Redis 哨兵（sentinel）模式集群配置（5.0.3版本）
2024-08-21 11:04:57
一、准备工作
1.系统环境：centos6.4
2.服务器六台（1主5从）：
192.168.1.161（master）
192.168.1.162（slave）
192.168.1.163（slave）
192.168.1.141（slave）
192.168.1.142（slave）
192.168.1.143（slave）
2.redis版本：5.0.3
3.安装：
进入到目录：cd /usr/local
下载redis：wget http://download.redis.io/releases/redis-5.0.3.tar.gz
下载完成后解压：tar zxvf redis-5.0.3.tar.gz
重命名为redis文件夹（这一步纯属个人喜好）：mv redis-5.0.3 redis
进入到redis文件夹：cd redis
编译及安装：make && make install
特别说明：官方文档只给出了make（编译），没有给出make install（安装）
二、配置
1.Master配置，把master的redis.conf拷贝了一份命名为redis_master.conf
设置：
### NETWORK 设置：
# bind 127.0.0.1 //绑定监控的网卡IP，注释掉bind，任何ip均可访问，
protected-mode no //保护模式关闭，这里使用密码访问
port 17000 //设置端口，建议测试时可以使用默认端口，我这里改掉了，建议生产环境均使用自定义端口
timeout 30 //Client 端空闲断开连接的时间
### GENERAL 设置：
daemonize yes //后台模式运行
pidfile /var/run/redis_17000.pid //pid进程文件名
logfile /usr/local/redis/logs/redis.log //日志文件的位置
### SNAPSHOTTING 设置：
dir /usr/local/redis/datas //快照文件的路径
### APPEND ONLY MODE 设置：
appendonly yes //默认值是No，意思是不使用AOF增量持久化的方式，使用RDB全量持久化的方式。把No值改成Yes，使用AOF增量持久化的方式
appendfsync always
###SECURITY 设置密码:，生产环境一定要使用复杂密码
requirepass 123456
2.Slave配置，我把master的redis.conf拷贝了一份命名为redis_slave.conf
### NETWORK 设置：
# bind 127.0.0.1 //注释掉bind，任何ip均可访问
port 17000 //设置端口
protected-mode no //保护模式关闭，使用密码访问
timeout 30 //Client 端空闲断开连接的时间
### GENERAL 设置：
daemonize yes //后台模式运行
pidfile /var/run/redis_17000.pid
logfile /usr/local/redis/logs/redis.log //日志文件的位置
### SNAPSHOTTING 设置：
dir /usr/local/redis/datas //SNAPSHOTTING文件的路径
### REPLICATION 设置：
replicaof 192.168.1.161 17000 //主服务器的Ip地址和Port端口号
replica-serve-stale-data no //如果slave 无法与master 同步，设置成slave不可读，方便监控脚本发现问题。
masterauth 123456 //master的密码
### APPEND ONLY MODE 设置：
appendonly yes //默认值是No，意思是不使用AOF增量持久化的方式，使用RDB全量持久化的方式。把No值改成Yes，使用AOF增量持久化的方式
appendfsync always
###SECURITY 设置密码:，生产环境一定要使用复杂密码
requirepass 123456
3.Sentinel（哨兵）配置，配置文件为sentinel.conf
port 16000 //哨兵端口号
protected-mode no //关闭保护模式
daemonize yes //守护进程
dir /usr/local/redis/sentinel/ //哨兵程序的工作路径
sentinel auth-pass mymaster 123456 //master的访问密码
//Sentinel去监视一个名为mymaster的主redis实例，这个主实例的IP地址为本机地址192.168.1.161，端口号为17000，而将这个主实例判断为失效至少需要1个 Sentinel进程的同意，只要同意Sentinel的数量不达标，自动failover就不会执行
sentinel monitor mymaster 192.168.1.161 17000 1
sentinel down-after-milliseconds mymaster 5000 //哨兵程序每5秒检测一次Master是否正常
//指定了在执行故障转移时，最多可以有多少个从Redis实例在同步新的主实例，在从Redis实例较多的情况下这个数字越小，同步的时间越长，完成故障转移所需的时间就越长
sentinel parallel-syncs mymaster 2
sentinel failover-timeout mymaster 300000 //如果在该时间（ms）内未能完成failover操作，则认为该failover失败，生产环境需要根据数据量设置该值
三、启动Redis服务
1.在各服务器上先建立配置文件中需要的文件夹
mkdir -p /usr/local/redis/datas
mkdir -p /usr/local/redis/logs
mkdir -p /usr/local/redis/sentinel
2.启动服务
先开启master服务，也就是192.168.1.161服务器
进入到redis文件夹中：cd /usr/local/redis
### 启动master:
src/redis-server redis_master.conf
注意： 这里指定了配置文件redis_master.conf
使用redis-cli访问服务端，查看状态
src/redis-cli -p 17000 //由于修改了端口，所以在使用redis-cli工具时，需要指定端口
看到127.0.0.1:17000> 表示连接到服务端
输入info replication 查看集群状态，这里报错：NOAUTH Authentication required.
