How Synchronization and Backup are Implemented
All supported versions of Redis support backup and synchronization between the primary and replica clusters. However, the way that backup and synchronization is implemented varies depending on the Redis version.
Redis Version 2.8.22 and Later
During the forkless process, if the write loads are heavy, writes to the cluster are delayed to ensure that you don't accumulate too many changes and thus prevent a successful snapshot.
Redis Versions Prior to 2.8.22
Redis backup and synchronization in versions prior to 2.8.22, is a three-step process.
Fork, and in the background process, serialize the cluster data to disk. This creates a point-in-time snapshot.
In the foreground, accumulate a change log in the client output buffer.
If the change log exceeds the client output buffer size, the backup or synchronization fails. For more information, see Ensuring You Have Sufficient Memory to Create a Redis Snapshot.
Finally, transmit the cache data and then the change log to the replica cluster.