docker sentinel的配置文件_用Docker-compose编排令你头疼的Redis集群

redis常用架构

主从

• 一主多从
• 主从读写分离

daemonize no docker要求容器内至少有一条前台进程,故不能设置为守护进程

从节点设置主节点ip replicaof 

version: '3'services:  master:    image: redis    container_name: redis-master    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-master-follower/master/conf:/etc/redis      - /usr/local/etc/docker-compose/redis-master-follower/master/data:/data      - /usr/local/etc/docker-compose/redis-master-follower/master/log:/log    ports:      - 6380:6379    networks:      redisNet:        ipv4_address: 192.168.88.80  follower:    image: redis    container_name: redis-follower    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-master-follower/follower/conf:/etc/redis      - /usr/local/etc/docker-compose/redis-master-follower/follower/data:/data      - /usr/local/etc/docker-compose/redis-master-follower/follower/log:/log    ports:      - 6381:6379    networks:      redisNet:        ipv4_address: 192.168.88.81networks:  redisNet:    external: true

sentinel哨兵

• 保证高可用,解决主节点宕机后无法写入的问题
• 主节点挂掉后哨兵会发布sdown(主观宕机),当sdown数量达到预设值时即为odown(客观宕机)
• 一般配置成半数sdown即odown,所以sentinel架构最小单位应为1主2从3哨兵
• 哨兵启动成功后写入myid

设置哨兵监控的主服务器和odown数(sdown数累计达到odown即开启故障转移主从切换)sentinel monitor mymaster 192.168.11.128 6379 2x秒不回应即单sentinel判定为sdownsentinel down-after-milliseconds单次故障转移最长时间,超过则认定为故障转移失败sentinel failover-timeout mymaster 10000 当发生failover主备切换时最多可以有多少个slave同时对新的master同步,数字越小favilover越快.通常设为1sentinel parallel-syncs 哨兵启动后自动写入的唯一id,每个sentinel不一样,故不能用同一份配置文件sentinel myid

这里哨兵容器应该等待redis容器先启动,使用depends_on控制

version: '3'services:  master:    image: redis    container_name: redis-master    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-sentinel/master/conf:/etc/redis      - /usr/local/etc/docker-compose/redis-sentinel/master/data:/data      - /usr/local/etc/docker-compose/redis-sentinel/master/log:/log    ports:      - 6380:6379    networks:      redisNet:        ipv4_address: 192.168.88.80  follower1:    image: redis    container_name: redis-follower1    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-sentinel/follower1/conf:/etc/redis      - /usr/local/etc/docker-compose/redis-sentinel/follower1/data:/data      - /usr/local/etc/docker-compose/redis-sentinel/follower1/log:/log    ports:      - 6381:6379    networks:      redisNet:        ipv4_address: 192.168.88.81  follower2:    image: redis    container_name: redis-follower2    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-sentinel/follower2/conf:/etc/redis      - /usr/local/etc/docker-compose/redis-sentinel/follower2/data:/data      - /usr/local/etc/docker-compose/redis-sentinel/follower2/log:/log    ports:      - 6382:6379    networks:      redisNet:        ipv4_address: 192.168.88.82        sentinel1:    image: redis    depends_on:      - master      - follower1      - follower2    container_name: redis-sentinel1    command: /bin/bash -c "redis-sentinel /etc/redis/redis-sentinel.conf"    volumes:      - /usr/local/etc/docker-compose/redis-sentinel/sentinel/sentinel1:/etc/redis      - /usr/local/etc/docker-compose/redis-sentinel/sentinel/sentinel1/data:/data      - /usr/local/etc/docker-compose/redis-sentinel/sentinel/sentinel1/log:/log    ports:      - 26379:26379    networks:      redisNet:        ipv4_address: 192.168.88.83  sentinel2:    image: redis    depends_on:      - master      - follower1      - follower2    container_name: redis-sentinel2    command: /bin/bash -c "redis-sentinel /etc/redis/redis-sentinel.conf"    volumes:      - /usr/local/etc/docker-compose/redis-sentinel/sentinel/sentinel2:/etc/redis      - /usr/local/etc/docker-compose/redis-sentinel/sentinel/sentinel2/data:/data      - /usr/local/etc/docker-compose/redis-sentinel/sentinel/sentinel2/log:/log    ports:      - 26380:26379    networks:      redisNet:        ipv4_address: 192.168.88.84  sentinel3:    image: redis    depends_on:      - master      - follower1      - follower2    container_name: redis-sentinel3    command: /bin/bash -c "redis-sentinel /etc/redis/redis-sentinel.conf"    volumes:      - /usr/local/etc/docker-compose/redis-sentinel/sentinel/sentinel3:/etc/redis      - /usr/local/etc/docker-compose/redis-sentinel/sentinel/sentinel3/data:/data      - /usr/local/etc/docker-compose/redis-sentinel/sentinel/sentinel3/log:/log    ports:      - 26381:26379    networks:      redisNet:        ipv4_address: 192.168.88.85networks:  redisNet:    external: true

