当前位置：网站首页>Redis sentinel cluster working principle and architecture deployment ＃ yyds dry goods inventory ＃

Redis sentinel cluster working principle and architecture deployment ＃ yyds dry goods inventory ＃

2022-07-02 20:23:00 【Jiangxiaolong's technology blog】

Redis Working principle and architecture deployment of sentry cluster

Redis Working principle and architecture deployment of sentry cluster

1.redis Sentinel mode principle
2. build redis The sentry cluster

2.1. Environmental preparation
2.2. Deploy on all machines redis
2.3. Three stations redis Deployment completed
2.4. To configure redis Master-slave
2.5. Deploy sentinel process sentinel
2.6. Start the sentry to observe the changes in the configuration file
2.7. Simulate the main library failure and verify whether the application is available
2.8. The main library hangs up the changes in the configuration files of other nodes

1.redis Sentinel mode principle

redis Lack of master-slave replication ： When the main database goes down ,slave You can't become the main library by yourself , Write data , Every time, it needs to be manually configured to change from the library to the main library to write data , After the master database is repaired, it is also necessary to manually configure the import slave database host before configuring the master-slave replication

redis The advantages of sentinel mode ： redis Sentinels are built on master-slave , There is a monitoring function , Monitor whether the main database is abnormal , When the main database is abnormal, it will automatically send a slave Change to main library , It eliminates human configuration

redis Sentinel mode principle ： Sentinel mode is based on master-slave replication , Will be in every redis Open a... On the node sentinel monitor , Every sentinel Each process has its own port number and IP, all sentinel Share cluster information , All in the cluster sentinel Can communicate , When the main database goes down , On the main library sentinel It will send messages to other nodes in the cluster sentinel Send a message saying that the main library is down , Need to vote from the library master, At this time, on each node sentinel Compare yourself redis Configured ID Number , such as slave1 Of ID Large ,slave1 Will be elected master As main storehouse , When the failed primary database rejoins the cluster , On the main library sentinel Will be sent to other nodes sentinel Ask who is the main library , At this time slave1 Upper sentinel Will tell the on the main library sentinel say slave1 It's the main warehouse , Rejoining the main database of the cluster will find slave1 Synchronous data , If the rejoined master database wants to become the master database again , You only need to execute the authorization command to become the main library again .

Program configuration during master-slave replication redis When the address is written, the address of the main database is written , You need to manually modify the application every time the main library goes down

With sentinel mode , Configuration in program code is not redis Address , But the addresses of all sentinels configured , Form an address pool , Even if one of the sentinels in the cluster breaks down , There are two other sentinels , Every time you need to find redis When writing data , The program will first find the sentinel process , Information sharing between sentinels , Will immediately tell the program who is the main library , Then the program gets what the sentry told it redis Main database address , Will go to find the master inventory data , So even if the main library is broken , There is no need to modify the program code

The sentry's configuration file after starting the sentry service , Try not to modify , Because the Sentry will automatically add configuration

The recommended number of sentinel clusters is odd , such as 3/5/7

With sentinels , When the main library hangs up , The sentry elected a new Treasury , The configuration file will be automatically modified to the address of the latest main library

Sentinel election rules ： The first judgment slave-priority Weight priority , Whoever is taller is elected master Main library , If all agree , Compare the of each node for so long id, Who's big, who's elected

Comparison between sentinel mode architecture and master-slave replication architecture
Redis Working principle and architecture deployment of sentry cluster #yyds Dry inventory #_ Master slave copy

2. build redis The sentry cluster

2.1. Environmental preparation

IP	service
192.168.81.210	redis+sentinel+master
192.168.81.220	redis+sentinel+slave
192.168.81.230	redis+sentinel+slave

Steps for configuring sentinel clusters ：

1. At all nodes redis

2. Configure master-slave replication , One master and two slaves

3. Configure... On all nodes sentinel, start-up sentinel after , The configuration file will be automatically added

2.2. Deploy on all machines redis

192.168.81.210 To configure

1. establish redis Deployment path 
[[email protected] ~]# mkdir -p /data/redis_cluster/redis_6379/{conf,pid,logs,data}
    
