之前研究过基于主从+分片实现的集群，现在研究下基于哨兵模式的集群模式。

1. 下载redis

版本：   3.2.100

2. 开始搭建集群

(1) 复制集三个redis 目录，分别命名为 redis-63791、 redis-63792、 redis-63793    63791是master节点，剩下两个是slave 节点

(2) redis-63791 目录下修改：

1》redis.windows.conf 大概79 行左右： 修改端口号

port 63791

2》 新建startup.bat

title master_63791
redis-server.exe redis.windows.conf

3》 新建sentinel.conf

port 36379
sentinel monitor mymaster 127.0.0.1 63791 1
sentinel down-after-milliseconds mymaster 5000
sentinel config-epoch mymaster 10

4》新建startup-sentinel.bat

title sentinel_63791
redis-server.exe sentinel.conf --sentinel

(2) redis-63792 目录下修改：

1》redis.windows.conf 大概79 行左右：    修改端口和声明是从服务器

port 63792
slaveof 127.0.0.1 63791

2》 新建startup.bat

title slave_63792
redis-server.exe redis.windows.conf

3》 新建sentinel.conf

port 36380
sentinel monitor mymaster 127.0.0.1 63791 1
sentinel down-after-milliseconds mymaster 5000
sentinel config-epoch mymaster 10

4》新建startup-sentinel.bat

title sentinel_63792
redis-server.exe sentinel.conf --sentinel

(3) redis-63793 目录下修改：

1》redis.windows.conf 大概79 行左右：

port 63793
slaveof 127.0.0.1 63791

2》 新建startup.bat

title slave_63793
redis-server.exe redis.windows.conf

3》 新建sentinel.conf

port 36381
sentinel monitor mymaster 127.0.0.1 63791 1
sentinel down-after-milliseconds mymaster 5000
sentinel config-epoch mymaster 10

4》新建startup-sentinel.bat

title sentinel_63793
redis-server.exe sentinel.conf --sentinel

title 是windows 批处理命令，用于设置cmd窗口的title。

3. 启动服务

1. 先启动三个redis 服务startup.bat

依次启动63791/63792/63793， 查看63791 服务的打印的日志如下：

[8928] 24 Apr 15:16:54.001 # Server started, Redis version 3.2.100
[8928] 24 Apr 15:16:54.002 * DB loaded from disk: 0.001 seconds
[8928] 24 Apr 15:16:54.002 * The server is now ready to accept connections on port 63791
[8928] 24 Apr 15:17:12.189 * Slave 127.0.0.1:63792 asks for synchronization
[8928] 24 Apr 15:17:12.189 * Full resync requested by slave 127.0.0.1:63792
[8928] 24 Apr 15:17:12.190 * Starting BGSAVE for SYNC with target: disk
[8928] 24 Apr 15:17:12.259 * Background saving started by pid 9400
[8928] 24 Apr 15:17:12.661 # fork operation complete
[8928] 24 Apr 15:17:12.664 * Background saving terminated with success
[8928] 24 Apr 15:17:12.674 * Synchronization with slave 127.0.0.1:63792 succeeded
[8928] 24 Apr 15:17:26.079 * Slave 127.0.0.1:63793 asks for synchronization
[8928] 24 Apr 15:17:26.080 * Full resync requested by slave 127.0.0.1:63793
[8928] 24 Apr 15:17:26.080 * Starting BGSAVE for SYNC with target: disk
[8928] 24 Apr 15:17:26.158 * Background saving started by pid 8224
[8928] 24 Apr 15:17:26.759 # fork operation complete
[8928] 24 Apr 15:17:26.761 * Background saving terminated with success
[8928] 24 Apr 15:17:26.767 * Synchronization with slave 127.0.0.1:63793 succeeded

63792 打印的日志如下：

[7556] 24 Apr 15:17:12.180 # Server started, Redis version 3.2.100
[7556] 24 Apr 15:17:12.181 * DB loaded from disk: 0.001 seconds
[7556] 24 Apr 15:17:12.182 * The server is now ready to accept connections on port 63792
[7556] 24 Apr 15:17:12.182 * Connecting to MASTER 127.0.0.1:63791
[7556] 24 Apr 15:17:12.184 * MASTER <-> SLAVE sync started
[7556] 24 Apr 15:17:12.185 * Non blocking connect for SYNC fired the event.
[7556] 24 Apr 15:17:12.186 * Master replied to PING, replication can continue...
[7556] 24 Apr 15:17:12.188 * Partial resynchronization not possible (no cached master)
[7556] 24 Apr 15:17:12.260 * Full resync from master: f797667812e495c1e89823ee97f454db6a97b51e:1
[7556] 24 Apr 15:17:12.671 * MASTER <-> SLAVE sync: receiving 380 bytes from master
[7556] 24 Apr 15:17:12.834 * MASTER <-> SLAVE sync: Flushing old data
[7556] 24 Apr 15:17:12.834 * MASTER <-> SLAVE sync: Loading DB in memory
[7556] 24 Apr 15:17:12.835 * MASTER <-> SLAVE sync: Finished with success

