Redis从3.0版本开始支持原生的集群模式，即 Redis Cluster。Redis Cluster，主要是针对海量数据下的高并发、高可用场景，海量数据就是说单机Redis无法完全容纳数据，需要进行数据分片。
本章，我就来讲解如何搭建一个3主3从的Redis Cluster。关于Redis Cluster的基本原理，读者可以参考进阶篇中的《分布式框架之高性能：Redis集群模式》。
一、集群部署
Redis Cluster至少要求有3个Master节点，同时为了保证高可用，每个Master都建议至少有一个Slave，同时Master不能跟自己的Slave部署在同一台机器上。所以，我会在ressmix-dsf01、ressmix-dsf02、ressmix-dsf03这三台机器上分别部署一个Master一个Slave。首先，我将ressmix-dsf01、ressmix-dsf02、ressmix-dsf03上的哨兵和Redis进程都停止掉。
1.1 集群配置
我们先分配下端口，6个Redis节点，端口依次为7001、7002、7003、7004、7005、7006。ressmix-dsf01上分配7001、7002，ressmix-dsf02上分配7003、7004，ressmix-dsf03上分配7005、7006。
我以ressmix-dsf01为例，说明下需要建立的目录：
#存放redis的配置文件
mkdir -p /etc/redis
#cluster模式下，Redis将集群状态保存在那里，包括集群机器信息等
mkdir -p /etc/redis-cluster
#存放redis的日志
mkdir -p /var/log/redis
#运行时工作目录，注意端口
mkdir -p /var/redis/7001
mkdir -p /var/redis/7002
然后，将redis安装包目录下redis.conf配置文件拷贝到各台机器的/etc/redis/目录下，分别重名为7001.conf、7002.conf……7006.conf。我以7001.conf为例，说下Redis Cluster模式要修改的配置：
#运行端口
port 7001
#启用集群模式
cluster-enabled yes
#集群状态信息文件位置
cluster-config-file /etc/redis-cluster/nodes7001.conf
#节点超时时间(秒)，超过该时间则认为节点宕机，master宕机会触发主备切换，slave宕机就不会提供服务
cluster-node-timeout 15000
daemonize yes
pidfile /var/run/redis_7001.pid
dir /var/redis/7001
logfile /var/log/redis/7001.log
#宿主机IP
bind 192.168.0.107
appendonly yes
接着，准备生产环境的启动脚本，将Redis安装包util目录下的redis_init_script脚本拷贝到/etc/init.d目录，分别命名为: redis_7001, redis_7002, redis_7003, redis_7004, redis_7005, redis_7006，每个启动脚本内，都修改REDISPORT为对应的端口号：
#!/bin/sh
#
# Simple Redis init.d script conceived to work on Linux systems
# as it does use of the /proc filesystem.
# chkconfig: 2345 90 10
# description: Redis is a persistent key-value database
最后，执行/etc/init.d/redis_XXXX start分别启动6个节点，我们可以到/var/log/redis/目录中看下启动日志，确认下是否启动成功：
[root@ressmix-dsf01 redis]# cat 7001.log
1087:C 28 Apr 2020 23:03:04.763 # oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo
1087:C 28 Apr 2020 23:03:04.763 # Redis version=5.0.8, bits=32, commit=00000000, modified=0, pid=1087, just started
1087:C 28 Apr 2020 23:03:04.763 # Configuration loaded
1088:M 28 Apr 2020 23:03:04.765 * Increased maximum number of open files to 10032 (it was originally set to 1024).
1088:M 28 Apr 2020 23:03:04.765 # Warning: 32 bit instance detected but no memory limit set. Setting 3 GB maxmemory limit with 'noeviction' policy now.
1088:M 28 Apr 2020 23:03:04.765 * No cluster configuration found, I'm 6ac3762d35e60af9d8d12f3ac148c8ede11b94af
_._
_.-``__ ''-._
_.-`` `. `_. ''-._ Redis 5.0.8 (00000000/0) 32 bit
.-`` .-```.```\/ _.,_ ''-._
( ' , .-` | `, ) Running in cluster mode
|`-._`-...-` __...-.``-._|'` _.-'| Port: 7001
| `-._ `._ / _.-' | PID: 1088
`-._ `-._ `-./ _.-' _.-'
|`-._`-._ `-.__.-' _.-'_.-'|
| `-._`-._ _.-'_.-' | http://redis.io
`-._ `-._`-.__.-'_.-' _.-'
|`-._`-._ `-.__.-' _.-'_.-'|
| `-._`-._ _.-'_.-' |
`-._ `-._`-.__.-'_.-' _.-'
`-._ `-.__.-' _.-'
`-._ _.-'
`-.__.-'
1088:M 28 Apr 2020 23:03:04.776 # WARNING: The TCP backlog setting of 511 cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of 128.
1088:M 28 Apr 2020 23:03:04.776 # Server initialized
1088:M 28 Apr 2020 23:03:04.776 # WARNING overcommit_memory is set to 0! Background save may fail under low memory condition. To fix this issue add 'vm.overcommit_memory = 1' to /etc/sysctl.conf and then reboot or run the command 'sysctl vm.overcommit_memory=1' for this to take effect.
