Docker Compose搭建Redis主从复制与哨兵模式

1.概述

在Docker中搭建Redis哨兵模式非常方便，下面是一个示例，演示一下如何使用Docker Compose搭建一个Redis哨兵模式环境。首先，确保我们本地环境已经安装了Docker和Docker Compose。

示例代码地址

2.搭建主从模式

2.1.创建文件夹

mkdir -p redis-sentinel-replication/redis/.data
mkdir -p redis-sentinel-replication/redis/redis-server1
mkdir -p redis-sentinel-replication/redis/redis-server2
mkdir -p redis-sentinel-replication/redis/redis-server3

2.2.创建配置文件

给出的样例仅仅用于windows wsl docker本地测试，不具有正式环境部署参考价值

2.2.1.redis-server1/redis.conf

bind 0.0.0.0
loglevel debug
logfile "/data/redis-6379.log"
save 3600 1
save 300 100
save 60 10000
stop-writes-on-bgsave-error no
rdbcompression yes
rdbchecksum yes
dbfilename dump.rdb
rdb-del-sync-files no
dir /data/
requirepass 123456
appendonly yes
appendfilename "appendonly.aof"
appendfsync everysec
no-appendfsync-on-rewrite no
auto-aof-rewrite-percentage 100
auto-aof-rewrite-min-size 64mb
aof-load-truncated yes
aof-use-rdb-preamble yes
masterauth 123456
replica-announce-ip 172.200.0.2
replica-announce-port 6379
# 这里不开启的话，slave不一定能同步成功。windows wsl docker必须这样设置
repl-diskless-load on-empty-db

2.2.2.redis-server2/redis.conf

bind 0.0.0.0
loglevel debug
logfile "/data/redis-6380.log"
save 3600 1
save 300 100
save 60 10000
stop-writes-on-bgsave-error no
rdbcompression yes
rdbchecksum yes
dbfilename dump.rdb
rdb-del-sync-files no
dir /data/
requirepass 123456
appendonly yes
appendfilename "appendonly.aof"
appendfsync everysec
no-appendfsync-on-rewrite no
auto-aof-rewrite-percentage 100
auto-aof-rewrite-min-size 64mb
aof-load-truncated yes
aof-use-rdb-preamble yes
masterauth 123456
replica-announce-ip 172.200.0.3
replica-announce-port 6379
# 这里不开启的话，slave不一定能同步成功。windows wsl docker必须这样设置
repl-diskless-load on-empty-db

slaveof 172.200.0.2 6379
slave-read-only yes
slave-serve-stale-data yes

2.2.3.redis-server3/redis.conf

bind 0.0.0.0
loglevel debug
logfile "/data/redis-6381.log"
save 3600 1
save 300 100
save 60 10000
stop-writes-on-bgsave-error no
rdbcompression yes
rdbchecksum yes
dbfilename dump.rdb
rdb-del-sync-files no
dir /data/
requirepass 123456
appendonly yes
appendfilename "appendonly.aof"
appendfsync everysec
no-appendfsync-on-rewrite no
auto-aof-rewrite-percentage 100
auto-aof-rewrite-min-size 64mb
aof-load-truncated yes
aof-use-rdb-preamble yes
masterauth 123456
replica-announce-ip 172.200.0.4
replica-announce-port 6379
# 这里不开启的话，slave不一定能同步成功。windows wsl docker必须这样设置
repl-diskless-load on-empty-db

slaveof 172.200.0.2 6379
slave-read-only yes
slave-serve-stale-data yes

3.搭建哨兵集群

首先需要有一个redis主从集群，才能接着做redis哨兵。具体可以参考《第2小节搭建主从模式》

3.1.创建文件夹

mkdir -p redis-sentinel-replication/sentinel/.data
mkdir -p redis-sentinel-replication/sentinel/conf/redis-sentinel1
mkdir -p redis-sentinel-replication/sentinel/conf/redis-sentinel2
mkdir -p redis-sentinel-replication/sentinel/conf/redis-sentinel3

