redis-集群
1、基础
1.1、非集群的问题点
容量不够,redis如何进行扩容?
并发写操作, redis如何分摊?
另外,主从模式,主机宕机,导致ip地址发生变化,应用程序中配置需要修改对应的主机地址、端口等信息。
之前通过代理主机来解决,但是redis3.0中提供了解决方案。就是无中心化集群配置。
1.2、什么是redis集群
Redis 集群实现了对Redis的水平扩容,即启动N个redis节点,将整个数据库分布存储在这N个节点中,每个节点存储总数据的1/N。
Redis 集群通过分区(partition)来提供一定程度的可用性(availability): 即使集群中有一部分节点失效或者无法进行通讯, 集群也可以继续处理命令请求。
2、环境准备
2.1、主机准备
redis-01 192.168.10.19
redis-02 192.168.10.20
2.2、实例规划
每台主机3个实例,端口分别:6379、6380、6381
3、配置集群
3.1、主配置文件redis.conf
]# cat /usr/local/redis-6.2.5/etc/redis.conf | grep -Ev '^$|^#' bind * port 6379 daemonize yes
...
3.2、redis-01配置
3.2.1、6379.conf
redis-01 ~]# cat /usr/local/redis-6.2.5/etc/6379.conf include /usr/local/redis-6.2.5/etc/redis.conf logfile "/usr/local/redis-6.2.5/log/6379.log" dbfilename "dump_6379.rdb" pidfile "/var/run/redis_6379.pid" dir "/usr/local/redis-6.2.5/data" daemonize yes port 6379 cluster-enabled yes cluster-config-file nodes-6379.conf cluster-node-timeout 15000
3.2.2、6380.conf
redis-01 ~]# cat /usr/local/redis-6.2.5/etc/6380.conf include /usr/local/redis-6.2.5/etc/redis.conf logfile "/usr/local/redis-6.2.5/log/6380.log" dbfilename "dump_6380.rdb" pidfile "/var/run/redis_6380.pid" dir "/usr/local/redis-6.2.5/data" daemonize yes port 6380 cluster-enabled yes cluster-config-file nodes-6380.conf cluster-node-timeout 15000
3.2.3、6381.conf
redis-01 ~]# cat /usr/local/redis-6.2.5/etc/6381.conf include /usr/local/redis-6.2.5/etc/redis.conf logfile "/usr/local/redis-6.2.5/log/6381.log" dbfilename "dump_6381.rdb" pidfile "/var/run/redis_6381.pid" dir "/usr/local/redis-6.2.5/data" daemonize yes port 6381 cluster-enabled yes cluster-config-file nodes-6381.conf cluster-node-timeout 15000
3.3、redis-02配置
3.3.1、6379.conf
redis-02 ~]# cat /usr/local/redis-6.2.5/etc/6379.conf include /usr/local/redis-6.2.5/etc/redis.conf logfile "/usr/local/redis-6.2.5/log/6379.log" dbfilename "dump_6379.rdb" pidfile "/var/run/redis_6379.pid" dir "/usr/local/redis-6.2.5/data" daemonize yes port 6379 cluster-enabled yes cluster-config-file nodes-6379.conf cluster-node-timeout 15000
3.3.2、6380.conf
redis-02 ~]# cat /usr/local/redis-6.2.5/etc/6380.conf include /usr/local/redis-6.2.5/etc/redis.conf logfile "/usr/local/redis-6.2.5/log/6380.log" dbfilename "dump_6380.rdb" pidfile "/var/run/redis_6380.pid" dir "/usr/local/redis-6.2.5/data" daemonize yes port 6380 cluster-enabled yes cluster-config-file nodes-6380.conf cluster-node-timeout 15000
3.3.3、6381.conf