2. download redis    
[[email protected] ~]# mkdir /data/soft
[[email protected] ~]# cd /data/soft
[[email protected] /data/soft]# wget https://repo.huaweicloud.com/redis/redis-3.2.9.tar.gz

3. Easy to install redis
[[email protected] /data/soft]# tar xf redis-3.2.9.tar.gz -C /data/redis_cluster/
[[email protected] /data/soft]# cd /data/redis_cluster/
[[email protected] /data/redis_cluster]# ln -s redis-3.2.9/ redis
[[email protected] /data/redis_cluster]# cd redis/src
[[email protected] /data/redis_cluster/redis]# make && make install

4. Prepare the configuration file 
[[email protected] ~]# vim /data/redis_cluster/redis_6379/conf/redis_6379.conf 
daemonize yes
bind 192.168.81.210 127.0.0.1
port 6379
pidfile /data/redis_cluster/redis_6379/pid/redis_6379.pid
logfile /data/redis_cluster/redis_6379/logs/redis_6379.log
databases 16
dbfilename redis_6379.rdb
dir /data/redis_cluster/redis_6379/data/
save 900 1
save 300 100
save 60 10000

5. start-up redis
[[email protected] ~]# redis-server /data/redis_cluster/redis_6379/conf/redis_6379.conf 

     
      1.
      2.
      3.
      4.
      5.
      6.
      7.
      8.
      9.
      10.
      11.
      12.
      13.
      14.
      15.
      16.
      17.
      18.
      19.
      20.
      21.
      22.
      23.
      24.
      25.
      26.
      27.
      28.
      29.
      30.
      31.

192.168.81.220 To configure

because redis-1 A set has been deployed redis, We can copy it directly

1. Use rsync take redis-1 Of redis Copy the directory and you 
[[email protected] ~]# rsync -avz /data [email protected]:/

2. View the copied directory file 
[[email protected] ~]# ls  /data/redis_cluster/
redis  redis-3.2.9  redis_6379
[[email protected] ~]# ls  /data/redis_cluster/redis_6379/
conf  data  logs  pid

3. Compilation and installation redis, Enable the system to use redis command 
 Direct execution make install that will do , Because the compilation step is redis-1 It has been done. 
[[email protected] ~]# cd /data/redis_cluster/redis-3.2.9/
[[email protected] /data/redis_cluster/redis-3.2.9]# make install

4. modify redis The configuration file 
[[email protected] ~]# vim /data/redis_cluster/redis_6379/conf/redis_6379.conf 
bind 192.168.81.220 127.0.0.1

5. start-up redis
[[email protected] ~]# redis-server /data/redis_cluster/redis_6379/conf/redis_6379.conf


     
      1.
      2.
      3.
      4.
      5.
      6.
      7.
      8.
      9.
      10.
      11.
      12.
      13.
      14.
      15.
      16.
      17.
      18.
      19.
      20.
      21.

192.168.81.230 To configure

because redis-1 A set has been deployed redis, We can copy it directly

1. Use rsync take redis-1 Of redis Copy the directory and you 
[[email protected] ~]# rsync -avz /data [email protected]:/

2. View the copied directory file 
[[email protected] ~]# ls  /data/redis_cluster/
redis  redis-3.2.9  redis_6379
[[email protected] ~]# ls  /data/redis_cluster/redis_6379/
conf  data  logs  pid

3. Compilation and installation redis, Enable the system to use redis command 
 Direct execution make install that will do , Because the compilation step is redis-1 It has been done. 
[[email protected] ~]# cd /data/redis_cluster/redis-3.2.9/
[[email protected] /data/redis_cluster/redis-3.2.9]# make install

4. modify redis The configuration file 
[[email protected] ~]# vim /data/redis_cluster/redis_6379/conf/redis_6379.conf 
bind 192.168.81.230 127.0.0.1

5. start-up redis
[[email protected] ~]# redis-server /data/redis_cluster/redis_6379/conf/redis_6379.conf

     
      1.
      2.
      3.
      4.
      5.
      6.
      7.
      8.
      9.
      10.
      11.
      12.
      13.
      14.
      15.
      16.
      17.
      18.
      19.
      20.