redis-cluter

• redis5.0之后可以直接使用redis-cli创建集群,不再需要ruby.
• redis会将slot平均分配给master.所有数据操作都会通过CRC16(key) mod 16384得到的值决定此次操作所在分区.

cluster  cluster-enabled yes 开启集群模式cluster-node-timeout：连接超时时间cluster-announce-ip：集群各节点IP地址(以下不配置也行,redis会自动搜索)cluster-announce-port：集群节点映射端口cluster-announce-bus-port：集群总线端口cluster-require-full-coverage no　redis cluster需要16384个slot都正常的时候才能对外提供服务，换句话说，只要任何一个slot异常那么整个cluster不对外提供服务。 因此生产环境一般为no

version: '3'services:  master1:    image: redis    container_name: redis-cluster1-master    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-cluster/cluster1/master:/etc/redis      - /usr/local/etc/docker-compose/redis-cluster/cluster1/master:/data      - /usr/local/etc/docker-compose/redis-cluster/cluster1/master:/log    ports:      - 6379:6379    networks:      redisNet:        ipv4_address: 192.168.88.80  follow1:    image: redis    container_name: redis-cluster1-follower    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-cluster/cluster1/follower:/etc/redis      - /usr/local/etc/docker-compose/redis-cluster/cluster1/follower:/data      - /usr/local/etc/docker-compose/redis-cluster/cluster1/follower:/log    ports:      - 6380:6379    networks:      redisNet:        ipv4_address: 192.168.88.81  master2:    image: redis    container_name: redis-cluster2-master    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-cluster/cluster2/master:/etc/redis      - /usr/local/etc/docker-compose/redis-cluster/cluster2/master:/data      - /usr/local/etc/docker-compose/redis-cluster/cluster2/master:/log    ports:      - 6381:6379    networks:      redisNet:        ipv4_address: 192.168.88.82  follow2:    image: redis    container_name: redis-cluster2-follower    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-cluster/cluster2/follower:/etc/redis      - /usr/local/etc/docker-compose/redis-cluster/cluster2/follower:/data      - /usr/local/etc/docker-compose/redis-cluster/cluster2/follower:/log    ports:      - 6382:6379    networks:      redisNet:        ipv4_address: 192.168.88.83  master3:    image: redis    container_name: redis-cluster3-master    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-cluster/cluster3/master:/etc/redis      - /usr/local/etc/docker-compose/redis-cluster/cluster3/master:/data      - /usr/local/etc/docker-compose/redis-cluster/cluster3/master:/log    ports:      - 6383:6379    networks:      redisNet:        ipv4_address: 192.168.88.84  follow3:    image: redis    container_name: redis-cluster3-follower    command: /bin/bash -c "redis-server /etc/redis/redis.conf"    volumes:      - /usr/local/etc/docker-compose/redis-cluster/cluster3/follower:/etc/redis      - /usr/local/etc/docker-compose/redis-cluster/cluster3/follower:/data      - /usr/local/etc/docker-compose/redis-cluster/cluster3/follower:/log    ports:      - 6384:6379       networks:      redisNet:        ipv4_address: 192.168.88.85networks:  redisNet:    external: true

这里yml中的master/follower没有实际意义,具体的主从分配是redis决定的.

docker-compose成功启动六个redis容器之后,进入任意一个,配置redis-cluster集群.

编排容器docker-compose up -d进入容器docker exec -it container_id -bin/bash配置redis集群,每个master一个followerredis-cli --cluster create192.168.88.80:6379192.168.88.81:6379192.168.88.82:6379192.168.88.83:6379192.168.88.84:6379192.168.88.85:6379--cluster-replicas 1

如图,16384个slot平均分配给三台master.绑定关系为

83 replicaof 82

85 replicaof 81

84 replicaof 80