3.2.创建配置文件

3.2.1.redis-sentinel1/sentinel.conf

protected-mode no
port 26379
daemonize no
pidfile "/var/run/redis-sentine1.pid"
logfile "/data/sentinel-1.log"
sentinel announce-ip "172.200.0.5"
sentinel announce-port 26379
dir "/data"
sentinel monitor mymaster 172.200.0.2 6379 2
sentinel auth-pass mymaster 123456
acllog-max-len 128
sentinel deny-scripts-reconfig yes

3.2.2.redis-sentinel2/sentinel.conf

protected-mode no
port 26379
daemonize no
pidfile "/var/run/redis-sentine2.pid"
logfile "/data/sentinel-2.log"
sentinel announce-ip "172.200.0.6"
sentinel announce-port 26379
dir "/data"
sentinel monitor mymaster 172.200.0.2 6379 2
sentinel auth-pass mymaster 123456
acllog-max-len 128
sentinel deny-scripts-reconfig yes

3.2.3.redis-sentinel3/sentinel.conf

protected-mode no
port 26379
daemonize no
pidfile "/var/run/redis-sentine3.pid"
logfile "/data/sentinel-3.log"
sentinel announce-ip "172.200.0.7"
sentinel announce-port 26379
dir "/data"
sentinel monitor mymaster 172.200.0.2 6379 2
sentinel auth-pass mymaster 123456
acllog-max-len 128
sentinel deny-scripts-reconfig yes

4.创建docker-compose.yml文件

services:
  # redis-master
  redis-master:
    image: redis:7.4.5
    container_name: redis-master
    restart: always
    ports:
      - 6379:6379
    environment:
      TZ: "Asia/Shanghai"
    volumes:
      - ./redis/conf/redis-master.conf:/usr/local/etc/redis/redis.conf
      - ./redis/.data/redis-master/:/data:Z
    command: ["redis-server", "/usr/local/etc/redis/redis.conf"]
    networks:
      redis-network:
        ipv4_address: 172.200.0.2

  # redis slave 1
  redis-slave1:
    image: redis:7.4.5
    container_name: redis-slave1
    restart: always
    ports:
      - 6380:6379
    environment:
      TZ: "Asia/Shanghai"
    volumes:
      - ./redis/conf/redis-slave1.conf:/usr/local/etc/redis/redis.conf
      - ./redis/.data/redis-slave1/:/data:Z
    command: ["redis-server", "/usr/local/etc/redis/redis.conf"]
    networks:
      redis-network:
        ipv4_address: 172.200.0.3

  # redis slave 2
  redis-slave2:
    image: redis:7.4.5
    container_name: redis-slave2
    restart: always
    ports:
      - 6381:6379
    environment:
      TZ: "Asia/Shanghai"
    volumes:
      - ./redis/conf/redis-slave2.conf:/usr/local/etc/redis/redis.conf
      - ./redis/.data/redis-slave2/:/data:Z
    command: ["redis-server", "/usr/local/etc/redis/redis.conf"]
    networks:
      redis-network:
        ipv4_address: 172.200.0.4

  # sentinel 1
  redis-sentinel1:
    image: redis:7.4.5
    container_name: redis-sentinel1
    restart: always
    environment:
      TZ: "Asia/Shanghai"
    ports:
      - 26379:26379
    volumes:
      - ./sentinel/conf/redis-sentinel1:/usr/local/etc/redis/conf
      - ./sentinel/.data:/data:Z
    command: redis-sentinel /usr/local/etc/redis/conf/sentinel.conf
    networks:
      redis-network:
        ipv4_address: 172.200.0.5

  # sentinel 2
  redis-sentinel2:
    image: redis:7.4.5
    container_name: redis-sentinel2
    restart: always
    environment:
      TZ: "Asia/Shanghai"
    ports:
      - 26380:26379
    volumes:
      - ./sentinel/conf/redis-sentinel2:/usr/local/etc/redis/conf
      - ./sentinel/.data:/data:Z
    command: redis-sentinel /usr/local/etc/redis/conf/sentinel.conf
    networks:
      redis-network:
        ipv4_address: 172.200.0.6

