Plan A - Copy
It is quite reasonable, even with a mixture of Engines, to stop a Slave, copy mysql's directory tree to another Slave, verify that the Slaves have my.cnf
files with different server_id
, start both slaves.
After that, both Slaves will catch up with replication from the same spot, and be online (hopefully) pretty soon.
That process is the same as for bringing a new Slave online.
Plan B - LVM
With some planning ahead, you could have LVM set up. With that, you can take a "snapshot" of a Slave with only a minute of downtime -- even for 700GB. Then take as long as you need to copy the snapshot to the other Slave (recovered / new). LVM requires OS work at the file system level, so it is not easily retrofitted.
Copying the snapshot (to another machine) can give you a backup. Or it can give you a copy of production for use in QA/Dev/etc. The copy will be not just the 700GB, but the entire file system.
Plan C - Galera
See also Galera (PXC / MariaDB / Galera addon), which has the concept of IST ("Incremental State Transfer"), which it uses when a node comes back online after a crash. (That only works if it has not been offline "too long", else a full state transfer will be performed.)