Redis master-slave synchronization principle

1. Foreword

In order to ensure the high availability of redis, a cluster mode is generally built, which is the master-slave mode.

The master-slave mode can ensure the high availability of redis, so how does redis ensure the data consistency of the master-slave server? Next, let's briefly talk about the principle of redis master (master) and slave (slave) synchronization.

2. Initial full synchronization

#When a redis server sends the salveof command to the master server for the first time, the redis slave server will perform a full synchronization , the synchronization steps are shown in the figure below:

• The slave server sends the psync command to the master (the psync command sent at this time is psync ? -1 ), tell the master that I need to synchronize data.
• After receiving the psync command, the master will execute the BGSAVE command to generate a snapshot of the RDB file.
• After the generation is completed, the RDB file will be sent to the slave.
• After the slave receives the file, it will load the RDB snapshot and change the database status to be consistent with the status of the master when executing BGSAVE.
• The master will send all the write commands saved in the buffer to tell the slave that it can be synchronized.
• The slave executes these write commands.

3. Command propagation

The slave has synchronized with the master, then if the subsequent master For a write operation, such as a simple set name redis, after the master executes the current command, it will send the current command to the slave for execution to achieve data consistency.

4. Re-copy

When the slave is disconnected and reconnected, it will be resynchronized. The resynchronization is divided into full synchronization and partial synchronization

First let’s take a look at the general trend of partial synchronization

• ##When the slave disconnects and reconnects, the psync command will be sent to master.
• The master will return a continue reply after receiving psync, indicating that the slave can perform partial synchronization.
• master sends the write command after disconnection to slave
• slave executes the write command.

In fact, after the slave sends the psync command to the master, the master still needs to judge whether to perform partial synchronization based on the following three points.

Let’s first introduce the three aspects:

• Server running ID

Every After a redis server is started, a running ID will be generated.

When performing the initial synchronization, the master will tell the slave its ID, and the slave will record it. When the slave is disconnected and reconnected, it will try to perform the synchronization if it finds that the ID belongs to this master. Partial resync. A complete resynchronization will be performed when the ID is different from the currently connected master.

• Replication offset

Replication offset includes master replication offset When the replication offset of the two databases is the same after the initial synchronization, then the master executes a write command, then the offset of the master is 1, the master will write the command to the slave, the slave executes it once, and the slave offset Measure 1 so that the versions will be consistent.

• Replication backlog buffer

The replication backlog buffer is fixed and maintained by the master A first-in-first-out queue of length.

When the slave sends psync, it will also send its own offset to the master. When the data after the slave's offset still exists in the buffer, it will return continue to notify the slave. Perform a partial resynchronization.

When the data after the slave offset is no longer in the buffer, a complete resynchronization will be performed.

Combining the above three points, we can summarize:

• When the slave disconnects and reconnects, it will send the psync command to the master.
• The master will first judge the server running id. If it is the same as its own, it will judge the offset.
• The master will judge its own bias. Whether the shift amount is consistent with the offset of the slave.
• If it is inconsistent, the master will go to the buffer to determine whether the data after the slave's offset exists.
• If it exists, a continue reply will be returned, indicating that the slave can perform partial synchronization.
• master sends the write command after disconnection to slave
• slave executes the write command.

5. The final process of master-slave synchronization

6. Conclusion

Recently, when the company needed it, I built a redis master-slave cluster and used sentinels to monitor and implement master-slave switching. Therefore, I sorted out the master-slave principle of redis based on "Redis Design and Implementation" to deepen my impression.

## Recommended tutorial: "

redis tutorial"

The above is the detailed content of Redis master-slave synchronization principle. For more information, please follow other related articles on the PHP Chinese website!