可以看到主从复制是通过sync 命令进行同步数据的

2. 接下来启动三个哨兵进程startup-sentinel.bat

3. 客户端连接到三个服务查看集群信息

(1) 63791查看

E:\redis-cluster\redis-63791>redis-cli -p 63791
127.0.0.1:63791> info replication
# Replication
role:master
connected_slaves:2
slave0:ip=127.0.0.1,port=63792,state=online,offset=26578,lag=0
slave1:ip=127.0.0.1,port=63793,state=online,offset=26728,lag=0
master_repl_offset:26728
repl_backlog_active:1
repl_backlog_size:1048576
repl_backlog_first_byte_offset:2
repl_backlog_histlen:26727

　　可以看到是master 节点，并且可以看到两个从节点。

(2) 查看63792 的端口

E:\redis-cluster\redis-63791>redis-cli -p 63792
127.0.0.1:63792> info replication
# Replication
role:slave
master_host:127.0.0.1
master_port:63791
master_link_status:up
master_last_io_seconds_ago:0
master_sync_in_progress:0
slave_repl_offset:36862
slave_priority:100
slave_read_only:1
connected_slaves:0
master_repl_offset:0
repl_backlog_active:0
repl_backlog_size:1048576
repl_backlog_first_byte_offset:0
repl_backlog_histlen:0

　　可以看到是从节点，并且 slave_read_only 只读属性为true， 也就是从节点只能读取数据。

(3) 查看63793集群信息

E:\redis-cluster\redis-63791>redis-cli -p 63793
127.0.0.1:63793> info replication
# Replication
role:slave
master_host:127.0.0.1
master_port:63791
master_link_status:up
master_last_io_seconds_ago:0
master_sync_in_progress:0
slave_repl_offset:53022
slave_priority:100
slave_read_only:1
connected_slaves:0
master_repl_offset:0
repl_backlog_active:0
repl_backlog_size:1048576
repl_backlog_first_byte_offset:0
repl_backlog_histlen:0

3. 模拟63791 主节点挂掉：

将63791 哨兵以及redis-server 都停掉，再次查看从63792/63793 查看集群信息

(1)从哨兵控制台可以看到产生了新的master 节点，日志如下：

[11188] 24 Apr 15:29:15.388 # +sdown master mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:15.388 # +odown master mymaster 127.0.0.1 63791 #quorum 1/1
[11188] 24 Apr 15:29:15.388 # +new-epoch 11
[11188] 24 Apr 15:29:15.388 # +try-failover master mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:15.393 # +vote-for-leader 956fe7dd1644c38db24792da8af749847bd3c247 11
[11188] 24 Apr 15:29:15.401 # 10765c1132658cf10050164915ea69a5b77da4eb voted for 956fe7dd1644c38db24792da8af749847bd3c24
7 11
[11188] 24 Apr 15:29:15.493 # +elected-leader master mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:15.493 # +failover-state-select-slave master mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:15.555 # +selected-slave slave 127.0.0.1:63792 127.0.0.1 63792 @ mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:15.555 * +failover-state-send-slaveof-noone slave 127.0.0.1:63792 127.0.0.1 63792 @ mymaster 127.0.
0.1 63791
[11188] 24 Apr 15:29:15.623 * +failover-state-wait-promotion slave 127.0.0.1:63792 127.0.0.1 63792 @ mymaster 127.0.0.1
63791
[11188] 24 Apr 15:29:16.466 # +promoted-slave slave 127.0.0.1:63792 127.0.0.1 63792 @ mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:16.467 # +failover-state-reconf-slaves master mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:16.524 * +slave-reconf-sent slave 127.0.0.1:63793 127.0.0.1 63793 @ mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:17.482 * +slave-reconf-inprog slave 127.0.0.1:63793 127.0.0.1 63793 @ mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:18.513 * +slave-reconf-done slave 127.0.0.1:63793 127.0.0.1 63793 @ mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:18.583 # +failover-end master mymaster 127.0.0.1 63791
[11188] 24 Apr 15:29:18.583 # +switch-master mymaster 127.0.0.1 63791 127.0.0.1 63792
[11188] 24 Apr 15:29:18.585 * +slave slave 127.0.0.1:63793 127.0.0.1 63793 @ mymaster 127.0.0.1 63792
[11188] 24 Apr 15:29:18.585 * +slave slave 127.0.0.1:63791 127.0.0.1 63791 @ mymaster 127.0.0.1 63792
[11188] 24 Apr 15:29:19.291 # +sdown sentinel 71a27c573f9eb89554a1db4dae04c406e19fd85a 127.0.0.1 36379 @ mymaster 127.0.
0.1 63792
[11188] 24 Apr 15:29:23.665 # +sdown slave 127.0.0.1:63791 127.0.0.1 63791 @ mymaster 127.0.0.1 63792