  # sentinel 3
  redis-sentinel3:
    image: redis:7.4.5
    container_name: redis-sentinel3
    restart: always
    environment:
      TZ: "Asia/Shanghai"
    ports:
      - 26381:26379
    volumes:
      - ./sentinel/conf/redis-sentinel3:/usr/local/etc/redis/conf
      - ./sentinel/.data:/data:Z
    command: redis-sentinel /usr/local/etc/redis/conf/sentinel.conf
    networks:
      redis-network:
        ipv4_address: 172.200.0.7

networks:
  redis-network:
    driver: bridge
    ipam:
      config:
        - subnet: 172.200.0.0/24

4.1.启动

1. 启动容器

docker compose up -d

2. 查看容器状态

docker ps

4.2.简单测试

1. 查看Sentinel集群是否生效
   1. 进入 Sentinel 容器，使用 Sentinel API 查看监控情况：

docker exec -it redis-sentinel1 /bin/bash
redis-cli -p 26379
sentinel master mymaster # 查看redis主信息
sentinel slaves mymaster # 查看从redis信息

2. 执行上述指令，当看到以下的信息，即集群已经生效

......
31) "num-slaves"
32) "2"
33) "num-other-sentinels"
34) "2"
35) "quorum"
36) "2"
......

2. 我们来手动停止redis-master查看故障转移过程

docker stop redis-master

3. redis-sentinel1日志分析

# 当前Sentinel 实例自己无法通过心跳检测连接到主节点，所以它先将其标记为“主观下线”
1:X 24 Aug 2025 06:46:59.289 # +sdown master mymaster 172.200.0.2 6379
1:X 24 Aug 2025 06:46:59.302 * Sentinel new configuration saved on disk
# Sentinel 集群开启了一个新的纪元（Epoch），编号为 1。每次故障转移操作都会有一个唯一的、更大的 epoch 编号。所有配置更新和领导选举都基于这个 epoch，它确保了整个集群的状态一致性，让所有 Sentinel 都知道哪次故障转移是最新的
1:X 24 Aug 2025 06:46:59.304 # +new-epoch 1
# 新的 epoch 信息也被持久化到磁盘配置中
1:X 24 Aug 2025 06:46:59.312 * Sentinel new configuration saved on disk
# 当前这个 Sentinel 节点 (1:X) 正在参与一次投票。它投票给 Sentinel 节点 f4a85f4091...，支持它成为负责执行本次（epoch 1）故障转移的领导者（Leader）
1:X 24 Aug 2025 06:46:59.313 # +vote-for-leader f4a85f409178a652d59e669e127bb144dbaeb5a3 1
# 这是故障转移的关键触发点。Sentinel 节点 1:X 收到足够多的投票，现在宣布主节点 mymaster 为 “客观下线” (Objectively Down)
1:X 24 Aug 2025 06:46:59.343 # +odown master mymaster 172.200.0.2 6379 #quorum 3/2
# 这个 Sentinel 节点自己计算了一下，它不会立即发起故障转移。它设置了一个延迟时间（6分钟），在这个时间之前它自己不会尝试成为领导者去执行故障转移
1:X 24 Aug 2025 06:46:59.344 * Next failover delay: I will not start a failover before Sun Aug 24 06:52:59 2025
# 当前这个 Sentinel 节点 (1:X) 收到了来自故障转移领导者（f4a85f4091...，其运行在 172.200.0.6:26379）的广播消息。消息内容是已经完成的、针对 mymaster 的新配置
1:X 24 Aug 2025 06:46:59.669 # +config-update-from sentinel f4a85f409178a652d59e669e127bb144dbaeb5a3 172.200.0.6 26379 @ mymaster 172.200.0.2 6379
# 这是最核心的操作日志！ Sentinel 正式宣布：主节点 mymaster 已经从 172.200.0.2:6379 (旧主，即 redis-master) 切换（Failover） 到了 172.200.0.3:6379 (新主，即 redis-slave1)。
1:X 24 Aug 2025 06:46:59.671 # +switch-master mymaster 172.200.0.2 6379 172.200.0.3 6379
# 领导者 Sentinel 已经重新配置了集群的拓扑结构，并更新了所有 Sentinel 的视图。
# 它发现了从节点 172.200.0.4 (redis-slave2) 现在已经成功地复制（replicate） 新的主节点 (172.200.0.3)
1:X 24 Aug 2025 06:46:59.673 * +slave slave 172.200.0.4:6379 172.200.0.4 6379 @ mymaster 172.200.0.3 6379
# 它尝试将旧的、故障的主节点 (172.200.0.2) 也重新配置为新的主节点的一个从节点
1:X 24 Aug 2025 06:46:59.675 * +slave slave 172.200.0.2:6379 172.200.0.2 6379 @ mymaster 172.200.0.3 6379
# 当前 Sentinel 节点将接收到的新集群配置（包括新主节点、所有从节点信息）再次持久化到本地磁盘。现在它的配置文件已经完全更新
1:X 24 Aug 2025 06:46:59.684 * Sentinel new configuration saved on disk
# 在大约 30 秒后，这个 Sentinel 节点发现，那个旧的、被降级为从节点的主节点 (172.200.0.2) 依然无法连接。
1:X 24 Aug 2025 06:47:29.752 # +sdown slave 172.200.0.2:6379 172.200.0.2 6379 @ mymaster 172.200.0.3 6379

