Define what must survive
List application data, databases, configuration and the information needed to regain access. A filesystem copy may not capture a consistent database while writes are in progress. Follow the database backup method and check that the result can be read.
Understand the failure boundary
A snapshot can make a change easier to reverse, but it may depend on the same platform or storage as the original. Redundant storage can tolerate some device failures, but it does not stop an authorised command deleting a file. Keep an independent copy outside the environment you are protecting.
Protect the recovery copy
Limit access to backups and consider encryption where the data requires it. Record retention and removal rules. A copy that anyone with a compromised server account can delete may not survive the incident you need it for.
Test the return journey
Restore a sample into a separate environment and verify the application can use it. Record the steps, dependencies and time required. A recovery plan should remain available when the original server is unavailable. Revisit the test after significant application or storage changes.