原因是：master和slave均设置了密码，如果要查看replication，则需要输入密码授权：
AUTH 123456 --密码就是123456（配置文件里配置的），看到OK表示授权通过，再次输入info replication便可以显示集群状态：
[root@mongodb- redis]# src/redis-server redis_master.conf
[root@mongodb- redis]# src/redis-cli -p
127.0.0.1:> info replication
NOAUTH Authentication required.
127.0.0.1:> AUTH
OK
127.0.0.1:> info replication
# Replication
role:master
connected_slaves:
master_replid:96942e0268bff14dba3b6d1e22a9b55e9e77343b
master_replid2:
master_repl_offset:
second_repl_offset:-
repl_backlog_active:
repl_backlog_size:
repl_backlog_first_byte_offset:
repl_backlog_histlen:
127.0.0.1:>
这里只是刚启动了master一个服务，所以connected_slaves:0
再启动192.168.1.162服务器上的redis
src/redis-server redis_slave.conf
注意：这里指定了配置文件redis_slave.conf
启动完成后使用redis-cli -p 17000 查看状态
[root@mongodb- redis]# src/redis-server redis_slave.conf
[root@mongodb- redis]# src/redis-cli -p
127.0.0.1:> AUTH
OK
127.0.0.1:> info replication
# Replication
role:slave
master_host:192.168.1.161
master_port:
master_link_status:up
master_last_io_seconds_ago:
master_sync_in_progress:
slave_repl_offset:
slave_priority:
slave_read_only:
connected_slaves:
master_replid:b3f15ec093bb8ab2b1e908b41740f4bdabbfeba4
master_replid2:
master_repl_offset:
second_repl_offset:-
repl_backlog_active:
repl_backlog_size:
repl_backlog_first_byte_offset:
repl_backlog_histlen:
127.0.0.1:>
可以看到当前的状态是slave，连接的master是192.168.1.161，这里再来查看192.168.1.161的master上的状态
127.0.0.1:> info replication
# Replication
role:master
connected_slaves:
slave0:ip=192.168.1.162,port=,state=online,offset=,lag=
master_replid:b3f15ec093bb8ab2b1e908b41740f4bdabbfeba4
master_replid2:
master_repl_offset:
second_repl_offset:-
repl_backlog_active:
repl_backlog_size:
repl_backlog_first_byte_offset:
repl_backlog_histlen:
127.0.0.1:>
可以看到连接到master有一个slave了，继续把剩余服务器上的redis都启动起来，然后再查看master状态
127.0.0.1:> info replication
# Replication
role:master
connected_slaves:
slave0:ip=192.168.1.162,port=,state=online,offset=,lag=
slave1:ip=192.168.1.163,port=,state=online,offset=,lag=
slave2:ip=192.168.1.141,port=,state=online,offset=,lag=
slave3:ip=192.168.1.142,port=,state=online,offset=,lag=
slave4:ip=192.168.1.143,port=,state=online,offset=,lag=
master_replid:b3f15ec093bb8ab2b1e908b41740f4bdabbfeba4
master_replid2:
master_repl_offset:
second_repl_offset:-
repl_backlog_active:
repl_backlog_size:
repl_backlog_first_byte_offset:
repl_backlog_histlen:
127.0.0.1:>
这里5个slave均已成功连接到master上了。
四、启动Sentinel（哨兵）进程
哨兵进程不一定与redis数量一致，也不一定要放在redis服务器上，sentinel的作用是监控所有服务及与其它哨兵通信，若sentinel单独放其它服务器上，则也需要安装redis，sentinel只是redis软件包中的一个服务
每台服务器上都放了一个sentinel进程
启动命令：src/redis-sentinel sentinel.conf
启动后查看日志：
[root@redis redis]# src/redis-sentinel sentinel.conf
[root@redis redis]# cat sentinel/sentinel.log
:X Mar ::52.042 # oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo
:X Mar ::52.042 # Redis version=5.0., bits=, commit=, modified=, pid=, just started
:X Mar ::52.042 # Configuration loaded
:X Mar ::52.059 * Increased maximum number of open files to (it was originally set to ).