2.3. Three stations redis Deployment completed

[[email protected] ~]# ps aux | grep redis
root      21860  0.1  0.5 139020  9740 ?        Ssl  09:36   0:16 redis-server 192.168.81.210:6379
root      25296  0.0  0.0 112724   984 pts/0    S+   13:15   0:00 grep --color=auto redis

[[email protected] ~]# ssh 192.168.81.220 "ps aux | grep redis"
root      47658  0.1  0.5 141068 10780 ?        Ssl  1 month 28   1:24 redis-server 192.168.81.220:6379
root      63254  0.0  0.0 113176  1588 ?        Ss   13:15   0:00 bash -c ps aux | grep redis
root      63271  0.0  0.0 112724   968 ?        S    13:15   0:00 grep redis

[[email protected] ~]# ssh 192.168.81.230 "ps aux | grep redis"
root      56584  0.1  0.7 136972  7548 ?        Ssl  13:13   0:00 redis-server 192.168.81.230:6379
root      56644  0.0  0.1 113176  1588 ?        Ss   13:15   0:00 bash -c ps aux | grep redis
root      56661  0.0  0.0 112724   968 ?        S    13:15   0:00 grep redis


     
      1.
      2.
      3.
      4.
      5.
      6.
      7.
      8.
      9.
      10.
      11.
      12.
      13.
      14.

2.4. To configure redis Master-slave

On both sets slave Synchronize master database configuration on

1. Configure master-slave replication 
[[email protected] ~]# redis-cli 
127.0.0.1:6379> SLAVEOF 192.168.81.210 6379
OK  

[[email protected] ~]# redis-cli 
127.0.0.1:6379> SLAVEOF 192.168.81.220 6379
OK


2. Create a new... In the main library key
127.0.0.1:6379> set name jiangxl
OK


3. Check whether to copy from the library 
[[email protected] ~]# redis-cli 
127.0.0.1:6379> get name
"jiangxl"
[[email protected] ~]# redis-cli 
127.0.0.1:6379> get name
"jiangxl"


     
      1.
      2.
      3.
      4.
      5.
      6.
      7.
      8.
      9.
      10.
      11.
      12.
      13.
      14.
      15.
      16.
      17.
      18.
      19.
      20.
      21.
      22.
      23.

2.5. Deploy sentinel process sentinel

Profile explanation

sentinel monitor mymaster 192.168.81.210 6379 2			// Set master node information ,mymaster Is the primary node alias , Just pick a name , Then fill in the of the master node ip Address ,2 Indicates that when the primary node hangs up , Yes 2 individual sentinel A new... Will not be elected until it is agreed master, A group of sentinels , Write the names the same 

sentinel down-after-milliseconds mymaster 3000			// How many seconds does the main library go down , Switch from the library , Because sometimes because of network fluctuations , If you switch between master and slave as soon as the master library goes down , that redis May have been switching 
sentinel parallel-syncs mymaster 1					// Allow several nodes to synchronize data to the master database at the same time 
sentinel failover-timeout mymaster 18000			// Failover timeout , When the slave database synchronizes with the master database rdb file , It will be considered as timeout if the synchronization is not completed for a long time 

     
      1.
      2.
      3.
      4.
      5.

Three stations redis Servers should be configured as follows , The... In the configuration file has been bind Written as a system variable , In coordination cat Write to file , Therefore, you can directly execute the following commands

1. Create sentinel service configuration path 
mkdir -p /data/redis_cluster/redis_26379/{conf,data,pid,logs} 

2. Write sentry profile 
cat > /data/redis_cluster/redis_26379/conf/redis_26379.conf <<EOF
bind $(ifconfig | awk 'NR==2{print $2}')
port 26379
daemonize yes
logfile /data/redis_cluster/redis_26379/logs/redis_26379.log
dir /data/redis_cluster/redis_26379/data
sentinel monitor mymaster 192.168.81.210 6379 2
sentinel down-after-milliseconds mymaster 3000
sentinel parallel-syncs mymaster 1
sentinel failover-timeout mymaster 18000
EOF


     
      1.
      2.
      3.
      4.
      5.
      6.
      7.
      8.
      9.
      10.
      11.
      12.
      13.
      14.
      15.
      16.

