IT Infrastructure & Cloud
Backup Restore Testing: How to Make Sure Your Server Backups Actually Work
A backup logged as successful is not necessarily recoverable. This guide explains how a company plans and runs restore tests for server backups: scope, schedule, RTO and RPO, and documenting the results.
SEO and Content Specialist
7 min read
Short answer: a server backup restore test means restoring data or a system from backup to a separate environment, then checking whether the application runs, the data is intact, and the time required is still within the set limit. First set the RTO (how long the system may be down) and the RPO (how much data may be lost), write a test plan with pass criteria, run it regularly and whenever the system changes, then record the results and the fixes.
Backup reports that always say “successful” give a sense of security that may not be justified. A successful status only means the copy process finished; it does not prove that the copy is complete, readable, and fast enough to restore when a server fails or is hit by ransomware. The article The 3-2-1 Backup Strategy for Company Servers covers how to store backups. This article covers the next step: how a company plans and runs restore tests, so there is evidence that its backups can actually be used.
What a server backup restore test proves
A restore test proves three things that cannot be seen from backup logs: the backup can be read and restored, the restored system can be used, and the time is realistic. NIST SP 800-34 Rev. 1, NIST’s contingency planning guide for information systems, describes tests as evaluation tools that use quantifiable metrics to validate that a system is operable, and notes that tests often focus on recovery and backup operations. The same guide includes controls for testing backup information to verify media reliability and information integrity, and for using a sample of backup information to restore selected system functions as part of contingency plan testing.
Many problems only surface through tests like this: a database that was never included in the backup, a backup encryption key that is not stored anywhere else, or a restore that turns out to take far longer than expected.
Set RTO and RPO before testing
A restore test without targets produces a story, not a conclusion. Two figures need to be set first, with definitions from NIST SP 800-34:
| Term | Meaning per NIST SP 800-34 | What is measured during the test |
|---|---|---|
| RTO (Recovery Time Objective) | The maximum time a system resource can remain unavailable before the impact becomes unacceptable | Time from the start of the restore until the system is usable again |
| RPO (Recovery Point Objective) | The point in time before a disruption to which data can be recovered from the most recent backup | The gap between the restored backup and the time of the disruption |
| MTD (Maximum Tolerable Downtime) | The total business process downtime the system owner is willing to accept | The upper limit for RTO; RTO must be shorter |
These figures are a business decision, not purely a technical one. A daily transaction system and an archive of old documents can reasonably have different RTOs and RPOs. If a test shows that the restore takes longer than the RTO, the options are to change how backup and recovery are done, or to agree on new targets in writing.
Types of restore test, from light to comprehensive
A restore test does not always have to restore an entire server. NIST SP 800-34 distinguishes discussion-based (tabletop) exercises, functional exercises, and full-scale functional exercises, with depth increasing according to how important the system is. In practice, the types can be arranged like this:
| Type of test | What is done | Suitable for |
|---|---|---|
| Sample file restore | Restore a few random files or folders, then open them | Routine checks that the backup is readable |
| Database restore | Restore the database to a test server, then run the application against it | Systems whose data changes every day |
| Full server restore | Rebuild the server from backup in a separate environment until the service runs | Proving the RTO for important systems |
| Scenario exercise (tabletop) | The team walks through recovery steps for one scenario, without touching systems | Testing roles, contacts, and the order of decisions |
For moderate-impact systems, NIST recommends a functional exercise that includes an element of system recovery from backup media. For high-impact systems, the exercise may include recovering a server or database from backup, and should include a full recovery of the system to a known state.
Writing a test plan
NIST SP 800-34 recommends a test plan with clear objectives and success criteria, and a written scope, scenario, logistics, schedule, and list of participants. For a server backup restore test, the plan can be put together in six steps:
- Choose the system and the scenario. For example “the finance application database server is completely lost” or “one shared folder is encrypted by ransomware”.
- Write the pass criteria. For example: the application can be opened and logged into, record counts match, the expected latest transactions are present, and the restore time is under the RTO.
- Prepare a separate environment. The restored system must not be connected to the production network, so that scheduled jobs, email sending, or synchronization do not run a second time against the real data.
- Make sure the recovery materials are available. Credentials, backup encryption keys, licenses, and configuration notes must be accessible without the server that is “down”.
- Set participants and roles. Who runs the restore, who checks the data from the user side, and who records the time.
- Schedule it. Pick a time that does not disrupt operations, and inform those affected.
Running the test and measuring the results
During the test, record the time at every stage: when the restore started, when the data finished copying, when the application became usable. These records give the actual RTO. Also record when the restored backup was taken; the gap between that and the time of the scenario disruption is the actual RPO.
Integrity checks should be done by people who know the data, not only by the IT team. A finance officer will notice that yesterday’s transactions are missing faster than a server administrator will. NIST also recommends that testing be conducted as close to the operating environment as possible, so a test that only restores files without running the application does not yet prove that the service can come back.
How often to run the test
There is no single schedule that suits every company. NIST SP 800-34 says tests and exercises should be conducted periodically, following organizational or system changes, and as otherwise needed, at a frequency the organization defines itself. CISA, in its #StopRansomware guide, recommends regularly testing the availability and integrity of backups in a disaster recovery scenario.
Besides the routine schedule, some events should always be followed by a restore test:
- A server or application moves to a new location, including migration to the cloud.
- The backup software or backup destination changes.
- A new system is added to the list of what is backed up.
- The person who usually runs restores changes.
The depth of the test can be adjusted: sample file restores more often, full server restores for important systems on a less frequent schedule.
Documenting the results
NIST SP 800-34 recommends that each test be recorded in an after-action report, with lessons learned and corrective actions used to update the recovery plan. A restore test report can be brief, as long as it contains:
| Report section | Content |
|---|---|
| Scope | The system, the scenario, and the backup used (date and location) |
| Results against targets | Actual RTO and RPO compared with the targets |
| Integrity checks | What was checked, by whom, and the result |
| Problems found | Steps that failed, materials that were unavailable, time overruns |
| Corrective actions | What is changed, by whom, and when it will be retested |
This report is also useful during a real incident: the team does not need to guess the order of steps, because that order has already been run and recorded.
How Legia does it
In the Server Installation & Management service, backups run automatically and restoring them is tested when they are first set up. Periodic restore tests are scheduled if agreed in the management contract. For servers already in operation, Legia starts with an infrastructure audit (3-7 working days) that checks the server’s condition, security configuration, and how backups are being run, then addresses critical findings including fixes to the backups. The scope covers physical servers in the office, virtual servers, and the cloud; considerations for choosing where to run them are in the article Office Server, Cloud, or Shared Hosting, and the other services in this pillar are under IT Infrastructure & Cloud.
Frequently asked questions
What is the difference between a successful backup and a tested backup?
A successful backup means the copy process finished without errors. A tested backup means that copy has been restored, the system ran, the data was checked, and the time was recorded.
Can a restore test be done without disrupting the running server?
Yes, by restoring to a separate environment, for example a test server or a virtual machine that is not connected to the production network. The main server keeps running as usual.
Who should be involved in a restore test?
The team or party managing the server runs the restore, and a user representative from the department that owns the data checks its integrity. The system owner approves the RTO and RPO targets and the corrective actions.
What should be done if a restore test fails?
Record the cause, fix the backup method or the restore steps, then repeat the test for the same system. A failure found during a test has far lighter consequences than one discovered during a real incident.