:X Mar ::52.061 * Running mode=sentinel, port=.
:X Mar ::52.061 # WARNING: The TCP backlog setting of cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of .
:X Mar ::52.071 # Sentinel ID is 53e37f3442336b7b229fa42c655bdca9eaa6578c
:X Mar ::52.072 # +monitor master mymaster 192.168.1.161 quorum
:X Mar ::52.073 * +slave slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::52.075 * +slave slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::52.076 * +slave slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::52.078 * +slave slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::52.079 * +slave slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::52.515 * +sentinel sentinel 4a32c88aa8bff64295e45f8bb9c25154f4533d2a 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::54.057 * +sentinel sentinel 93da0f13c96f22e0263e4288c68390cce5e11cea 192.168.1.161 @ mymaster 192.168.1.161
:X Mar ::54.078 * +sentinel sentinel 605bf637a218de833308aa590cb3346de0eaacaf 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::55.520 * +sentinel sentinel 3607a3f65b224ee03e49a825604b40d1cc1fbeaf 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::58.040 * +sentinel sentinel d8e333f5daec7809605b8f276b1f2b44c1711459 192.168.1.143 @ mymaster 192.168.1.161
这里有个警告是由于TCP backlog 设置了511，而系统目前仅支持128，可以先忽略。
目前所有master-slave服务均正常。
五、故障转移
模拟master宕机，这里直接把master进程杀掉
查看redis进程：
[root@mongodb-161 redis]# ps -e | grep redis
? :: redis-server
? :: redis-sentinel
redis-server进程号为2730
按进程号杀掉：kill -9 2730
[root@mongodb- redis]# kill - 2730
[root@mongodb- redis]# ps -e | grep redis
2748 ? :: redis-sentinel
[root@mongodb- redis]#
杀掉后再查看进程，已经没有了redis-server进程了。
然后查看192.168.1.161上的sentinel日志
[root@mongodb- redis]# cat sentinel/sentinel.log
:X Mar ::42.113 # oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo
:X Mar ::42.113 # Redis version=5.0., bits=, commit=, modified=, pid=, just started
:X Mar ::42.113 # Configuration loaded
:X Mar ::42.127 * Increased maximum number of open files to (it was originally set to ).
:X Mar ::42.129 * Running mode=sentinel, port=.
:X Mar ::42.129 # WARNING: The TCP backlog setting of cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of .
:X Mar ::42.139 # Sentinel ID is 93da0f13c96f22e0263e4288c68390cce5e11cea
:X Mar ::42.139 # +monitor master mymaster 192.168.1.161 quorum
:X Mar ::42.141 * +slave slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::42.143 * +slave slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::42.144 * +slave slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::42.146 * +slave slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::42.147 * +slave slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::50.335 * +sentinel sentinel 605bf637a218de833308aa590cb3346de0eaacaf 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::52.836 * +sentinel sentinel 4a32c88aa8bff64295e45f8bb9c25154f4533d2a 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::54.374 * +sentinel sentinel 53e37f3442336b7b229fa42c655bdca9eaa6578c 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::55.841 * +sentinel sentinel 3607a3f65b224ee03e49a825604b40d1cc1fbeaf 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::58.361 * +sentinel sentinel d8e333f5daec7809605b8f276b1f2b44c1711459 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::46.300 # +new-epoch
:X Mar ::46.301 # +vote-for-leader 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.336 # +sdown master mymaster 192.168.1.161
:X Mar ::46.336 # +odown master mymaster 192.168.1.161 #quorum /
:X Mar ::46.336 # Next failover delay: I will not start a failover before Fri Mar ::
:X Mar ::46.635 # +config-update-from sentinel 4a32c88aa8bff64295e45f8bb9c25154f4533d2a 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::46.635 # +switch-master mymaster 192.168.1.161 192.168.1.162
:X Mar ::46.636 * +slave slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.162
:X Mar ::46.636 * +slave slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.162
:X Mar ::46.636 * +slave slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.162
:X Mar ::46.636 * +slave slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.162
:X Mar ::46.636 * +slave slave 192.168.1.161: 192.168.1.161 @ mymaster 192.168.1.162
:X Mar ::51.638 # +sdown slave 192.168.1.161: 192.168.1.161 @ mymaster 192.168.1.162
可以看到时间在16:25:46时master宕机后，已经进行了故障转移，新的master是192.168.1.162，查看192.168.1.162的sentinel日志：
[root@mongodb- redis]# cat sentinel/sentinel.log
:X Mar ::48.245 # oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo
:X Mar ::48.245 # Redis version=5.0., bits=, commit=, modified=, pid=, just started
:X Mar ::48.245 # Configuration loaded
:X Mar ::48.250 * Increased maximum number of open files to (it was originally set to ).