1088:M 28 Apr 2020 23:03:04.776 * Ready to accept connections
1.2 集群创建
配置完并启动上述的6个节点后，它们还不在一个集群里面，不能互相发现。我们需要通过命令完成集群的创建：
redis-cli --cluster create 192.168.0.107:7001 192.168.0.107:7002 192.168.0.109:7003 192.168.0.109:7004 192.168.0.110:7005 192.168.0.110:7006 --cluster-replicas 1
其中，--cluster-replicas 1表示每个Master有一个Slave，Redis会保证尽量不将同一个Master和Slave分配在一台机器上。执行完命令后，输出如下：
>>> Performing hash slots allocation on 6 nodes...
Master[0] -> Slots 0 - 5460
Master[1] -> Slots 5461 - 10922
Master[2] -> Slots 10923 - 16383
Adding replica 192.168.0.109:7004 to 192.168.0.107:7001
Adding replica 192.168.0.110:7006 to 192.168.0.109:7003
Adding replica 192.168.0.107:7002 to 192.168.0.110:7005
M: 6ac3762d35e60af9d8d12f3ac148c8ede11b94af 192.168.0.107:7001
slots:[0-5460] (5461 slots) master
S: 359dec6ff52d85c983f49c97997b98d91758f049 192.168.0.107:7002
replicates 94b96a1d009232fc013f9d15535000087c82248a
M: 50d9fb7c86d15847233fc691f6e54e6173fd5d85 192.168.0.109:7003
slots:[5461-10922] (5462 slots) master
S: 477af6a0d870cd61930d654667992eb4abeac920 192.168.0.109:7004
replicates 6ac3762d35e60af9d8d12f3ac148c8ede11b94af
M: 94b96a1d009232fc013f9d15535000087c82248a 192.168.0.110:7005
slots:[10923-16383] (5461 slots) master
S: e8091d7c48647d55a34fa83949600d025d68b879 192.168.0.110:7006
replicates 50d9fb7c86d15847233fc691f6e54e6173fd5d85
Can I set the above configuration? (type 'yes' to accept): yes
>>> Nodes configuration updated
>>> Assign a different config epoch to each node
>>> Sending CLUSTER MEET messages to join the cluster
Waiting for the cluster to join
...
>>> Performing Cluster Check (using node 192.168.0.107:7001)
M: 6ac3762d35e60af9d8d12f3ac148c8ede11b94af 192.168.0.107:7001
slots:[0-5460] (5461 slots) master
13816612910207598593 additional replica(s)
M: 50d9fb7c86d15847233fc691f6e54e6173fd5d85 192.168.0.109:7003
slots:[5461-10922] (5462 slots) master
653934029518667777 additional replica(s)
S: e8091d7c48647d55a34fa83949600d025d68b879 192.168.0.110:7006
slots: (0 slots) slave
replicates 50d9fb7c86d15847233fc691f6e54e6173fd5d85
M: 94b96a1d009232fc013f9d15535000087c82248a 192.168.0.110:7005
slots:[10923-16383] (5461 slots) master
653934854152388609 additional replica(s)
S: 477af6a0d870cd61930d654667992eb4abeac920 192.168.0.109:7004
slots: (0 slots) slave
replicates 6ac3762d35e60af9d8d12f3ac148c8ede11b94af
S: 359dec6ff52d85c983f49c97997b98d91758f049 192.168.0.107:7002
slots: (0 slots) slave
replicates 94b96a1d009232fc013f9d15535000087c82248a
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.
至此，3主3从的Redis Cluster集群搭建完毕。
二、集群操作
2.1 数据读写
我们可以尝试在ressmix-dsf01这台机器上，给Redis的Master节点写入一条数据：
[root@ressmix-dsf01 src]# redis-cli -h 192.168.0.107 -p 7001
192.168.0.107:7001> set k1 v1
(error) MOVED 12706 192.168.0.110:7005
192.168.0.107:7001> set k2 v2
OK
192.168.0.107:7001> KEYS *
1) "k2"
可以看到，对于k1这个键，返回了一个MOVED指令，说明k1对应的槽（slot）不在192.168.0.107:7001这个Master节点上，而是在192.168.0.110:7005上。
我们如果希望Redis自动进行重定向，可以加上一个
-c参数：redis-cli -h 192.168.0.107 -p 7001 -c 。
2.2 高可用
我们可以kill掉一个Master节点，然后通过命令redis-cli --cluster check 192.168.0.107:7001看下集群状态，看看是否会发生故障自动转移和恢复。
kill掉192.168.0.107:7001之前：
[root@ressmix-dsf01 src]# redis-cli --cluster check 192.168.0.107:7001
192.168.0.107:7001 (6ac3762d...) -> 1 keys | 5461 slots | 1 slaves.
192.168.0.109:7003 (50d9fb7c...) -> 0 keys | 5462 slots | 1 slaves.
192.168.0.110:7005 (94b96a1d...) -> 0 keys | 5461 slots | 1 slaves.
