Test a backup by restoring the complete workflow
Check the database, files and access together. A worked timeline separates the recoverable data gap from recovery time.
A successful backup notification shows that a backup job has completed. To find out whether your team can work with it again, restore and test a complete business workflow. Include files, permissions and the connections it needs.
Start with a safe test scope
Use a separate test environment and synthetic data. Disable outgoing email, payments and other real actions. Do not restore over production as an experiment. Required keys and credentials must be available through the planned recovery process without being copied into the test record.
This fictional example tests an order application. The chosen objectives are a maximum data gap of 60 minutes, the Recovery Point Objective or RPO, and no more than 120 minutes until the service is usable again, the Recovery Time Objective or RTO. These are invented example targets, not promised service levels. AWS recommends setting such objectives according to business impact.
A completed timeline
Time | Event in the example |
|---|---|
09:25 | Consistent data state, including test order T-17 and its attachment |
09:30 | Backup job containing that data state finishes |
10:00 | Assumed outage; the recovery clock starts |
10:40 | Database technically restored |
10:50 | Login works; attachments are still missing |
11:05 | Attachments, permissions and the complete order workflow pass testing |
The verified data gap is 35 minutes, from 09:25 to 10:00. Recovery takes 65 minutes, from 10:00 to 11:05. The example therefore meets both chosen objectives. A backup job finishing at 09:30 does not mean the data it contains is current to that time.
The test would not have passed at 10:40. A running database without required attachments does not support the agreed workflow. Successful login at 10:50 is also insufficient.
Demonstrate business completion
Check T-17 against its expected order lines and open its attachment. Sign in with a designated ordinary role: it must see permitted orders and be unable to open protected areas outside its permissions. Then create a new test order and take it through the intended final step. External handovers use test destinations only.
The example record initially says “Attachment missing; test failed”. After correction and retesting, it says “T-17 complete; roles checked; new test order completed”. Record the backup's data state and the actual times. A tick without the test result will not explain later what was restored.
Let the finding determine the change
Missing files require a change to backup scope. If a key is missing, its controlled availability belongs in the recovery plan. If recovery is too slow, identify the actual bottleneck and repeat the affected steps. More frequent backups alone will not speed up a slow import, for example.
The example timings do not replace your own measurements. Repeat the full exercise particularly after changes to storage locations, encryption, roles or the recovery procedure. Keep a clear result so the next exercise has a useful point of comparison.
Sources
Read more
All articles
Cloud & Operations with ASES
Hosting, deployment, maintenance and support