redis-02 ~]# cat /usr/local/redis-6.2.5/etc/6381.conf include /usr/local/redis-6.2.5/etc/redis.conf logfile "/usr/local/redis-6.2.5/log/6381.log" dbfilename "dump_6381.rdb" pidfile "/var/run/redis_6381.pid" dir "/usr/local/redis-6.2.5/data" daemonize yes port 6381 cluster-enabled yes cluster-config-file nodes-6381.conf cluster-node-timeout 15000
3.4、启动redis服务、检查端口状态、查询配置文件生成
3.4.1、redis-01
# redis-01 redis-server /usr/local/redis-6.2.5/etc/6379.conf redis-server /usr/local/redis-6.2.5/etc/6380.conf redis-server /usr/local/redis-6.2.5/etc/6381.conf redis-01 ~]# netstat -tunlp | grep redis tcp 0 0 0.0.0.0:6379 0.0.0.0:* LISTEN 29696/redis-server tcp 0 0 0.0.0.0:6380 0.0.0.0:* LISTEN 29702/redis-server tcp 0 0 0.0.0.0:6381 0.0.0.0:* LISTEN 29708/redis-server tcp 0 0 0.0.0.0:16379 0.0.0.0:* LISTEN 29696/redis-server tcp 0 0 0.0.0.0:16380 0.0.0.0:* LISTEN 29702/redis-server tcp 0 0 0.0.0.0:16381 0.0.0.0:* LISTEN 29708/redis-server redis-01 ~]# ll /usr/local/redis-6.2.5/data/ 总用量 12 -rw-r--r--. 1 root root 126 5月 16 20:52 nodes-6379.conf -rw-r--r--. 1 root root 126 5月 16 20:52 nodes-6380.conf -rw-r--r--. 1 root root 126 5月 16 20:52 nodes-6381.conf
3.4.2、redis-02
# redis-02 redis-server /usr/local/redis-6.2.5/etc/6379.conf redis-server /usr/local/redis-6.2.5/etc/6380.conf redis-server /usr/local/redis-6.2.5/etc/6381.conf redis-02 ~]# netstat -tunlp | grep redis tcp 0 0 0.0.0.0:6379 0.0.0.0:* LISTEN 27761/redis-server tcp 0 0 0.0.0.0:6380 0.0.0.0:* LISTEN 27767/redis-server tcp 0 0 0.0.0.0:6381 0.0.0.0:* LISTEN 27773/redis-server tcp 0 0 0.0.0.0:16379 0.0.0.0:* LISTEN 27761/redis-server tcp 0 0 0.0.0.0:16380 0.0.0.0:* LISTEN 27767/redis-server tcp 0 0 0.0.0.0:16381 0.0.0.0:* LISTEN 27773/redis-server redis-02 ~]# ll /usr/local/redis-6.2.5/data/ 总用量 12 -rw-r--r--. 1 root root 114 5月 16 20:53 nodes-6379.conf -rw-r--r--. 1 root root 114 5月 16 20:53 nodes-6380.conf -rw-r--r--. 1 root root 114 5月 16 20:53 nodes-6381.conf
4、将六个节点合成一个集群
4.1、前提条件
组合之前,请确保所有redis实例启动后,nodes-xxxx.conf文件都生成正常。
4.2、合成集群
4.2.1、报错的解决
[ERR] Node 192.168.10.19:6380 is not empty. Either the node already knows other nodes (check with CLUSTER NODES) or contains some key in database 0. 解决方法 将所有的节点都flushall一下或者重置集群: redis-cli -p 6379 flushall redis-cli -p 6380 flushall redis-cli -p 6381 flushall redis-cli -p 6380 CLUSTER reset redis-cli -p 6379 CLUSTER reset redis-cli -p 6381 CLUSTER reset
4.2.2、合并成为一个集群
