An article takes you to understand the sentinel mode of redis

Reading guide redis Sentry mode , It is used in a master-slave cluster environment , If the primary server goes down , It will automatically set one of the slave servers as new master, And put the rest slave The configuration file of is automatically modified , In this way, a new set of master-slave services will be switched , No human intervention is required , And will not affect the use of services .

DSC0000.png(60.56 KB, amount of downloads : 0)

Download the attachment

The day before yesterday  10:19 Upload

background In the production environment , In order to ensure redis High availability of services , Usually, master-slave will be built . We know that the principle of master-slave is to obtain from the server rdb Full copy of files + Incremental replication of write operations to jointly ensure data consistency , So when configuring the slave server , A very important configuration item is to identify the main server ip And port number , I have to know which server is my master, not , Like in the following figure replicaof To configure .

DSC0001.png(46.85 KB, amount of downloads : 0)

Download the attachment

The day before yesterday  10:19 Upload

Easy to understand redis Sentry mode
For the deployment scheme in the figure , If the primary server goes down , We can only intervene manually , Choose a slave server to be the master server again , Then modify the configuration file of the other two slave servers , take replicaof The configuration of is changed to the new primary server address . Manual intervention takes time and effort without saying , what's more , This will make the service unavailable for a period of time . In this case , Sentinel mode came into being .
What is sentinel mode Sentinelredis Sentry mode , It is used in a master-slave cluster environment , If the primary server goes down , It will automatically set one of the slave servers as new master, And put the rest slave The configuration file of is automatically modified , In this way, a new set of master-slave services will be switched , No human intervention is required , And will not affect the use of services .
So how does it work ？ First look at the picture below ：

DSC0002.png(28.65 KB, amount of downloads : 0)

Download the attachment

The day before yesterday  10:19 Upload

Easy to understand redis Sentry mode

DSC0003.png(28.65 KB, amount of downloads : 0)

Download the attachment

The day before yesterday  10:19 Upload

Sentinel mode structure diagram
First of all, sentry is a service independent of master-slave service , It is also a cluster service . The sentinel instance will continuously send Ping, The master server receives , Return a valid reply , So the sentinel instance thinks that the server is normal .
Subjective offline Suppose the primary server goes down , sentry 1 Within the specified time ( Configurable ) No valid reply was received from the master server , Then the Sentry will mark the server offline , It's called subjective offline SDOWN.
Note that only one sentinel is marked as offline , In fact, the sentinel did not receive a reply for many reasons , Maybe the server did hang up , It is also possible that the server is not connected , No reply was received due to network reasons , All in all , The failure of a sentinel to receive a reply does not prove that the main server is down .
Objective offline sentry 2 Also sent Ping command , Also did not receive a reply , sentry 2 The primary server will also be marked SDOWN. This is the time ,3 Among the Sentinels 2 Sentinels reported SDOWN, After the Sentinels communicated with each other , It is believed that there are enough instances to prove that the service is no longer available , therefore , The sentinel instance will mark the server as offline ODOWN.
Sufficient quantity here is configurable , Usually half the number of sentinels plus 1, such as 3 Sentinels are set to 2.
Vote for , Fail over When the sentinel instance marks the service as objectively offline , There will be an election . In the remaining slave server instances , Select one as the master node , At the same time, modify the configuration files of other slave servers , Use the new master node as the source of data synchronization , Then restart the service , Finish switching .
thus , A complete sentinel automatic failover process is completed .
springboot Configure one master and multiple slaves + sentry If our environment changes from master-slave to master-slave + sentry , Modifying the configuration is also relatively simple , Comment out the original host and port Configuration of , Replace with the configuration of sentinel , Here's the picture ：
Easy to understand redis Sentry mode

DSC0004.png(21.52 KB, amount of downloads : 0)

Download the attachment

The day before yesterday  10:19 Upload

It should be noted that , here nodes The sentry cluster is configured in IP+ port , Instead of master-slave nodes , Don't make a mistake .

From the original ：