:X Mar ::48.251 * Running mode=sentinel, port=.
:X Mar ::48.252 # WARNING: The TCP backlog setting of cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of .
:X Mar ::48.254 # Sentinel ID is 605bf637a218de833308aa590cb3346de0eaacaf
:X Mar ::48.254 # +monitor master mymaster 192.168.1.161 quorum
:X Mar ::48.256 * +slave slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::48.257 * +slave slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::48.259 * +slave slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::48.261 * +slave slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::48.262 * +slave slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::48.270 * +sentinel sentinel 93da0f13c96f22e0263e4288c68390cce5e11cea 192.168.1.161 @ mymaster 192.168.1.161
:X Mar ::52.815 * +sentinel sentinel 4a32c88aa8bff64295e45f8bb9c25154f4533d2a 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::54.355 * +sentinel sentinel 53e37f3442336b7b229fa42c655bdca9eaa6578c 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::55.820 * +sentinel sentinel 3607a3f65b224ee03e49a825604b40d1cc1fbeaf 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::58.341 * +sentinel sentinel d8e333f5daec7809605b8f276b1f2b44c1711459 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::46.277 # +new-epoch
:X Mar ::46.279 # +vote-for-leader 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.309 # +sdown master mymaster 192.168.1.161
:X Mar ::46.309 # +odown master mymaster 192.168.1.161 #quorum /
:X Mar ::46.309 # Next failover delay: I will not start a failover before Fri Mar ::
:X Mar ::46.612 # +config-update-from sentinel 4a32c88aa8bff64295e45f8bb9c25154f4533d2a 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::46.612 # +switch-master mymaster 192.168.1.161 192.168.1.162
:X Mar ::46.614 * +slave slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.162
:X Mar ::46.614 * +slave slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.162
:X Mar ::46.614 * +slave slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.162
:X Mar ::46.614 * +slave slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.162
:X Mar ::46.614 * +slave slave 192.168.1.161: 192.168.1.161 @ mymaster 192.168.1.162
:X Mar ::51.622 # +sdown slave 192.168.1.161: 192.168.1.161 @ mymaster 192.168.1.162
跟161上日志几乎是一样的，那么162是怎么被选举出作为master的呢，继续查看163日志：
[root@mongodb- redis]# cat sentinel/sentinel.log
:X Mar ::50.731 # oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo
:X Mar ::50.731 # Redis version=5.0., bits=, commit=, modified=, pid=, just started
:X Mar ::50.731 # Configuration loaded
:X Mar ::50.747 * Increased maximum number of open files to (it was originally set to ).
:X Mar ::50.749 * Running mode=sentinel, port=.
:X Mar ::50.749 # WARNING: The TCP backlog setting of cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of .