redis-02 ~]# redis-cli --cluster create --cluster-replicas 1 192.168.10.19:6379 192.168.10.19:6380 192.168.10.19:6381 192.168.10.20:6379 192.168.10.20:6380 192.168.10.20:6381 >>> Performing hash slots allocation on 6 nodes... Master[0] -> Slots 0 - 5460 Master[1] -> Slots 5461 - 10922 Master[2] -> Slots 10923 - 16383 Adding replica 192.168.10.20:6381 to 192.168.10.19:6379 Adding replica 192.168.10.19:6381 to 192.168.10.20:6379 Adding replica 192.168.10.20:6380 to 192.168.10.19:6380 M: 124c4d088133786aaebe5820f8d8fc093dd3289e 192.168.10.19:6379 slots:[0-5460] (5461 slots) master M: 9a7befa99f0a574cd18b17d21c725c69cf43be23 192.168.10.19:6380 slots:[10923-16383] (5461 slots) master S: 24813b313e72693904292fea9acfefeae46e6b71 192.168.10.19:6381 replicates 5b38b6e5f726d78f75780619ac383d53b08afa84 M: 5b38b6e5f726d78f75780619ac383d53b08afa84 192.168.10.20:6379 slots:[5461-10922] (5462 slots) master S: c3ddf00c20e389802d7d0ffb0747bed1a6c6a976 192.168.10.20:6380 replicates 9a7befa99f0a574cd18b17d21c725c69cf43be23 S: 794078cb7c59e6d73aa3e2c58b55487ee064cafe 192.168.10.20:6381 replicates 124c4d088133786aaebe5820f8d8fc093dd3289e Can I set the above configuration? (type 'yes' to accept): yes >>> Nodes configuration updated >>> Assign a different config epoch to each node >>> Sending CLUSTER MEET messages to join the cluster Waiting for the cluster to join . >>> Performing Cluster Check (using node 192.168.10.19:6379) M: 124c4d088133786aaebe5820f8d8fc093dd3289e 192.168.10.19:6379 slots:[0-5460] (5461 slots) master 1 additional replica(s) S: 24813b313e72693904292fea9acfefeae46e6b71 192.168.10.19:6381 slots: (0 slots) slave replicates 5b38b6e5f726d78f75780619ac383d53b08afa84 M: 5b38b6e5f726d78f75780619ac383d53b08afa84 192.168.10.20:6379 slots:[5461-10922] (5462 slots) master 1 additional replica(s) S: c3ddf00c20e389802d7d0ffb0747bed1a6c6a976 192.168.10.20:6380 slots: (0 slots) slave replicates 9a7befa99f0a574cd18b17d21c725c69cf43be23 S: 794078cb7c59e6d73aa3e2c58b55487ee064cafe 192.168.10.20:6381 slots: (0 slots) slave replicates 124c4d088133786aaebe5820f8d8fc093dd3289e M: 9a7befa99f0a574cd18b17d21c725c69cf43be23 192.168.10.19:6380 slots:[10923-16383] (5461 slots) master 1 additional replica(s) [OK] All nodes agree about slots configuration. >>> Check for open slots... >>> Check slots coverage... [OK] All 16384 slots covered. # 说明集群成功
属性参数解析: 一个集群至少要有三个主节点。 选项 --cluster-replicas 1 表示我们希望为集群中的每个主节点创建一个从节点。 分配原则尽量保证每个主数据库运行在不同的IP地址,每个从库和主库不在一个IP地址上。
4.2.3、查询节点状态
