Guides

Testing restores - will your backup work when disaster strikes?

A green backup job does not prove that you will get your data back. How to plan, automate and document restore tests - with a test plan, Veeam features and a test report template.

Lennart from myveeam.cloud · Published on 6 October 2026 · 12 min read

Only an actual restore proves that your backup works when disaster strikes - not the green status of the backup job. You should therefore test in stages: automatically with integrity checks and boot tests (in Veeam, for example, Health Check and SureBackup) and additionally with scheduled tests ranging from a single file to a complete disaster recovery run. Measure whether the recovered data state and the time taken meet your RPO and RTO targets, and document the result.

This guide shows what the BSI (Germany’s Federal Office for Information Security), the GDPR (EU General Data Protection Regulation) and the EU NIS-2 Directive require, which tests make sense for small and medium-sized businesses and what Veeam Backup & Replication 13 can automate. At the end, you will find a template for your test report.

Why does ‘job successful’ not mean ‘recoverable’?

A successful backup job only tells you that data has been read and stored. It does not show whether the system boots, whether applications are consistent or whether the restore succeeds within an acceptable time. The BSI describes exactly this threat in module CON.3: regular backups do not automatically guarantee that the data can actually be restored.

Typical reasons why a restore fails despite green jobs:

  • Backup data on the storage has been corrupted without anyone noticing.
  • New servers or additional drives were never added to the job.
  • The application-consistent backup failed, but depending on the settings the job still only ended with a warning.
  • The system boots, but without Active Directory and DNS nobody can log in.
  • The password for encrypted backups is stored only on the failed server.

How wide the gap between expectation and outcome can be is shown by the Veeam Data Trust and Resilience Report 2026 (14 April 2026, more than 900 respondents): 90% of organisations are confident they can recover after a cyberattack, yet only 28% of those hit by ransomware got all their data back - on average, they recovered 72%. According to Veeam, attackers also targeted the backup repositories in 89% of ransomware incidents.

Veeam therefore extends the 3-2-1 rule to 3-2-1-1-0: the ‘0’ stands for zero errors during recovery, verified by automated checks and regular tests. This is a vendor recommendation, not a standard. How to set up the geographically separate copy is explained in our guide Off-site backup for SMEs.

What do the BSI, GDPR and NIS-2 require?

The IT-Grundschutz module CON.3 Data Backup Concept (2023 edition) makes testing a basic requirement. According to CON.3.A15, ‘it MUST be tested regularly whether the data backups work as intended, in particular whether backed-up data can be restored correctly and within an appropriate time’ (our translation). So time explicitly counts as well. The backup plans should also record the order of recovery (A4), the concept should state the RPO for each IT system, and the recovery information should remain available even if the backup systems fail (A6).

In its guideline LSI-Info T#01, which is primarily aimed at local authorities, the Bavarian State Office for Information Security (LSI) recommends checking the restore on a test system at least once a year - and explicitly applies the documentation requirement to recovery tests.

In Art. 32(1)(c), the GDPR requires the ability to restore the availability of personal data in a timely manner in the event of an incident, and in point (d) a process for regularly testing the effectiveness of the measures. A documented restore test is an obvious way of providing evidence of this.

For companies covered by NIS-2, Section 30 BSIG (German Act on the Federal Office for Information Security, the ‘BSI Act’) explicitly lists backup management, recovery after an emergency and the assessment of the effectiveness of the measures. Find out more in the article NIS-2 and data backup.

Note: This article provides a technical overview and does not constitute legal advice.

RPO and RTO - what do you measure the test against?

Without target values, there is no success criterion. Together with the business departments, define two values for each important system:

  • RPO (Recovery Point Objective): the maximum tolerable data loss. At 24 hours, a nightly backup may be sufficient; at four hours, it is not.
  • RTO (Recovery Time Objective): the maximum time it may take until the system is usable again.

During the test, you measure both: which restore point was actually available, and how long did it take from the decision to sign-off by the business department? The pure copy time is only part of this - providing replacement hardware, finding credentials and starting dependent services also take time.

Measuring is particularly important for the off-site or cloud copy. A worked example: transferring 2 TB over a 100 Mbit/s line takes around 44 hours in purely mathematical terms, at full utilisation and without overhead. At 1 Gbit/s, it would take around 4.5 hours. Only a test over your actual line shows whether this fits your RTO.

What should you test - and how often?

The frequencies in the table are suggestions for typical SME environments, not a standard - align them with your risk assessment and your RTO targets.

