  @ mymaster 172.60.0.2
:X  Apr ::05.724 # +sdown master mymaster 172.60.0.2
:X  Apr ::05.895 # +new-epoch
:X  Apr ::05.895 # +vote-for-leader dfce433e021aa3e82276974aa12fa0684fb0b4f0
:X  Apr ::06.276 # +config-update-from sentinel dfce433e021aa3e82276974aa12fa0684fb0b4f0 172.60.0.7  @ mymaster 172.60.0.2
:X  Apr ::06.276 # +switch-master mymaster 172.60.0.2  172.60.0.3 
:X  Apr ::06.276 * +slave slave 172.60.0.4: 172.60.0.4  @ mymaster 172.60.0.3
:X  Apr ::06.276 * +slave slave 172.60.0.2: 172.60.0.2  @ mymaster 172.60.0.3 

查看sentinel节点的日志信息

此时，我们把pkill掉的172.60.0.2(旧主节点)重启，在看看新任Master主节点172.60.0.3有何变化

重启的旧主节点成为从节点并加入进来，此时正在进行复制动作，offset的偏移量还没有同步一致，

【小结】

进入哨兵节点客户端执行SENTINEL masterts命令显示被监控的所有master以及状态

127.0.0.1:> SENTINEL masters
)  ) "name"
    ) "mymaster"        #被监控主节点的名称
    ) "ip"
    ) "172.60.0.3"      #被监控主节点的IP
    ) "port"
    ) ""
    ) "runid"
    ) "dd3696a2793e4e19892fca48793d75cec3f07bea"   #被监控主节点的runid值
    ) "flags"
   ) "master"
   ) "link-pending-commands"
   ) ""
   ) "link-refcount"
   ) ""
   ) "last-ping-sent"
   ) ""
   ) "last-ok-ping-reply"
   ) ""
   ) "last-ping-reply"
   ) ""
   ) "down-after-milliseconds"
   ) ""                 #监控节点不可达超时时间
   ) "info-refresh"
   ) ""
   ) "role-reported"
   ) "master"
   ) "role-reported-time"
   ) ""
   ) "config-epoch"
   ) ""
   ) "num-slaves"                   #检测剩余slave节点个数
   ) ""
   ) "num-other-sentinels"          #检测其它sentinel节点个数
   ) ""
   ) "quorum"                 #允许主节点不可用的sentinels的数量（最多允许两个sentinel节点故障）
   ) ""
   ) "failover-timeout"     #延迟时间
   ) ""
   ) "parallel-syncs"          #复制转移数量
   ) ""

 执行SENTINEL slaves mymastert查看从节点信息，此时你会看到两个从节点

)  ) "name"
    ) "172.60.0.2:6379"
    ) "ip"
    ) "172.60.0.2"
    ) "port"
    ) ""
    ) "runid"
    ) "3ede55439a3ce6fb1ab171ed7fd6b6c639725966"
    ) "flags"
   ) "slave"
   ) "link-pending-commands"
   ) ""
   ) "link-refcount"
   ) ""
   ) "last-ping-sent"
   ) ""
   ) "last-ok-ping-reply"
   ) ""
   ) "last-ping-reply"
   ) ""
   ) "down-after-milliseconds"
   ) ""
   ) "info-refresh"
   ) ""
   ) "role-reported"
   ) "slave"
   ) "role-reported-time"
   ) ""
   ) "master-link-down-time"
   ) ""
   ) "master-link-status"
   ) "ok"
   ) "master-host"
   ) "172.60.0.3"
   ) "master-port"
   ) ""
   ) "slave-priority"
   ) ""
   ) "slave-repl-offset"
   ) ""
)  ) "name"
    ) "172.60.0.4:6379"
    ) "ip"
    ) "172.60.0.4"
    ) "port"
    ) ""
    ) "runid"
    ) "f17cfcfc4b9217e1f5a3c0d0a2c55d82da46c37e"
    ) "flags"
   ) "slave"
   ) "link-pending-commands"
   ) ""
   ) "link-refcount"
   ) ""
   ) "last-ping-sent"
   ) ""
   ) "last-ok-ping-reply"
   ) ""
   ) "last-ping-reply"
   ) ""
   ) "down-after-milliseconds"
   ) ""
   ) "info-refresh"
   ) ""
   ) "role-reported"
   ) "slave"
   ) "role-reported-time"
   ) ""
   ) "master-link-down-time"
   ) ""
   ) "master-link-status"
   ) "ok"
   ) "master-host"
   ) "172.60.0.3"
   ) "master-port"
   ) ""
   ) "slave-priority"
   ) ""
   ) "slave-repl-offset"
   ) ""

 查看主节点的端口

127.0.0.1:> sentinel get-master-addr-by-name mymaster
) "172.60.0.3"
) ""

执行sentinel failover mymaster命令，强制切换主节点！下图所示，发现主节点已经由172.60.0.3变为172.60.0.2了，

待更。。。。