After configuration, remember to check the configuration file bind Whether a column belongs to the respective host ip Address
Redis Working principle and architecture deployment of sentry cluster #yyds Dry inventory #_ The configuration file _02

2.6. Start the sentry to observe the changes in the configuration file

All three machines operate like this. Start the sentry

Observe the changes of configuration file before and after sentry startup

Before startup
Redis Working principle and architecture deployment of sentry cluster #yyds Dry inventory #_ Master slave copy _03

After starting

Redis Working principle and architecture deployment of sentry cluster #yyds Dry inventory #_ Master slave copy _04

Each sentinel host automatically adds one myid Configuration of , This is when the main database hangs up , The basis for the sentinel election , Judge whose myid Whoever is older will be elected as the main library

Each sentinel host also automatically adds sentinel known-sentinel This configuration , In this configuration, each sentinel will record the data of other nodes in the cluster id Number , In this way, information sharing can be realized , Even when the application asks the sentinel process who is the main library , At this time, because each sentinel process has information about other nodes , So you can tell the application who is the main library

Redis Working principle and architecture deployment of sentry cluster #yyds Dry inventory #_ Master slave copy _05

2.7. Simulate the main library failure and verify whether the application is available

After the sentinel is deployed , Each node has cluster information sharing , When the main library hangs up , Sentinel process confirms that the main library is offline , The Sentinels according to their id Size elects a new main library , Take over the work of the main warehouse , Ensure that the application is not affected , When the main library is repaired , First synchronize the data of the current master database by raising the right , In making itself the main library

1. Close the main library redis service ,reids Normally shut down ,sentinel direct kill
[[email protected] ~]# redis-cli shutdown
[[email protected] ~]# pkill redis

4. Check the configuration file to see whose myid Big 
redis-2 Of myid The larger 
[[email protected] ~]# grep 'known-sentinel' /data/redis_cluster/redis_26379/conf/redis_26379.conf 
sentinel known-sentinel mymaster 192.168.81.220 26379 df44bb3e9fdf8c635628b1ae724b2db7d3ef144c
sentinel known-sentinel mymaster 192.168.81.230 26379 de282d14bb0a79df90603eb92243cd1f362dd46d

2. test redis-2 Whether data can be written 
 Can write data ,redis-2 Selected as primary Library 
[[email protected] ~]# redis-cli -h 192.168.81.220 set gzzy_test guzhangzhuanyi
OK
[[email protected] ~]#  redis-cli -h 192.168.81.220 config get slaveof
1) "slaveof"
2) ""

4. test redis-3 Whether data can be written 
 Failed to write data , And the synchronization is redis-2 The data of , therefore redis-2 Give priority to the library 
[[email protected] ~]# redis-cli -h 192.168.81.230 set kkkk111 vvv
(error) READONLY You can't write against a read only slave.
[[email protected] ~]# redis-cli -h 192.168.81.230 config get slaveof
1) "slaveof"
2) "192.168.81.220 6379"


     
      1.
      2.
      3.
      4.
      5.
      6.
      7.
      8.
      9.
      10.
      11.
      12.
      13.
      14.
      15.
      16.
      17.
      18.
      19.
      20.
      21.
      22.
      23.
      24.
      25.
      26.

Redis Working principle and architecture deployment of sentry cluster #yyds Dry inventory #_redis_06

2.8. The main library hangs up the changes in the configuration files of other nodes

After the main library hangs up , The other two nodes elect master after , The configuration file will also be filled in as new master The address of

Redis Working principle and architecture deployment of sentry cluster #yyds Dry inventory #_ The configuration file _07

原网站

版权声明
本文为[Jiangxiaolong's technology blog]所创，转载请带上原文链接，感谢
https://yzsam.com/2022/02/202202151355201556.html