:X Mar ::50.751 # Sentinel ID is 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::50.751 # +monitor master mymaster 192.168.1.161 quorum
:X Mar ::50.754 * +slave slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::50.756 * +slave slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::50.757 * +slave slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::50.759 * +slave slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::50.760 * +slave slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::52.329 * +sentinel sentinel 93da0f13c96f22e0263e4288c68390cce5e11cea 192.168.1.161 @ mymaster 192.168.1.161
:X Mar ::52.331 * +sentinel sentinel 605bf637a218de833308aa590cb3346de0eaacaf 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::54.330 * +sentinel sentinel 53e37f3442336b7b229fa42c655bdca9eaa6578c 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::55.796 * +sentinel sentinel 3607a3f65b224ee03e49a825604b40d1cc1fbeaf 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::58.316 * +sentinel sentinel d8e333f5daec7809605b8f276b1f2b44c1711459 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::46.247 # +sdown master mymaster 192.168.1.161
:X Mar ::46.248 # +odown master mymaster 192.168.1.161 #quorum /
:X Mar ::46.248 # +new-epoch
:X Mar ::46.248 # +try-failover master mymaster 192.168.1.161
:X Mar ::46.250 # +vote-for-leader 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.253 # 53e37f3442336b7b229fa42c655bdca9eaa6578c voted for 53e37f3442336b7b229fa42c655bdca9eaa6578c
:X Mar ::46.254 # 93da0f13c96f22e0263e4288c68390cce5e11cea voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.255 # 605bf637a218de833308aa590cb3346de0eaacaf voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.255 # 3607a3f65b224ee03e49a825604b40d1cc1fbeaf voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.255 # d8e333f5daec7809605b8f276b1f2b44c1711459 voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.316 # +elected-leader master mymaster 192.168.1.161
:X Mar ::46.316 # +failover-state-select-slave master mymaster 192.168.1.161
:X Mar ::46.372 # +selected-slave slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::46.372 * +failover-state-send-slaveof-noone slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::46.424 * +failover-state-wait-promotion slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::46.528 # +promoted-slave slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::46.528 # +failover-state-reconf-slaves master mymaster 192.168.1.161
:X Mar ::46.584 * +slave-reconf-sent slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::46.584 * +slave-reconf-sent slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::47.546 * +slave-reconf-inprog slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::47.546 * +slave-reconf-done slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::47.546 * +slave-reconf-inprog slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::47.546 * +slave-reconf-done slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::47.602 * +slave-reconf-sent slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::47.602 * +slave-reconf-sent slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::48.588 * +slave-reconf-inprog slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::48.588 * +slave-reconf-done slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::48.588 * +slave-reconf-inprog slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::49.661 * +slave-reconf-done slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::49.732 # +failover-end master mymaster 192.168.1.161
:X Mar ::49.732 # +switch-master mymaster 192.168.1.161 192.168.1.162
:X Mar ::49.733 * +slave slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.162
:X Mar ::49.733 * +slave slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.162
:X Mar ::49.733 * +slave slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.162
:X Mar ::49.733 * +slave slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.162
:X Mar ::49.733 * +slave slave 192.168.1.161: 192.168.1.161 @ mymaster 192.168.1.162
:X Mar ::54.743 # +sdown slave 192.168.1.161: 192.168.1.161 @ mymaster 192.168.1.162
在16:25:46.247时sentinel已将161进行了主观下线（sdown，标记为不可用），然后紧接着进行了客观下线（odown，sentinel认为确实不可用），这里说明一下，为什么主观下线后紧跟着客观下线，原因是配置了：
sentinel monitor mymaster 192.168.1.161 17000 1
最后这个1的意思是只要大于等于1个sentinel认为sdown，则认为可以进行odown了，所以日志中有一句：odown master mymaster 192.168.1.161 17000 #quorum 1/1，所以odown紧跟在sdown后面。
在163的sentinel的日志中：Sentinel ID is 4a32c88aa8bff64295e45f8bb9c25154f4533d2a 表示163上的sentinel的ID，后面会用到这个。
当sentinel设置161为odown（客观下线）后，这些sentinel开始进行投票选举，这里特别注意，选举的不是redis_server服务，而是sentinel服务，意思就是先选举一个sentinel，然后由它决定将哪个slave提升为master。
查看141服务器上的sentinel日志：
[root@redis redis]# cat sentinel/sentinel.log
:X Mar ::52.042 # oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo
:X Mar ::52.042 # Redis version=5.0., bits=, commit=, modified=, pid=, just started
:X Mar ::52.042 # Configuration loaded
:X Mar ::52.059 * Increased maximum number of open files to (it was originally set to ).
:X Mar ::52.061 * Running mode=sentinel, port=.
:X Mar ::52.061 # WARNING: The TCP backlog setting of cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of .
