13. Redis cluster, master-slave replication, sentry

1. Redis master-slave replication

        Master-slave replication refers to copying the data of one redis server (master node-master) to other redis servers (slave node-slave). By default, each redis server is a master node, and each master node can have multiple or There are no slave nodes, but a slave node can only have one master node

1.1 The role of master-slave replication

        Data redundancy: Master-slave replication implements hot backup of data, which is a data redundancy method other than persistence

        Failure recovery: when the master node fails, the slave node can provide services

        Load balancing: on the basis of master-slave replication with read-write separation, the master node provides write services, and the slave nodes provide read services to share server load

        The cornerstone of high availability: master-slave replication is the basis for the implementation of sentinels and clusters

1.2 Common master-slave structure

        One master and one slave: used for failover of the master node from the slave node, mainly for data security and avoiding the impact of persistence on the master node

        One master and multiple slaves: mainly for scenarios with a large number of reads, the reads are shared by multiple nodes, but it affects the stability of the master node and affects the bandwidth

        Tree-like master-slave: In order to alleviate the problem of high pressure on the master node caused by one master and multiple slaves, a hierarchical structure is formed, master-slave-slave and then master-slave

1.3 Realize master-slave replication

        Copy multiple copies of redis.conf and modify the configuration

92          prot    6379
158        pidfile /var/run/redis_6379.pid
171        logfile "redis6379.log"
253        dbfilename dump6379.rdb

        Can be replaced in batches with commands

:%s/6379/6380/g  

        Start different servers and open different terminals to connect to different servers

redis-server redis6379.conf

redis-cli -p 6379

        Enter info replication to view information such as the role of the current server

        Enter the command under the server to set the current service as the slave server and who is its master server

salveof main server ip port number

2. redis sentry

        When the main server is down, in order to maintain the normal implementation of the function, we need a sentry to select a slave server to become the main server; (the sentinel is also a separate redis server, connected to the main server, and connected to the slave through the main)

        In order to prevent a single sentry from misjudgment of downtime, we can also use multiple sentries (must be an odd number), set a voting parameter, and only when this number of sentries think that the master server is down, will the master-slave switch

        Sentinel's sentinel.conf should be set as follows (remove comments)

protected-mode no # Turn off protected mode, convenient for testing
port 26379 # Sentinel port
sentinel monitor mymaster 192.168.41.226 6379 1        

# 192.168.41.226: host ip 6379: port 1: failover when at least a few sentinels think the host is offline

         Enter redis-sentinel sentinel6379.conf to start the sentinel

3. Redis cluster

        Redis-cluster clusters are supported in versions after redis 3.0, which means that multiple redis master servers are connected together to form a cluster.

This is to solve the problem that although the sentinel mode can meet most requirements, it cannot handle a large number of write requests;

         This is also a classic P2P (decentralization)

3.1 Implementation of redis cluster

• Create a redis-cluster directory under /usr/local/redis

• Copy redis6379 to redis-cluster and name it redis8001.conf

• 92 Modify the port number in redis8001.conf

• 158 modify pidfile

• 171 Modify log files

• 253 Modify rdb file

• 699 Enable aof and set it to yes

• 703 specify aof file name

• 832 Enable the cluster and set it to yes

• 840 Specify nodes configuration file

        In order to facilitate the startup of multiple redis servers, we can write a startup script, create a start_all.sh file under /user/local/redis, and add it to the file

redis-server redis-cluster/redis8001.conf
redis-server redis-cluster/redis8002.conf
redis-server redis-cluster/redis8003.conf
redis-server redis-cluster/redis8004.conf
redis-server redis-cluster/redis8005.conf
redis-server redis-cluster/redis8006.conf
ps -ef|grep redis

        Save and exit, add execution permission to the start_all.sh file

chmod +x start_all.sh

         start redis

./start_all

        create cluster

redis-cli --cluster create 192.168.74.130:8001 192.168.74.130:8002 192.168.74.130:8003 192.168.74.130:8004 192.168.74.130:8005 192.168.74.130:8006 --cluster-replicas 1 

        Enter yes during creation 

        To connect to the redis cluster, use -c

redis-cli -c -p 8001

        Create a cluster shutdown script

#!/bin/bash
PORT=8001
ENDPORT=8006
while [ $((PORT <= ENDPORT)) != "0" ]; do
   echo "Stopping Redis $PORT"
   redis-cli -p $PORT shutdown
   PORT=$((PORT+1))
done
echo "done"
exit 0 

        Add execute permission 

chmod +x stop_all.sh