]# redis-cli cluster nodes c3ddf00c20e389802d7d0ffb0747bed1a6c6a976 192.168.10.20:6380@16380 slave 9a7befa99f0a574cd18b17d21c725c69cf43be23 0 1684248605230 2 connected 9a7befa99f0a574cd18b17d21c725c69cf43be23 192.168.10.19:6380@16380 master - 0 1684248608248 2 connected 10923-16383 794078cb7c59e6d73aa3e2c58b55487ee064cafe 192.168.10.20:6381@16381 slave 124c4d088133786aaebe5820f8d8fc093dd3289e 0 1684248607241 1 connected 5b38b6e5f726d78f75780619ac383d53b08afa84 192.168.10.20:6379@16379 myself,master - 0 1684248605000 4 connected 5461-10922 24813b313e72693904292fea9acfefeae46e6b71 192.168.10.19:6381@16381 slave 5b38b6e5f726d78f75780619ac383d53b08afa84 0 1684248606237 4 connected 124c4d088133786aaebe5820f8d8fc093dd3289e 192.168.10.19:6379@16379 master - 0 1684248605000 1 connected 0-5460
4.3、写入数据测试
4.3.1、-c 采用集群策略连接,设置数据会自动切换到相应的写主机
# 设置值 redis-02 ~]# redis-cli -c 127.0.0.1:6379> set name cyc OK 127.0.0.1:6379> get name "cyc" # 其它节点获取值 redis-01 ~]# redis-cli -c 127.0.0.1:6379> get name -> Redirected to slot [5798] located at 192.168.10.20:6379 "cyc"
4.4、什么是slots
[OK] All nodes agree about slots configuration. >>> Check for open slots... >>> Check slots coverage... [OK] All 16384 slots covered. 一个 Redis 集群包含 16384 个插槽(hash slot), 数据库中的每个键都属于这 16384 个插槽的其中一个, 集群使用公式 CRC16(key) % 16384 来计算键 key 属于哪个槽, 其中 CRC16(key) 语句用于计算键 key 的 CRC16 校验和 。 集群中的每个节点负责处理一部分插槽。 举个例子, 如果一个集群可以有主节点, 其中: 节点 A 负责处理 0 号至 5460 号插槽。 节点 B 负责处理 5461 号至 10922 号插槽。 节点 C 负责处理 10923 号至 16383 号插槽。
4.5、集群查询
4.5.1、在集群中录入值
在redis-cli每次录入、查询键值,redis都会计算出该key应该送往的插槽,如果不是该客户端对应服务器的插槽,redis会报错,并告知应前往的redis实例地址和端口。 redis-cli客户端提供了 –c 参数实现自动重定向。 如 redis-cli -c –p 6379 登入后,再录入、查询键值对可以自动重定向。 不在一个slot下的键值,是不能使用mget,mset等多键操作。 127.0.0.1:6379> mset k1 v1 k2 v2 k3 v3 (error) CROSSSLOT Keys in request don't hash to the same slot 可以通过{}来定义组的概念,从而使key中{}内相同内容的键值对放到一个slot中去。 127.0.0.1:6379> mset k1{cust} v1 k2{cust} v2 k3{cust} v3 -> Redirected to slot [4847] located at 192.168.10.19:6379 OK
4.5.2、查询集群中的值
192.168.10.19:6379> CLUSTER KEYSLOT cust (integer) 4847 192.168.10.19:6379> CLUSTER COUNTKEYSINSLOT 4847 (integer) 3 192.168.10.19:6379> CLUSTER GETKEYSINSLOT 4847 10 1) "k1{cust}" 2) "k2{cust}" 3) "k3{cust}"
4.6、故障恢复
4.6.1、停掉其中一个master
]# redis-cli -c 127.0.0.1:6379> shutdown
4.6.2、查询日志情况
]# tail -f /usr/local/redis-6.2.5/log/6380.log 29702:M 16 May 2023 22:48:04.929 * Background saving started by pid 30056 30056:C 16 May 2023 22:48:04.933 * DB saved on disk 30056:C 16 May 2023 22:48:04.933 * RDB: 4 MB of memory used by copy-on-write 29702:M 16 May 2023 22:48:05.005 * Background saving terminated with success 29702:M 16 May 2023 22:48:05.006 * Synchronization with replica 192.168.10.20:6380 succeeded 29702:M 16 May 2023 22:48:07.924 # Cluster state changed: ok 29702:M 