:X Mar ::52.071 # Sentinel ID is 53e37f3442336b7b229fa42c655bdca9eaa6578c
:X Mar ::52.072 # +monitor master mymaster 192.168.1.161 quorum
:X Mar ::52.073 * +slave slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::52.075 * +slave slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::52.076 * +slave slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::52.078 * +slave slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::52.079 * +slave slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::52.515 * +sentinel sentinel 4a32c88aa8bff64295e45f8bb9c25154f4533d2a 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::54.057 * +sentinel sentinel 93da0f13c96f22e0263e4288c68390cce5e11cea 192.168.1.161 @ mymaster 192.168.1.161
:X Mar ::54.078 * +sentinel sentinel 605bf637a218de833308aa590cb3346de0eaacaf 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::55.520 * +sentinel sentinel 3607a3f65b224ee03e49a825604b40d1cc1fbeaf 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::58.040 * +sentinel sentinel d8e333f5daec7809605b8f276b1f2b44c1711459 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::45.962 # +sdown master mymaster 192.168.1.161
:X Mar ::45.963 # +odown master mymaster 192.168.1.161 #quorum /
:X Mar ::45.963 # +new-epoch
:X Mar ::45.963 # +try-failover master mymaster 192.168.1.161
:X Mar ::45.966 # +vote-for-leader 53e37f3442336b7b229fa42c655bdca9eaa6578c
:X Mar ::45.966 # 4a32c88aa8bff64295e45f8bb9c25154f4533d2a voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::45.967 # 93da0f13c96f22e0263e4288c68390cce5e11cea voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::45.968 # 605bf637a218de833308aa590cb3346de0eaacaf voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::45.968 # 3607a3f65b224ee03e49a825604b40d1cc1fbeaf voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::45.968 # d8e333f5daec7809605b8f276b1f2b44c1711459 voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.301 # +config-update-from sentinel 4a32c88aa8bff64295e45f8bb9c25154f4533d2a 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::46.301 # +switch-master mymaster 192.168.1.161 192.168.1.162
:X Mar ::46.302 * +slave slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.162
:X Mar ::46.302 * +slave slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.162
:X Mar ::46.303 * +slave slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.162
:X Mar ::46.303 * +slave slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.162
:X Mar ::46.303 * +slave slave 192.168.1.161: 192.168.1.161 @ mymaster 192.168.1.162
:X Mar ::51.320 # +sdown slave 192.168.1.161: 192.168.1.161 @ mymaster 192.168.1.162
由141上日志看出：选举后，141上的sentinel并没有进行操作failover，所以可以证实选举是sentinel。
日志说明：
这段日志就是在进行选举：
:X Mar ::46.248 # +new-epoch
:X Mar ::46.248 # +try-failover master mymaster 192.168.1.161
:X Mar ::46.250 # +vote-for-leader 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.253 # 53e37f3442336b7b229fa42c655bdca9eaa6578c voted for 53e37f3442336b7b229fa42c655bdca9eaa6578c
:X Mar ::46.254 # 93da0f13c96f22e0263e4288c68390cce5e11cea voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.255 # 605bf637a218de833308aa590cb3346de0eaacaf voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.255 # 3607a3f65b224ee03e49a825604b40d1cc1fbeaf voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
:X Mar ::46.255 # d8e333f5daec7809605b8f276b1f2b44c1711459 voted for 4a32c88aa8bff64295e45f8bb9c25154f4533d2a
可以看出id为4a32c88aa8bff64295e45f8bb9c25154f4533d2a的sentinel有4票（共6个sentinel），它决定哪个slave升为master：
:X Mar ::46.316 # +elected-leader master mymaster 192.168.1.161
:X Mar ::46.316 # +failover-state-select-slave master mymaster 192.168.1.161
:X Mar ::46.372 # +selected-slave slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::46.372 * +failover-state-send-slaveof-noone slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::46.424 * +failover-state-wait-promotion slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::46.528 # +promoted-slave slave 192.168.1.162: 192.168.1.162 @ mymaster 192.168.1.161