4. redis-sentinel2日志分析

# Sentinel 主观判定主节点 mymaster (172.200.0.2:6379) 已下线
1:X 24 Aug 2025 06:46:59.227 # +sdown master mymaster 172.200.0.2 6379

# Sentinel 集群达成共识，确认主节点客观下线 (quorum 2/2 表示2个Sentinel同意，达到法定人数2)
1:X 24 Aug 2025 06:46:59.282 # +odown master mymaster 172.200.0.2 6379 #quorum 2/2

# 开启一个新的纪元(epoch)，编号为1，用于标记这次故障转移操作
1:X 24 Aug 2025 06:46:59.284 # +new-epoch 1

# 开始尝试执行故障转移操作
1:X 24 Aug 2025 06:46:59.285 # +try-failover master mymaster 172.200.0.2 6379

# 将新的配置信息保存到磁盘
1:X 24 Aug 2025 06:46:59.293 * Sentinel new configuration saved on disk

# 当前Sentinel投票给ID为f4a85f409178a652d59e669e127bb144dbaeb5a3的Sentinel成为领导者，负责纪元1的故障转移
1:X 24 Aug 2025 06:46:59.295 # +vote-for-leader f4a85f409178a652d59e669e127bb144dbaeb5a3 1

# Sentinel e94cf26d... 投票给 f4a85f4091... 成为领导者
1:X 24 Aug 2025 06:46:59.314 * e94cf26d20f1d29e0f771f7af25a68a2df463e85 voted for f4a85f409178a652d59e669e127bb144dbaeb5a3 1

# Sentinel 69d264a9... 投票给 f4a85f4091... 成为领导者
1:X 24 Aug 2025 06:46:59.315 * 69d264a94c3ebf062e09d15580fb9f737888deac voted for f4a85f409178a652d59e669e127bb144dbaeb5a3 1

# f4a85f4091... 成功当选为故障转移的领导者
1:X 24 Aug 2025 06:46:59.387 # +elected-leader master mymaster 172.200.0.2 6379

# 故障转移进入选择从节点阶段，领导者正在评估哪个从节点最适合提升为新主节点
1:X 24 Aug 2025 06:46:59.388 # +failover-state-select-slave master mymaster 172.200.0.2 6379

# 已选择从节点 172.200.0.3:6379 作为新的主节点
1:X 24 Aug 2025 06:46:59.456 # +selected-slave slave 172.200.0.3:6379 172.200.0.3 6379 @ mymaster 172.200.0.2 6379

# 向选中的从节点发送 SLAVEOF NO ONE 命令，使其停止复制并成为新的主节点
1:X 24 Aug 2025 06:46:59.457 * +failover-state-send-slaveof-noone slave 172.200.0.3:6379 172.200.0.3 6379 @ mymaster 172.200.0.2 6379

# 等待从节点被提升为主节点的确认
1:X 24 Aug 2025 06:46:59.549 * +failover-state-wait-promotion slave 172.200.0.3:6379 172.200.0.3 6379 @ mymaster 172.200.0.2 6379

