前言：

　　缓存对应实时性要求比较高的业务需求可谓十分重要，但缓存中的实时数据一旦丢失，将直接影响整个业务功能。考虑到各场景的业务需求，redis推出了高可用的方案，即：主从+哨兵的HA机制。　　

准备

"redis-4.0.6.tar.gz"  下载地址（https://redis.io/download ）

$ tar -zxvf redis-4.0.6.tar.gz 

$ ln -s redis-4.0.6 redis

 

1. 背景

升级之前：

单实例 Jedis Instance

-------------------------------------

HA升级方案：

 

2. 实现步骤

step1:主节点test-zk02增加如下服务启动：

test-zk02 (master)

src/redis-sentinel.conf

配置文件的内容见“附1”

 

test-zk03 (slave)

src/redis-server redis6379.conf

复制主节点的配置后在末尾加上一行"slaveof test-zk02 6379"即可

 

step2:代码层修改如下：

Set
     
       sentinels = new HashSet
      
       ();sentinels.add("192.168.10.80:26379");JedisSentinelPool jedisPool = new JedisSentinelPool("mymaster", sentinels);
      
     

3. 注意

1).服务升级用户为root

2).配置哨兵日志 sentinel26379.log

3).修改代码层之后需及时同步redis配置

4).升级HA后观察连接数变动 " echo 'info clients' | /opt/redis-cli -h test-zk02 -p 6379"

附1

-------------------------------

sentinel.conf配置内容

bind 127.0.0.1 test-zk02

protected-mode no

daemonize yes

logfile "/opt/redis/s26379.log"

port 26379

dir /tmp

sentinel monitor test-master test-zk02 6379 1

sentinel down-after-milliseconds test-master 30000

sentinel parallel-syncs test-master 1

sentinel failover-timeout test-master 180000

附2

-------------------------------

主从切换调试命令

/opt/redis-cli -h localhost -p 9998 debug sleep 20

主从切换日志

55794:X 01 Jun 15:25:55.813 # +sdown master test-master 192.168.10.80 9999

55794:X 01 Jun 15:25:55.813 # +odown master test-master 192.168.10.80 9999 #quorum 1/1

55794:X 01 Jun 15:25:55.813 # +new-epoch 3

55794:X 01 Jun 15:25:55.813 # +try-failover master test-master 192.168.10.80 9999

55794:X 01 Jun 15:25:55.828 # +vote-for-leader 35e9cce79ec5d7a2f1cd9473d089166b533c4690 3

55794:X 01 Jun 15:25:55.828 # +elected-leader master test-master 192.168.10.80 9999

55794:X 01 Jun 15:25:55.828 # +failover-state-select-slave master test-master 192.168.10.80 9999

55794:X 01 Jun 15:25:55.886 # +selected-slave slave 192.168.10.80:9998 192.168.10.80 9998 @ test-master 192.168.10.80 9999

55794:X 01 Jun 15:25:55.886 * +failover-state-send-slaveof-noone slave 192.168.10.80:9998 192.168.10.80 9998 @ test-master 192.168.10.80 9999

55794:X 01 Jun 15:25:55.945 * +failover-state-wait-promotion slave 192.168.10.80:9998 192.168.10.80 9998 @ test-master 192.168.10.80 9999

55794:X 01 Jun 15:25:56.894 # +promoted-slave slave 192.168.10.80:9998 192.168.10.80 9998 @ test-master 192.168.10.80 9999

55794:X 01 Jun 15:25:56.894 # +failover-state-reconf-slaves master test-master 192.168.10.80 9999

55794:X 01 Jun 15:25:56.969 # +failover-end master test-master 192.168.10.80 9999

55794:X 01 Jun 15:25:56.969 # +switch-master test-master 192.168.10.80 9999 192.168.10.80 9998

55794:X 01 Jun 15:25:56.969 * +slave slave 192.168.10.80:9999 192.168.10.80 9999 @ test-master 192.168.10.80 9998

55794:X 01 Jun 15:26:25.531 * +convert-to-slave slave 192.168.10.80:9999 192.168.10.80 9999 @ test-master 192.168.10.80 9998

 

参考资料

https://redis.io/topics/sentinel