(2) 从63792 查看集群信息

127.0.0.1:63792> info replication
# Replication
role:master
connected_slaves:1
slave0:ip=127.0.0.1,port=63793,state=online,offset=12526,lag=0
master_repl_offset:12526
repl_backlog_active:1
repl_backlog_size:1048576
repl_backlog_first_byte_offset:2
repl_backlog_histlen:12525

　　可以看到升级为主节点。并且63793 为从节点。

4. 63791 再次上线

　　启动服务节点与哨兵节点。从63792 查看集群信息

127.0.0.1:63792> info replication
# Replication
role:master
connected_slaves:2
slave0:ip=127.0.0.1,port=63793,state=online,offset=25459,lag=1
slave1:ip=127.0.0.1,port=63791,state=online,offset=25459,lag=0
master_repl_offset:25595
repl_backlog_active:1
repl_backlog_size:1048576
repl_backlog_first_byte_offset:2
repl_backlog_histlen:25594

　　可以看到63791 上线后成为slave 节点。然后查看63791的配置文件：redis.windows.conf 和 sentinel.conf

(1) redis.windows.conf 

slaveof 127.0.0.1 63792

　　可以看到成为从节点

(2) 查看哨兵配置信息： 监控的主节点也变为63792

port 36379
sentinel myid 71a27c573f9eb89554a1db4dae04c406e19fd85a
sentinel monitor mymaster 127.0.0.1 63792 1
sentinel down-after-milliseconds mymaster 5000
# Generated by CONFIG REWRITE
dir "E:\\redis-cluster\\redis-63791"
sentinel config-epoch mymaster 11
sentinel leader-epoch mymaster 0
sentinel known-slave mymaster 127.0.0.1 63791
sentinel known-slave mymaster 127.0.0.1 63793
sentinel known-sentinel mymaster 127.0.0.1 36381 956fe7dd1644c38db24792da8af749847bd3c247
sentinel known-sentinel mymaster 127.0.0.1 36380 10765c1132658cf10050164915ea69a5b77da4eb
sentinel current-epoch 11

4. 主节点写入数据

　　注意主节点可以写入数据，从节点只能读取数据；主节点接到写入命令后会将命令同步到其他节点。

　　当然从节点可以用如下属性进行设置：  只是从节点不会同步至主节点，所以从节点写入数据没必要。

slave-read-only no

1.  63792 主节点进行操作

127.0.0.1:63792> set str value
OK
127.0.0.1:63792> keys *
1) "str"
127.0.0.1:63792> get str
"value"
127.0.0.1:63792> ttl str
(integer) -1

2. 63793 从节点查看

127.0.0.1:63793> keys *
1) "str"
127.0.0.1:63793> get str
"value"
127.0.0.1:63793> set kk kk
(error) READONLY You can't write against a read only slave.
127.0.0.1:63793> flushall
(error) READONLY You can't write against a read only slave.

　　可以看到63793 从节点只能读取数据，不能进行数据的修改操作。

总结：

1. redis.windows.conf 配置文件中主节点只需要修改端口即可，从节点需要声明是哪个节点的从节点。

2. sentinel.conf 需要声明三个不一样的端口。因为哨兵实际上是启动一个进程进行检测，所以需要声明为是三个不同的端口。

3. 哨兵声明检测的是master 节点。

4. 当主节点挂掉后会自动从从节点选取一个主节点，然后对应的redis.windows.conf 会改变相应的slave 信息； 哨兵配置文件 sentinel.conf 也会改变为对应的文件。

5. 如果不启动哨兵进程， 主节点挂掉之后将没有节点可写入数据，所以需要启动哨兵进程，主节点挂掉之后随机选择一个节点作为主节点， 此节点可以写入数据。

6. 启动各节点之后可以用redis-cli -p 63791 连接到之后用命令 info replication 查看节点的集群信息。

7. redis 主从节点同步是主节点收到命令会同步发送给客户端，并且在启动的时候客户端通过sync 命令同步主节点的数据信息。