Field Notes03 / Operations
Backups you can actually restore.
The useful question is not whether a copy exists. It is whether the recovery path works.
Start with what matters
List the information needed to rebuild the service: data, configuration, access requirements, and the software version. Note which files are generated and which contain the only copy of something important.
Keep recovery independent
A backup on the same disk shares many of the same failure modes as the original. Keep an independent copy, restrict access, and make sure the credentials needed to recover it are available when the primary machine is not.
Restore into an empty location
- Choose a small, representative backup.
- Restore it to a separate directory or test machine.
- Check file ownership and permissions.
- Open the data with the application that uses it.
- Record the time taken and anything the instructions missed.
A checksum can show that bytes match an expected copy. It cannot show that a database is consistent or that the application can use the recovered files. Include that application-level check.
Leave a short recovery note
Write the location of the backup, the restoration steps, and the date of the last successful test. Keep enough history to recover from a mistake discovered later, with retention suited to the data.