redis 分片集群(cluster)搭建（Centos6.10 + Centos7.6 + Reids 6.2.6）

1、在安装目录创建多个配置文件夹来启动多个redis服务

cd  /usr/local/redis
##递归创建redis配置文件夹
mkdir -p redis_cluster/6370 ## 6370  6371  6372

 2、将redis.conf给不同目录下都拷贝一份

cp /usr/local/redis/conf/redis.conf  /usr/local/redis/redis_cluster/6370/  ##每个目录都拷贝一份

3、修改不同端口下的redis.conf

port  6370                                        //端口6370,6371,6372        
bind 本机ip                                       //默认ip为127.0.0.1 需要改为其他节点机器可访问的ip 192.168.2.64 否则创建集群时无法访问对应的端口，无法创建集群
daemonize    yes                               //redis后台运行
pidfile  /var/run/redis_6370.pid              //pidfile文件对应7000,7001,7002
cluster-enabled  yes                           //开启集群  把注释#去掉
cluster-config-file  nodes_6370.conf         //集群的配置  配置文件首次启动自动生成 7000,7001,7002
cluster-node-timeout  15000                //请求超时  默认15秒，可自行设置
appendonly  yes                           //aof日志开启  有需要就开启，它会每次写操作都记录一条日志
dbfilename dump-6370.rdb                  //redis持久化时存储的文件名
requirepass  123456                       //密码
masterauth 123456                         //主服务器密码，建议都设置成一样的

 4、启动两台服务器中的多个节点

##第一台机器（192.168.2.64）
/usr/local/redis/bin/redis-server /usr/local/redis/redis_cluster/6370/redis.conf 
/usr/local/redis/bin/redis-server /usr/local/redis/redis_cluster/6371/redis.conf 
/usr/local/redis/bin/redis-server /usr/local/redis/redis_cluster/6372/redis.conf 
##第二台机器（192.168.2.24）
/usr/local/redis/bin/redis-server /usr/local/redis/redis_cluster/6373/redis.conf 
/usr/local/redis/bin/redis-server /usr/local/redis/redis_cluster/6374/redis.conf 
/usr/local/redis/bin/redis-server /usr/local/redis/redis_cluster/6375/redis.conf 

5、检查各个服务是否都已启动成功（如果配置有错，比如端口冲突则会启动失败，又因后台启动，前台无打印），我们可以看到除了6371端口启用  16371端口也启用了，故如果配置防火墙时都应打开

[root@ora redis_cluster]# ps -ef | grep redis
root 14963 1 0 13:30 ? 00:00:01 ./bin/redis-server 192.168.2.64:6370 [cluster]
root 14969 1 0 13:30 ? 00:00:01 ./bin/redis-server 192.168.2.64:6371 [cluster]
root 15044 1 0 13:34 ? 00:00:01 ./bin/redis-server 192.168.2.64:6372 [cluster]
root 15514 14554 0 14:04 pts/0 00:00:00 grep redis
[root@ora redis_cluster]# netstat -tnlp| grep redis
tcp 0 0 127.0.0.1:16370 0.0.0.0:* LISTEN 14963/./bin/redis-s
tcp 0 0 192.168.2.64:16370 0.0.0.0:* LISTEN 14963/./bin/redis-s
tcp 0 0 127.0.0.1:16371 0.0.0.0:* LISTEN 14969/./bin/redis-s
tcp 0 0 192.168.2.64:16371 0.0.0.0:* LISTEN 14969/./bin/redis-s
tcp 0 0 127.0.0.1:16372 0.0.0.0:* LISTEN 15044/./bin/redis-s
tcp 0 0 192.168.2.64:16372 0.0.0.0:* LISTEN 15044/./bin/redis-s
tcp 0 0 127.0.0.1:6370 0.0.0.0:* LISTEN 14963/./bin/redis-s
tcp 0 0 192.168.2.64:6370 0.0.0.0:* LISTEN 14963/./bin/redis-s
tcp 0 0 127.0.0.1:6371 0.0.0.0:* LISTEN 14969/./bin/redis-s
tcp 0 0 192.168.2.64:6371 0.0.0.0:* LISTEN 14969/./bin/redis-s
tcp 0 0 127.0.0.1:6372 0.0.0.0:* LISTEN 15044/./bin/redis-s
tcp 0 0 192.168.2.64:6372 0.0.0.0:* LISTEN 15044/./bin/redis-s

