Home >Database >Redis >What are the features of redis cluster?

What are the features of redis cluster?

WBOY
WBOYforward
2023-05-27 22:04:54851browse

redis3.0 cluster features

Master-slave replication (read-write separation)

The benefits of master-slave replication are 2 points:

  • 1 , Avoid redis single point of failure

  • 2. Build a read-write separation architecture to meet application scenarios with more reads and less writes

Setting the master-slave

Create 6379, 6380, and 6381 directories, and copy redis.conf in the installation directory to these three directories respectively.

Enter these three directories respectively, modify the configuration files respectively, and set the ports to: 6379 (Master), 6380 (Slave), and 6381 (Slave). At the same time, set the pidfile file to a different path.

There are two ways to set master and slave in redis:

  • 1. Set slaveof in redis.conf
    a) slaveof

  • 2. Use the redis-cli client to connect to the redis service and execute the slaveof command
    a) slaveof

Second method The master-slave replication relationship will be lost after restarting.

View master-slave information: INFO replication

Master:

role: Role

connected_slaves: Number of slaves

slave0: Slave information

Slave:

Master-slave-slave architecture

Read-only slave

By default, the redis database is read-only and cannot perform write operations when acting as a slave.
You can enable non-read-only in the configuration file: slave-read-only no

##Principle of the replication process
  • 1. When the MS relationship is established between the slave database and the master database, the SYNC command will be sent to the master database;

  • 2. After the master database receives the SYNC command, it will start saving in the background. Snapshot (RDB persistence process), and cache the write commands received during the period;

  • #3. When the snapshot is completed, the main Redis will cache the snapshot file and all cached write commands Sent to Redis;

  • 4. After receiving from Redis, the snapshot file will be loaded and the received cached command will be executed;

  • 5. Afterwards, whenever the master Redis receives a write command, it will send the command to the slave Redis to ensure data consistency;

No disk replication
Passed We learned from the previous replication process that the main library will execute the RDB process when it receives the SYNC command. Even if RDB persistence is disabled in the configuration file, it will be generated. If the disk IO performance of the server where the main library is located is poor, then this replication There will be a bottleneck in the process. Fortunately, Redis has implemented the diskless replication function in version 2.8.18 (but this function is still in the experimental stage).

Principle: Redis will not store the snapshot to the disk during replication initialization with the slave database, but will send it directly to the slave database through the network, avoiding poor IO performance question.

Enable diskless replication:

repl-diskless-sync yes

What should I do if there is a downtime in the replication architecture?

If there is a downtime in the master-slave replication architecture, you need to look at the situation:

1. Slave Redis downtime
    a)  这个相对而言比较简单,在Redis中从库重新启动后会自动加入到主从架构中,自动完成同步数据;
    b)  问题? 如果从库在断开期间,主库的变化不大,从库再次启动后,
                  主库依然会将所有的数据做RDB操作吗?还是增量更新?(从库有做持久化的前提下)

No Yes, because it has been implemented since Redis version 2.8, and incremental replication can be achieved when the master and slave recover after disconnection.

2. Main Redis downtime
This is relatively more complicated and requires the following 2 steps to complete

    i.  第一步,在从数据库中执行SLAVEOF NO ONE命令,断开主从关系并且提升为主库继续服务;
    ii. 第二步,将主库重新启动后,执行SLAVEOF命令,将其设置为其他库的从库,这时数据就能更新回来;

This manual recovery process is actually quite troublesome and error-prone. Is there a good way to solve it? Currently, Redis has improved sentinel functionality.

Sentinel

What is Sentinel
As the name suggests, the role of Sentinel is to monitor the operation of the Redis system. It is a independent process. It has two functions:

  • 1. Monitor whether the master database and slave database are running normally;

  • 2. After the master data fails Automatically convert from database to primary database;

原理

单个哨兵的架构:

多个哨兵的架构:

多个哨兵,不仅同时监控主从数据库,而且哨兵之间互为监控。

配置哨兵

启动哨兵进程首先需要创建哨兵配置文件:

vim sentinel.conf

输入内容:

sentinel monitor taotaoMaster 127.0.0.1 6379 1

说明:

    taotaoMaster:监控主数据的名称,自定义即可,可以使用大小写字母和“.-_”符号
    127.0.0.1:监控的主数据库的IP
    6379:监控的主数据库的端口
    1:最低通过票数

启动哨兵进程:

redis-sentinel ./sentinel.conf

由上图可以看到:

  • 1、 哨兵已经启动,它的id为9059917216012421e8e89a4aa02f15b75346d2b7

  • 2、 为master数据库添加了一个监控

  • 3、 发现了2个slave(由此可以看出,哨兵无需配置slave,只需要指定master,哨兵会自动发现slave)

配置多个哨兵
vim sentinel.conf

输入内容:

sentinel monitor taotaoMaster 127.0.0.1 6381 2
sentinel monitor taotaoMaster2 127.0.0.1 6381 1

集群

即使有了主从复制,每个数据库都要保存整个集群中的所有数据,容易形成木桶效应。

使用Jedis实现了分片集群,是由客户端控制哪些key数据保存到哪个数据库中,如果在水平扩容时就必须手动进行数据迁移,而且需要将整个集群停止服务,这样做非常不好的。

Redis3.0版本的一大特性就是集群(Cluster),接下来我们一起学习集群。

架构

  • (2)Redis节点之间互相连接并采用二进制协议以优化传输速度和带宽利用。同时也使用PING-PONG机制来确保连接稳定

  • (2)节点的fail是通过集群中超过半数的节点检测失效时才生效.