[OK] 1 keys in 3 masters.
0.00 keys per slot on average.
>>> Performing Cluster Check (using node 192.168.0.107:7001)
M: 6ac3762d35e60af9d8d12f3ac148c8ede11b94af 192.168.0.107:7001
slots:[0-5460] (5461 slots) master
13832323230561468417 additional replica(s)
M: 50d9fb7c86d15847233fc691f6e54e6173fd5d85 192.168.0.109:7003
slots:[5461-10922] (5462 slots) master
625984512660078593 additional replica(s)
S: e8091d7c48647d55a34fa83949600d025d68b879 192.168.0.110:7006
slots: (0 slots) slave
replicates 50d9fb7c86d15847233fc691f6e54e6173fd5d85
M: 94b96a1d009232fc013f9d15535000087c82248a 192.168.0.110:7005
slots:[10923-16383] (5461 slots) master
625985406013276161 additional replica(s)
S: 477af6a0d870cd61930d654667992eb4abeac920 192.168.0.109:7004
slots: (0 slots) slave
replicates 6ac3762d35e60af9d8d12f3ac148c8ede11b94af
S: 359dec6ff52d85c983f49c97997b98d91758f049 192.168.0.107:7002
slots: (0 slots) slave
replicates 94b96a1d009232fc013f9d15535000087c82248a
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.
然后kill掉192.168.0.107:7001，再check一下集群状态：
[root@ressmix-dsf01 run]# redis-cli --cluster check 192.168.0.107:7002
Could not connect to Redis at 192.168.0.107:7001: Connection refused
192.168.0.109:7004 (477af6a0...) -> 1 keys | 5461 slots | 0 slaves.
192.168.0.109:7003 (50d9fb7c...) -> 0 keys | 5462 slots | 1 slaves.
192.168.0.110:7005 (94b96a1d...) -> 0 keys | 5461 slots | 1 slaves.
[OK] 1 keys in 3 masters.
0.00 keys per slot on average.
>>> Performing Cluster Check (using node 192.168.0.107:7002)
S: 359dec6ff52d85c983f49c97997b98d91758f049 192.168.0.107:7002
slots: (0 slots) slave
replicates 94b96a1d009232fc013f9d15535000087c82248a
S: e8091d7c48647d55a34fa83949600d025d68b879 192.168.0.110:7006
slots: (0 slots) slave
replicates 50d9fb7c86d15847233fc691f6e54e6173fd5d85
M: 477af6a0d870cd61930d654667992eb4abeac920 192.168.0.109:7004
slots:[0-5460] (5461 slots) master
M: 50d9fb7c86d15847233fc691f6e54e6173fd5d85 192.168.0.109:7003
slots:[5461-10922] (5462 slots) master
643121260372426753 additional replica(s)
M: 94b96a1d009232fc013f9d15535000087c82248a 192.168.0.110:7005
slots:[10923-16383] (5461 slots) master
643122188085362689 additional replica(s)
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.
可以看到192.168.0.109:7004从Slave转变为了Master，说明故障自动转移和恢复是成功的。然后，我们再重新启动192.168.0.107:7001，再check一下集群状态，发现192.168.0.107:7001已经变成了Slave：
[root@ressmix-dsf01 init.d]# redis-cli --cluster check 192.168.0.107:7002
192.168.0.109:7004 (477af6a0...) -> 1 keys | 5461 slots | 1 slaves.
192.168.0.109:7003 (50d9fb7c...) -> 0 keys | 5462 slots | 1 slaves.
192.168.0.110:7005 (94b96a1d...) -> 0 keys | 5461 slots | 1 slaves.
[OK] 1 keys in 3 masters.
0.00 keys per slot on average.
>>> Performing Cluster Check (using node 192.168.0.107:7002)
S: 359dec6ff52d85c983f49c97997b98d91758f049 192.168.0.107:7002
slots: (0 slots) slave
replicates 94b96a1d009232fc013f9d15535000087c82248a
S: 6ac3762d35e60af9d8d12f3ac148c8ede11b94af 192.168.0.107:7001
slots: (0 slots) slave
replicates 477af6a0d870cd61930d654667992eb4abeac920
S: e8091d7c48647d55a34fa83949600d025d68b879 192.168.0.110:7006
slots: (0 slots) slave
replicates 50d9fb7c86d15847233fc691f6e54e6173fd5d85
M: 477af6a0d870cd61930d654667992eb4abeac920 192.168.0.109:7004
slots:[0-5460] (5461 slots) master
603169234066866177 additional replica(s)
M: 50d9fb7c86d15847233fc691f6e54e6173fd5d85 192.168.0.109:7003
slots:[5461-10922] (5462 slots) master
603169612023988225 additional replica(s)
M: 94b96a1d009232fc013f9d15535000087c82248a 192.168.0.110:7005
slots:[10923-16383] (5461 slots) master
603170539736924161 additional replica(s)
[OK] All nodes agree about slots configuration.
三、总结
本章，我介绍了redis cluster的部署方式，读者可以自己在本地尝试搭建一个redis cluster集群，以便加深理解。