16 May 2023 23:29:01.296 * FAIL message received from 24813b313e72693904292fea9acfefeae46e6b71 about 124c4d088133786aaebe5820f8d8fc093dd3289e 29702:M 16 May 2023 23:29:01.296 # Cluster state changed: fail 29702:M 16 May 2023 23:29:02.005 # Failover auth granted to 794078cb7c59e6d73aa3e2c58b55487ee064cafe for epoch 7 29702:M 16 May 2023 23:29:02.010 # Cluster state changed: ok
4.6.3、查询集群节点
127.0.0.1:6379> CLUSTER nodes c3ddf00c20e389802d7d0ffb0747bed1a6c6a976 192.168.10.20:6380@16380 slave 9a7befa99f0a574cd18b17d21c725c69cf43be23 0 1684251060731 2 connected 9a7befa99f0a574cd18b17d21c725c69cf43be23 192.168.10.19:6380@16380 master - 0 1684251061742 2 connected 10923-16383 794078cb7c59e6d73aa3e2c58b55487ee064cafe 192.168.10.20:6381@16381 master - 0 1684251059718 7 connected 0-5460 # 这个是就是从节点接管为master 5b38b6e5f726d78f75780619ac383d53b08afa84 192.168.10.20:6379@16379 myself,master - 0 1684251059000 4 connected 5461-10922 24813b313e72693904292fea9acfefeae46e6b71 192.168.10.19:6381@16381 slave 5b38b6e5f726d78f75780619ac383d53b08afa84 0 1684251058000 4 connected 124c4d088133786aaebe5820f8d8fc093dd3289e 192.168.10.19:6379@16379 master,fail - 1684250926110 1684250921000 1 disconnected # 已经失败,停用,master还是有3个,说明已经自动从变成主
4.6.4、恢复下掉的节点
redis-server /usr/local/redis-6.2.5/etc/6379.conf
4.6.5、分析集群节点状态【恢复后的节点,将一直变成slave】
127.0.0.1:6379> CLUSTER nodes c3ddf00c20e389802d7d0ffb0747bed1a6c6a976 192.168.10.20:6380@16380 slave 9a7befa99f0a574cd18b17d21c725c69cf43be23 0 1684251327352 2 connected 9a7befa99f0a574cd18b17d21c725c69cf43be23 192.168.10.19:6380@16380 master - 0 1684251326000 2 connected 10923-16383 794078cb7c59e6d73aa3e2c58b55487ee064cafe 192.168.10.20:6381@16381 master - 0 1684251328358 7 connected 0-5460 5b38b6e5f726d78f75780619ac383d53b08afa84 192.168.10.20:6379@16379 myself,master - 0 1684251327000 4 connected 5461-10922 24813b313e72693904292fea9acfefeae46e6b71 192.168.10.19:6381@16381 slave 5b38b6e5f726d78f75780619ac383d53b08afa84 0 1684251326000 4 connected 124c4d088133786aaebe5820f8d8fc093dd3289e 192.168.10.19:6379@16379 slave 794078cb7c59e6d73aa3e2c58b55487ee064cafe 0 1684251326000 7 connected # 这个就是恢复的节点
4.7、如果所有某一段插槽的主从节点都宕掉,redis服务是否还能继续?
如果某一段插槽的主从都挂掉,而cluster-require-full-coverage 为yes ,那么 ,整个集群都挂掉 如果某一段插槽的主从都挂掉,而cluster-require-full-coverage 为no ,那么,该插槽数据全都不能使用,也无法存储。 redis.conf中的参数 cluster-require-full-coverage
5、集群优点与缺点
5.1、优点
实现扩容
分摊压力
无中心配置相对简单
5.2、缺点
多键操作是不被支持的
多键的Redis事务是不被支持的。lua脚本不被支持
由于集群方案出现较晚,很多公司已经采用了其他的集群方案,而代理或者客户端分片的方案想要迁移至redis cluster,需要整体迁移而不是逐步过渡,复杂度较大。