另一台（192.168.2.24）

[root@project-deve redis]# ps -ef |grep redis
root 26281 1 0 14:05 ? 00:00:00 ./bin/redis-server 192.168.2.24:6373 [cluster]
root 26287 1 0 14:05 ? 00:00:00 ./bin/redis-server 192.168.2.24:6374 [cluster]
root 26293 1 0 14:05 ? 00:00:00 ./bin/redis-server 192.168.2.24:6375 [cluster]
root 26299 25341 0 14:05 pts/0 00:00:00 grep --color=auto redis
[root@project-deve redis]# netstat -tnlp | grep redis
tcp 0 0 127.0.0.1:6375 0.0.0.0:* LISTEN 26293/./bin/redis-s
tcp 0 0 192.168.2.24:6375 0.0.0.0:* LISTEN 26293/./bin/redis-s
tcp 0 0 127.0.0.1:16373 0.0.0.0:* LISTEN 26281/./bin/redis-s
tcp 0 0 192.168.2.24:16373 0.0.0.0:* LISTEN 26281/./bin/redis-s
tcp 0 0 127.0.0.1:16374 0.0.0.0:* LISTEN 26287/./bin/redis-s
tcp 0 0 192.168.2.24:16374 0.0.0.0:* LISTEN 26287/./bin/redis-s
tcp 0 0 127.0.0.1:16375 0.0.0.0:* LISTEN 26293/./bin/redis-s
tcp 0 0 192.168.2.24:16375 0.0.0.0:* LISTEN 26293/./bin/redis-s
tcp 0 0 127.0.0.1:6373 0.0.0.0:* LISTEN 26281/./bin/redis-s
tcp 0 0 192.168.2.24:6373 0.0.0.0:* LISTEN 26281/./bin/redis-s
tcp 0 0 127.0.0.1:6374 0.0.0.0:* LISTEN 26287/./bin/redis-s
tcp 0 0 192.168.2.24:6374 0.0.0.0:* LISTEN 26287/./bin/redis-s

6、创建集群，redis5.0以上已经集成了ruby，不需要单独安装ruby,并且安装集群命令有异 redis-cli（5.0以上）、redis-trib.rb(5.0以下)

##如果设置了密码，则需要加  -a 123456
/usr/local/redis/bin/redis-cli -a 123456 --cluster create 192.168.2.64:6370 192.168.2.64:6371 192.168.2.64:6372 192.168.2.24:6373 192.168.2.24:6374 192.168.2.24:6375 --cluster-replicas 1
#此处需要注意，创建集群时 --cluster-replicas 1 代表为每一个master分配1个slave 根据所提供的node数去分配主从，前三个节点为master，后三个为slave
##在正式环境切勿将全部主节点放在同一台主机，如果集群中超过半数以上master挂掉（需要选举），则无论是否有slave，集群都会进入fail状态，不可用。（比如断电）

7、出现如上截图说明redis集群已经安装成功，下来登陆集群，所谓登陆集群其实就是带参数 -c 登录任意一个节点

## -h ip -p port  -a password  -c 表示登录集群，不加表示登录单个节点
/usr/local/redis/bin/redis-cli -h 192.168.2.24 -p 6373 -a 123456 -c

测试过程中会发现一个很有意思的现象，在6370上写入c=200提示写入了7365数据槽，在24服务器的6373服务上，并且直接切换了过去，并且在此服务上查询无误。这或许就是cluster集群中的去中心化、数据分slot存储吧，后面再细细研究。

8、（以下34节点信息为新增的集群节点，不影响展示）模拟集群master节点挂掉，6373（主）、6372（从）互为主从。

192.168.2.64:6370> cluster nodes
c4729c87272ebab033d1b78ca00b6c93a15c02eb 192.168.2.24:6375@16375 slave 05afba26af286f583204ed8db95cef91cf422d9f 0 1634021664306 1 connected
05afba26af286f583204ed8db95cef91cf422d9f 192.168.2.64:6370@16370 myself,master - 0 1634021655000 1 connected 0-5460
36f8084c986ce896593ea7cf00496fdcdef70a11 192.168.2.24:6374@16374 slave bd6c1bd4328c3a8f0b17987285f007e7a3733776 0 1634021662301 2 connected
bd6c1bd4328c3a8f0b17987285f007e7a3733776 192.168.2.64:6371@16371 master - 0 1634021663304 2 connected 10923-16383
2f899821dbfe2e2d5aa2c22e9e023eb9a9f127f3 192.168.2.34:6378@16378 master - 0 1634021659000 8 connected
a0069cc8774b1139c4ff696c1c2dbfbe0b44d598 192.168.2.64:6372@16372 slave 088548d9ebb6e3b908bb74e0f8d8a397b2c53cd3 0 1634021661297 4 connected
0936b30e022b7fad6ddb3ce6aee1b96b8c1d1c6b 192.168.2.34:6376@16376 slave bd6c1bd4328c3a8f0b17987285f007e7a3733776 0 1634021665309 2 connected
088548d9ebb6e3b908bb74e0f8d8a397b2c53cd3 192.168.2.24:6373@16373 master - 0 1634021663000 4 connected 5461-10922
e954db69f71beb9426e01325064de520d156a67a 192.168.2.34:6377@16377 master - 0 1634021661000 0 connected

