Redis集群节点的选举:
当master挂掉之后，就会在该集群中的slave中选取一个来代替mater角色，
从而保证redis集群slot的完整性。
如果其中一个mster和它的slave都挂掉后，会导致slot不完整，整个集群都会挂掉。
集群节点信息：
192.168.2.200:6379> cluster nodes
3ff3a74f9dc41f8bc635ab845ad76bf77ffb0f69 192.168.2.201:6379 master - 0 1527145806504 5 connected 10923-16383
2faf68564a70372cfc06c1afff197019cc6a39f3 192.168.2.201:6380 slave 3ff3a74f9dc41f8bc635ab845ad76bf77ffb0f69 0 1527145805495 5 connected
227f51028bbe827f27b4e40ed7a08fcc7d8df969 192.168.2.200:6380 slave 098e7eb756b6047fde988ab3c0b7189e1724ecf5 0 1527145804485 4 connected
5844b4272c39456b0fdf73e384ff8c479547de47 192.168.2.200:6379 myself,master - 0 0 3 connected 5461-10922
7119dec91b086ca8fe69f7878fa42b1accd75f0f 192.168.2.100:6380 slave 5844b4272c39456b0fdf73e384ff8c479547de47 0 1527145802468 3 connected
098e7eb756b6047fde988ab3c0b7189e1724ecf5 192.168.2.100:6379 master - 0 1527145803476 1 connected 0-5460
新建3个key：
192.168.2.200:6379> set name zhangsan
OK
192.168.2.200:6379> set age 26
-> Redirected to slot [741] located at 192.168.2.100:6379
OK
192.168.2.100:6379> set home beijing
-> Redirected to slot [10814] located at 192.168.2.200:6379
OK
模拟192.168.2.100:6379挂掉：
# ps -ef |grep redis
root 19023 1 0 15:05 ? 00:00:01 redis-server 192.168.2.100:6379 [cluster]
root 19030 1 0 15:05 ? 00:00:01 redis-server 192.168.2.100:6380 [cluster]
root 19127 2912 0 15:13 pts/0 00:00:00 grep --color=auto redis
# kill 19023
查看集群节点状态：（可以看到集群中的一个slave的角色变成了master）
# redis-trib.rb check 192.168.2.200:6379
>>> Performing Cluster Check (using node 192.168.2.200:6379)
M: 5844b4272c39456b0fdf73e384ff8c479547de47 192.168.2.200:6379
slots:5461-10922 (5462 slots) master
1 additional replica(s)
M: 3ff3a74f9dc41f8bc635ab845ad76bf77ffb0f69 192.168.2.201:6379
slots:10923-16383 (5461 slots) master
1 additional replica(s)
S: 2faf68564a70372cfc06c1afff197019cc6a39f3 192.168.2.201:6380
slots: (0 slots) slave
replicates 3ff3a74f9dc41f8bc635ab845ad76bf77ffb0f69
M: 227f51028bbe827f27b4e40ed7a08fcc7d8df969 192.168.2.200:6380
slots:0-5460 (5461 slots) master
0 additional replica(s)
S: 7119dec91b086ca8fe69f7878fa42b1accd75f0f 192.168.2.100:6380
slots: (0 slots) slave
replicates 5844b4272c39456b0fdf73e384ff8c479547de47
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.
查看数据是否存在：
# redis-cli -h 192.168.2.200 -p 6380 -c
192.168.2.200:6380> keys *
1) "age"
模拟恢复故障节点：
# redis-server redis.conf
查看集群节点状态：（故障节点恢复后，是变成master还是slave？）
# redis-trib.rb check 192.168.2.200:6379
>>> Performing Cluster Check (using node 192.168.2.200:6379)
M: 5844b4272c39456b0fdf73e384ff8c479547de47 192.168.2.200:6379
slots:5461-10922 (5462 slots) master
1 additional replica(s)
M: 3ff3a74f9dc41f8bc635ab845ad76bf77ffb0f69 192.168.2.201:6379
slots:10923-16383 (5461 slots) master
1 additional replica(s)
S: 2faf68564a70372cfc06c1afff197019cc6a39f3 192.168.2.201:6380
slots: (0 slots) slave
replicates 3ff3a74f9dc41f8bc635ab845ad76bf77ffb0f69
M: 227f51028bbe827f27b4e40ed7a08fcc7d8df969 192.168.2.200:6380
slots:0-5460 (5461 slots) master
1 additional replica(s)
S: 7119dec91b086ca8fe69f7878fa42b1accd75f0f 192.168.2.100:6380
slots: (0 slots) slave
replicates 5844b4272c39456b0fdf73e384ff8c479547de47
S: 098e7eb756b6047fde988ab3c0b7189e1724ecf5 192.168.2.100:6379
slots: (0 slots) slave
replicates 227f51028bbe827f27b4e40ed7a08fcc7d8df969
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.
总结：master：192.168.2.100:6379 宕机后，
slave：192.168.2.200:6380 代替了其位置，变成了master角色。
当192.168.2.100:6379故障恢复后，并没有恢复到master的角色，
而是充当了slave角色。
原文地址：http://blog.51cto.com/13690439/2119973