:X Mar ::46.528 # +failover-state-reconf-slaves master mymaster 192.168.1.161
最终163上的sentinel选择了162为master，并进行了failover
其它各服务器则进行更新配置：
:X Mar ::46.584 * +slave-reconf-sent slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::46.584 * +slave-reconf-sent slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::47.546 * +slave-reconf-inprog slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::47.546 * +slave-reconf-done slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.161
:X Mar ::47.546 * +slave-reconf-inprog slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::47.546 * +slave-reconf-done slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.161
:X Mar ::47.602 * +slave-reconf-sent slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::47.602 * +slave-reconf-sent slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::48.588 * +slave-reconf-inprog slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::48.588 * +slave-reconf-done slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.161
:X Mar ::48.588 * +slave-reconf-inprog slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
:X Mar ::49.661 * +slave-reconf-done slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.161
最后，形成新的master-slave集群：
:X Mar ::49.733 * +slave slave 192.168.1.163: 192.168.1.163 @ mymaster 192.168.1.162
:X Mar ::49.733 * +slave slave 192.168.1.142: 192.168.1.142 @ mymaster 192.168.1.162
:X Mar ::49.733 * +slave slave 192.168.1.143: 192.168.1.143 @ mymaster 192.168.1.162
:X Mar ::49.733 * +slave slave 192.168.1.141: 192.168.1.141 @ mymaster 192.168.1.162
:X Mar ::49.733 * +slave slave 192.168.1.161: 192.168.1.161 @ mymaster 192.168.1.162
:X Mar ::54.743 # +sdown slave 192.168.1.161: 192.168.1.161 @ mymaster 192.168.1.162
查看162上的replication信息：
[root@mongodb- redis]# src/redis-cli -p
127.0.0.1:> AUTH
OK
127.0.0.1:> info replication
# Replication
role:master
connected_slaves:
slave0:ip=192.168.1.163,port=,state=online,offset=,lag=
slave1:ip=192.168.1.142,port=,state=online,offset=,lag=
slave2:ip=192.168.1.141,port=,state=online,offset=,lag=
slave3:ip=192.168.1.143,port=,state=online,offset=,lag=
master_replid:b6c12c4d48aa953cb473480198359da669f171cc
master_replid2:b3f15ec093bb8ab2b1e908b41740f4bdabbfeba4
master_repl_offset:
second_repl_offset:
repl_backlog_active:
repl_backlog_size:
repl_backlog_first_byte_offset:
repl_backlog_histlen:
127.0.0.1:>
再查看163上集群配置：
[root@mongodb- redis]# src/redis-cli -p
127.0.0.1:> AUTH
OK
127.0.0.1:> info replication
# Replication
role:slave
master_host:192.168.1.162
master_port:
master_link_status:up
master_last_io_seconds_ago:
master_sync_in_progress:
slave_repl_offset:
slave_priority:
slave_read_only:
connected_slaves:
master_replid:b6c12c4d48aa953cb473480198359da669f171cc
master_replid2:b3f15ec093bb8ab2b1e908b41740f4bdabbfeba4
master_repl_offset:
second_repl_offset:
repl_backlog_active:
repl_backlog_size:
repl_backlog_first_byte_offset:
repl_backlog_histlen:
127.0.0.1:>
最后，master宕机后，master-slave自动切换成功。
六、遇到的坑
1.slave不能连接到master，原因是防火墙开着的，解决方案是关闭防火墙，或把相应端口添加到防火墙中
2.master宕机后，不能自动选举新的master，查看日志报如下错误：
2230:X 15 Mar 2019 09:11:16.006 # -failover-abort-not-elected master mymaster 192.168.1.161 17000
2230:X 15 Mar 2019 09:11:16.007 # Next failover delay: I will not start a failover before Fri Mar 15 09:17:05 2019
原因是：原先开启了保护模式：protected-mode yes，而bind参数没有设置正确，实际上redis_master.conf、redis_slave.conf、sentinel.conf均需要设置正确，绑定上正确的网卡IP，我这里嫌麻烦改每台服务器配置（配置文件是做好后上传到服务器的），就把保护模式关闭了：protected-mode no。
官方配置中是默认开启的，一定要注意这点。不过文档中也建议，如果设置了访问密码，也可以关闭。
最新文章
- selenium 常见面试题以及答案（Java版）
- Java ThreadLocal
- 用UIImageView作出动画效果
- 查看局域网内某个ip的mac地址
- Oracle Create DBLink
- 简单模拟java动态动态代理机制的底层实现原理
- hibernate_validator_01
- libthrift0.9.0解析（三）之TProtocol&TTransport
- Linux vps无法发送邮件
- OMNeT++安装教程
- .net4.5的弱事件
- 转载：c# string.Format
- R实战 第三篇：数据处理（基础）
- Lesson 1-1
- 面试容易问到的Linux问题
- Codeforces ECR47F Dominant Indices（线段树合并）
- Kafka：ZK+Kafka+Spark Streaming集群环境搭建（三十）：使用flatMapGroupsWithState替换agg
- javascript定时器使用
- appium+python自动化测试
- Android基础知识之Manifest中的Intent-filter元素