# 将新的配置信息保存到磁盘
1:X 24 Aug 2025 06:46:59.618 * Sentinel new configuration saved on disk

# 从节点 172.200.0.3:6379 已成功提升为新的主节点
1:X 24 Aug 2025 06:46:59.619 # +promoted-slave slave 172.200.0.3:6379 172.200.0.3 6379 @ mymaster 172.200.0.2 6379

# 故障转移进入重新配置从节点阶段，开始将其他从节点指向新的主节点
1:X 24 Aug 2025 06:46:59.620 # +failover-state-reconf-slaves master mymaster 172.200.0.2 6379

# 已向从节点 172.200.0.4:6379 发送重新配置命令，使其复制新的主节点
1:X 24 Aug 2025 06:46:59.668 * +slave-reconf-sent slave 172.200.0.4:6379 172.200.0.4 6379 @ mymaster 172.200.0.2 6379

# 主节点不再处于客观下线状态（可能是因为故障转移已开始处理）
1:X 24 Aug 2025 06:47:00.408 # -odown master mymaster 172.200.0.2 6379

# 从节点 172.200.0.4:6379 的重新配置正在进行中
1:X 24 Aug 2025 06:47:00.647 * +slave-reconf-inprog slave 172.200.0.4:6379 172.200.0.4 6379 @ mymaster 172.200.0.2 6379

# 从节点 172.200.0.4:6379 的重新配置已完成
1:X 24 Aug 2025 06:47:00.649 * +slave-reconf-done slave 172.200.0.4:6379 172.200.0.4 6379 @ mymaster 172.200.0.2 6379

# 故障转移操作完成
1:X 24 Aug 2025 06:47:00.699 # +failover-end master mymaster 172.200.0.2 6379

# 主节点已切换：从 172.200.0.2:6379 切换到 172.200.0.3:6379
1:X 24 Aug 2025 06:47:00.701 # +switch-master mymaster 172.200.0.2 6379 172.200.0.3 6379

# 发现从节点 172.200.0.4:6379 现在复制新的主节点
1:X 24 Aug 2025 06:47:00.703 * +slave slave 172.200.0.4:6379 172.200.0.4 6379 @ mymaster 172.200.0.3 6379

# 尝试将旧的主节点 172.200.0.2:6379 配置为新主节点的从节点
1:X 24 Aug 2025 06:47:00.705 * +slave slave 172.200.0.2:6379 172.200.0.2 6379 @ mymaster 172.200.0.3 6379

# 将新的配置信息保存到磁盘
1:X 24 Aug 2025 06:47:00.716 * Sentinel new configuration saved on disk

# 约30秒后，Sentinel 发现旧的主节点（现在是作为从节点）仍然无法访问，标记为主观下线
1:X 24 Aug 2025 06:47:30.747 # +sdown slave 172.200.0.2:6379 172.200.0.2 6379 @ mymaster 172.200.0.3 6379

99.常见问题

1. Failed trying to load the MASTER synchronization DB from disk: No such file or directory

redis官方配置redis.conf

其中有这么一段

# -----------------------------------------------------------------------------
#
# Replica can load the RDB it reads from the replication link directly from the
# socket, or store the RDB to a file and read that file after it was completely
# received from the master.
#
# In many cases the disk is slower than the network, and storing and loading
# the RDB file may increase replication time (and even increase the master's
# Copy on Write memory and replica buffers).
# However, parsing the RDB file directly from the socket may mean that we have
# to flush the contents of the current database before the full rdb was
# received. For this reason we have the following options:
#
# "disabled"    - Don't use diskless load (store the rdb file to the disk first)
# "on-empty-db" - Use diskless load only when it is completely safe.
# "swapdb"      - Keep current db contents in RAM while parsing the data directly
#                 from the socket. Replicas in this mode can keep serving current
#                 data set while replication is in progress, except for cases where
#                 they can't recognize master as having a data set from same
#                 replication history.
#                 Note that this requires sufficient memory, if you don't have it,
#                 you risk an OOM kill.
repl-diskless-load disabled

我们将上面的属性设置为on-empty-db即可

2. WARNING: Sentinel was not able to save the new configuration on disk!!!: Device or resource busy

如果直接使用文件映射指定sentinel.conf到容器内，这么做有可能导致哨兵没有写入配置文件的权限。

解决方案：使用文件夹映射。