9、停掉6373，再查看集群状态，可以看到6373已经是fail状态，6372自动选举为主节点。

192.168.2.64:6370> cluster nodes
c4729c87272ebab033d1b78ca00b6c93a15c02eb 192.168.2.24:6375@16375 slave 05afba26af286f583204ed8db95cef91cf422d9f 0 1634022069602 1 connected
05afba26af286f583204ed8db95cef91cf422d9f 192.168.2.64:6370@16370 myself,master - 0 1634022067000 1 connected 0-5460
36f8084c986ce896593ea7cf00496fdcdef70a11 192.168.2.24:6374@16374 slave bd6c1bd4328c3a8f0b17987285f007e7a3733776 0 1634022065000 2 connected
bd6c1bd4328c3a8f0b17987285f007e7a3733776 192.168.2.64:6371@16371 master - 0 1634022067591 2 connected 10923-16383
2f899821dbfe2e2d5aa2c22e9e023eb9a9f127f3 192.168.2.34:6378@16378 master - 0 1634022069602 8 connected
a0069cc8774b1139c4ff696c1c2dbfbe0b44d598 192.168.2.64:6372@16372 master - 0 1634022065582 9 connected 5461-10922
0936b30e022b7fad6ddb3ce6aee1b96b8c1d1c6b 192.168.2.34:6376@16376 slave a0069cc8774b1139c4ff696c1c2dbfbe0b44d598 0 1634022068596 9 connected
088548d9ebb6e3b908bb74e0f8d8a397b2c53cd3 192.168.2.24:6373@16373 master,fail - 1634021993275 1634021987000 4 disconnected
e954db69f71beb9426e01325064de520d156a67a 192.168.2.34:6377@16377 master - 0 1634022066587 0 connected

10、重新启动6373，发现6373已经加入集群，并且为6372的从节点，即master节点如果挂掉，它的slave节点变为新master节点继续对外提供服务，而原来的master节点如果重启，则变为新master节点的slave节点。

192.168.2.64:6370> cluster nodes
c4729c87272ebab033d1b78ca00b6c93a15c02eb 192.168.2.24:6375@16375 slave 05afba26af286f583204ed8db95cef91cf422d9f 0 1634022235270 1 connected
05afba26af286f583204ed8db95cef91cf422d9f 192.168.2.64:6370@16370 myself,master - 0 1634022232000 1 connected 0-5460
36f8084c986ce896593ea7cf00496fdcdef70a11 192.168.2.24:6374@16374 slave bd6c1bd4328c3a8f0b17987285f007e7a3733776 0 1634022232261 2 connected
bd6c1bd4328c3a8f0b17987285f007e7a3733776 192.168.2.64:6371@16371 master - 0 1634022233264 2 connected 10923-16383
2f899821dbfe2e2d5aa2c22e9e023eb9a9f127f3 192.168.2.34:6378@16378 master - 0 1634022230000 8 connected
a0069cc8774b1139c4ff696c1c2dbfbe0b44d598 192.168.2.64:6372@16372 master - 0 1634022230255 9 connected 5461-10922
0936b30e022b7fad6ddb3ce6aee1b96b8c1d1c6b 192.168.2.34:6376@16376 slave a0069cc8774b1139c4ff696c1c2dbfbe0b44d598 0 1634022232762 9 connected
088548d9ebb6e3b908bb74e0f8d8a397b2c53cd3 192.168.2.24:6373@16373 slave a0069cc8774b1139c4ff696c1c2dbfbe0b44d598 0 1634022235771 9 connected
e954db69f71beb9426e01325064de520d156a67a 192.168.2.34:6377@16377 master - 0 1634022234266 0 connected

 ## 如果master节点不在数据槽，不在（0-16383）范围内，则master挂掉后slave不会切换为master，待验证。

 ## 数据槽只会分配给master节点，slave节点备用，当master挂掉后接管这片数据槽。