  • (3)客户端与redis节点直连,不需要中间proxy层.客户端不需要连接集群所有节点,连接集群中任何一个可用节点即可

  • (4)redis-cluster把所有的物理节点映射到[0-16383]slot(插槽)上,cluster 负责维护node<->slot<->value

修改配置文件

  • 1、 设置不同的端口,6379、6380、6381

  • 2、 开启集群,cluster-enabled yes

  • 3、 指定集群的配置文件,cluster-config-file “nodes-xxxx.conf”

创建集群

首先,进入redis的安装包路径下:
cd /usr/local/src/redis/redis-3.0.1/src/

执行命令:

./redis-trib.rb create --replicas 0 192.168.56.102:6379 192.168.56.102:6380 192.168.56.102:6381

–replicas 0:指定了从数据的数量为0

注意:这里不能使用127.0.0.1,否则在Jedis客户端使用时无法连接到!

redis-trib用法:

测试

什么情况??(error) MOVED 7638 127.0.0.1:6380

因为abc的hash槽信息是在6380上,现在使用redis-cli连接的6379,无法完成set操作,需要客户端跟踪重定向。

redis-cli -c

插槽的分配

通过cluster nodes命令可以查看当前集群的信息:

该信息反映出了集群中的每个节点的id、身份、连接数、插槽数等。

当我们执行set abc 123命令时,redis是如何将数据保存到集群中的呢?执行步骤:

1、  接收命令set abc 123
2、  通过key(abc)计算出插槽值,然后根据插槽值找到对应的节点。(abc的插槽值为:7638)
3、  重定向到该节点执行命令

整个Redis提供了16384个插槽,也就是说集群中的每个节点分得的插槽数总和为16384。

./redis-trib.rb 脚本实现了是将16384个插槽平均分配给了N个节点。

注意:如果插槽数有部分是没有指定到节点的,那么这部分插槽所对应的key将不能使用。

插槽和key的关系

计算key的插槽值:

key的有效部分使用CRC16算法计算出哈希值,再将哈希值对16384取余,得到插槽值。

什么是有效部分?

  • 1、 如果key中包含了{符号,且在{符号后存在}符号,并且{和}之间至少有一个字符,则有效部分是指{和}之间的部分;
    a) key={hello}_tatao的有效部分是hello

  • 2、 如果不满足上一条情况,整个key都是有效部分;
    a) key=hello_taotao的有效部分是全部

新增集群节点

再开启一个实例的端口为6382

执行脚本:

./redis-trib.rb add-node 192.168.56.102:6382 192.168.56.102:6379

已经添加成功!查看集群信息:

发现没有插槽数。

接下来需要给6382这个服务分配插槽,将6379的一部分(1000个)插槽分配给6382:

查看节点情况:

删除集群节点

想要删除集群节点中的某一个节点,需要严格执行2步:

1、 将这个节点上的所有插槽转移到其他节点上;

    a)  假设我们想要删除6380这个节点
    b)  执行脚本:./redis-trib.rb reshard 192.168.56.102:6380
    c)  选择需要转移的插槽的数量,因为3380有5128个,所以转移5128个
    d)  输入转移的节点的id,我们转移到6382节点:82ed0d63cfa6d19956dca833930977a87d6ddf7
    e)  输入插槽来源id,也就是6380的id
    f)  输入done,开始转移
    g)  查看集群信息,可以看到6380节点已经没有插槽了。

2、 使用redis-trib.rb删除节点

    a)  ./redis-trib.rb del-node 192.168.56.102:6380 4a9b8886ba5261e82597f5590fcdb49ea47c4c6c
    b)  del-node host:port node_id
    c)   
    d)  查看集群信息,可以看到已经没有6380这个节点了。

故障机制

1、  集群中的每个节点都会定期的向其它节点发送PING命令,并且通过有没有收到回复判断目标节点是否下线;
2、  集群中每一秒就会随机选择5个节点,然后选择其中最久没有响应的节点放PING命令;
3、  如果一定时间内目标节点都没有响应,那么该节点就认为目标节点疑似下线;
4、  当集群中的节点超过半数认为该目标节点疑似下线,那么该节点就会被标记为下线;
5、  当集群中的任何一个节点下线,就会导致插槽区有空档,不完整,那么该集群将不可用;
6、  如何解决上述问题?
a)  在Redis集群中可以使用主从模式实现某一个节点的高可用
b)  当该节点(master)宕机后,集群会将该节点的从数据库(slave)转变为(master)继续完成集群服务;

集群中的主从复制架构

架构:

出现故障:

4.9.3. 创建主从集群

需要启动6个redis实例,分别是:
6379(主) 6479(从)
6380(主) 6480(从)
6381(主) 6481(从)

启动redis实例:

cd 6379/ && redis-server ./redis.conf && cd ..
cd 6380/ && redis-server ./redis.conf && cd ..
cd 6381/ && redis-server ./redis.conf && cd ..
cd 6479/ && redis-server ./redis.conf && cd ..
cd 6480/ && redis-server ./redis.conf && cd ..
cd 6481/ && redis-server ./redis.conf && cd

创建集群,指定了从库数量为1,创建顺序为主库(3个)、从库(3个):

./redis-trib.rb create --replicas 1 192.168.56.102:6379 192.168.56.102:6380 192.168.56.102:6381 192.168.56.102:6479 192.168.56.102:6480 192.168.56.102:6481

The above is the detailed content of What are the features of redis cluster?. For more information, please follow other related articles on the PHP Chinese website!

Statement:
This article is reproduced at:yisu.com. If there is any infringement, please contact admin@php.cn delete