注意事项： 1.如果是使用redis-trib.rb工具构建集群，集群构建完成前不要配置密码，集群构建完毕再通过config set + config rewrite命令逐个机器设置密码 2.如果对集群设置密码，那么requirepass和masterauth都需要设置，否则发生主从切换时，就会遇到授权问题，可以模拟并观察日志 3.各个节点的密码都必须一致，否则Redirected就会失败
config set masterauth abcconfig set requirepass abcconfig rewrite
测试版本： redis3.0.7
测试环境： centos7
集群规模： 3主3从
集群构建成功前的redis配置：
port 8000cluster-enabled yescluster-config-file "nodes-8000.conf"cluster-node-timeout 15000dir "/opt/redisdata"appendonly yesappendfilename "appendonly-8000.aof"logfile "/opt/redisdata/8000.log"daemonize yespidfile "/var/run/redis-8000.pid"dbfilename "dump-8000.rdb"cluster-require-full-coverage no
集群构建成功后的redis配置：
port 8004cluster-enabled yescluster-config-file "nodes-8004.conf"cluster-node-timeout 15000dir "/opt/redisdata"appendonly yesappendfilename "appendonly-8004.aof"logfile "/opt/redisdata/8004.log"daemonize yespidfile "/var/run/redis-8004.pid"dbfilename "dump-8004.rdb"cluster-require-full-coverage no# Generated by CONFIG REWRITErequirepass "abc"masterauth "abc"
测试： 1.8002是主，8005是8002的从，现在要kill掉8002模拟主从切换
2.查看8005的日志，8005最终被选举成新的master 21227:S 14 Jun 07:28:49.343 # Connection with master lost. 21227:S 14 Jun 07:28:49.343 * Caching the disconnected master state. 21227:S 14 Jun 07:28:49.433 * Connecting to MASTER 127.0.0.1:8002 21227:S 14 Jun 07:28:49.433 * MASTER <-> SLAVE sync started 21227:S 14 Jun 07:28:49.434 # Error condition on socket for SYNC: Connection refused 21227:S 14 Jun 07:28:50.437 * Connecting to MASTER 127.0.0.1:8002
21227:S 14 Jun 07:29:06.205 # Start of election delayed for 564 milliseconds (rank #0, offset 337). 21227:S 14 Jun 07:29:06.507 * Connecting to MASTER 127.0.0.1:8002 21227:S 14 Jun 07:29:06.507 * MASTER <-> SLAVE sync started 21227:S 14 Jun 07:29:06.507 # Error condition on socket for SYNC: Connection refused 21227:S 14 Jun 07:29:06.808 # Starting a failover election for epoch 10. 21227:S 14 Jun 07:29:06.810 # Failover election won: I'm the new master. 21227:S 14 Jun 07:29:06.810 # configEpoch set to 10 after successful failover 21227:M 14 Jun 07:29:06.810 * Discarding previously cached master state.
3.再次启动8002，查看8005的日志，8002已经变成8005的slave 21227:M 14 Jun 07:32:53.511 * Slave 127.0.0.1:8002 asks for synchronization 21227:M 14 Jun 07:32:53.511 * Full resync requested by slave 127.0.0.1:8002 21227:M 14 Jun 07:32:53.511 * Starting BGSAVE for SYNC with target: disk 21227:M 14 Jun 07:32:53.511 * Background saving started by pid 21613 21613:C 14 Jun 07:32:53.513 * DB saved on disk 21613:C 14 Jun 07:32:53.513 * RDB: 0 MB of memory used by copy-on-write 21227:M 14 Jun 07:32:53.612 * Background saving terminated with success 21227:M 14 Jun 07:32:53.612 * Synchronization with slave 127.0.0.1:8002 succeeded
4.简单测试了GET SET命令，正常
原文出自：/daiyudong/article/details/51674169