Test typeFrequency (suggestion)Success criterion
Individual files and foldersmonthly as a spot checkFile from the required restore point opens, permissions are correct
Complete VM or serverautomated weekly, manually quarterly for each critical systemSystem boots in the isolated network, services are running, login works
Applications (Active Directory, SQL Server, Exchange)quarterlyLogin and name resolution work, databases consistent (e.g. DBCC CHECKDB without errors), mailboxes accessible
Physical server (bare metal)every six months and after hardware replacementBoot from recovery media (in Veeam: Veeam Recovery Media), system runs on replacement hardware or as a VM
Off-site or cloud copyat least annually, preferably every six monthsAccess without the production environment, decryption works, duration fits the RTO
Microsoft 365 itemsquarterlyEmail, OneDrive file, SharePoint library and Teams content restored
Complete disaster recovery runannuallyProcedure as per the disaster recovery manual, RPO and RTO met, business department confirms it can work

Also test after migrations, major updates or a change of hypervisor. For individual objects from Active Directory, SQL Server and Exchange, Veeam offers dedicated Explorers. How to back up Microsoft 365 is described under Veeam Backup for Microsoft 365.

How do you automate checks with Veeam?

Veeam Backup & Replication 13 (currently version 13.1.1, build 13.1.1.18 of 13 August 2026) includes several complementary verification mechanisms.

Health Check: Veeam stores checksums during backup and compares them again during the Health Check for the most recent restore point. By default, it runs on the last Saturday of the month at 5:00 am. A retry run fetches corrupted blocks again from the source; on immutable Linux repositories, no repair is possible and a new active full backup is required. The Health Check is also available for backup copy jobs, and therefore for copies in the cloud repository.

SureBackup works in two modes:

  • Full recoverability testing: Veeam starts the machines directly from the backup in an isolated virtual lab, together with an application group (such as domain controller and DNS), and checks heartbeat, ping and applications using a test script. Predefined roles exist for DNS, domain controllers, mail servers, SQL Server and web servers, among others. Optionally, Veeam verifies the backup file via CRC and scans it for malicious code with Veeam Threat Hunter, antivirus software or YARA rules. The report is sent by email; on request, the job tests randomly selected machines.
  • Backup verification and content scan only: integrity check and malware scan without a virtual lab.

The boot test in the lab is available for VMware vSphere and Microsoft Hyper-V; agent backups of physical servers can be started there as a VM. For Proxmox VE, Nutanix AHV and XCP-ng/XenServer, Veeam only documents the second mode in version 13.1. Backups from backup copy jobs and from cloud repositories cannot be linked to SureBackup. In addition, the feature depends on the licence edition; without it, the remaining option is a manual check via Instant Recovery in an isolated network.

Secure Restore: Before a restore, Veeam can scan the data for malicious code and, if something is found, abort or restore the VM with its network adapters disabled. Run through this once in a ransomware scenario.

What SureBackup does not replace

SureBackup proves that a system boots from the backup and that defined services respond. It does not prove:

  • that the business department can work with the system and that external dependencies such as licence servers, interfaces or multi-factor authentication work,
  • that the restore to production succeeds within the planned time, because in the lab the machines run directly from the backup storage,
  • that credentials and the disaster recovery manual are at hand in an emergency,
  • that the off-site or cloud copy can be restored.

Organisation - what belongs in the disaster recovery manual?

Plan for an emergency so that recovery also works without the person who set up the system.

Order of recovery: A typical pattern:

  1. Network, firewall, virtualisation hosts and access to the backups
  2. Active Directory and DNS, and DHCP if applicable
  3. Databases and file servers
  4. Business applications, email and terminal servers
  5. Clients and downstream systems

Responsibilities: Who declares the emergency, who selects the restore point (in the case of ransomware, one from before the compromise), who releases systems, and who informs management, staff and, where applicable, the authorities? Appoint deputies.

Credentials offline: Emergency administrator accounts, passwords for encrypted backups and the configuration backup, licence keys, provider logins and contact details of service providers should also be kept in a secure offline location, for example sealed in a safe. The BSI explicitly warns against keeping recovery information in digital form only. In Veeam, only an encrypted configuration backup contains the stored credentials and keys - test restoring it to a new backup server as well.

Evidence: Document every test, including failed ones, and also keep the reports outside the IT environment.

Template - what belongs in a test report?

  • Date, test lead and participants (IT, business department, service providers)
  • Test object, test type and source of the backup (local, off-site, cloud)
  • Restore point used, with date and time
  • Target values for RPO and RTO
  • Measured times: start, end of data transfer, system ready for operation, sign-off
  • Result: passed, partially passed or failed
  • Problems and deviations from the disaster recovery manual
  • Actions with owners and deadlines
  • Confirmation that test systems were isolated and removed
  • Signature and storage location

Frequently asked questions

How often do I need to test restores?

There is no fixed statutory frequency. The BSI requires ‘regularly’, and the LSI Bayern recommends at least once a year. Ongoing automated checks plus manual tests, graded according to the importance of the system, make sense.

Is a SureBackup report sufficient as evidence?

It is good technical evidence, but it does not replace a documented test with the business department, measurement of the real recovery time and a test of the off-site copy.

How do I test without disrupting day-to-day operations?

Restore systems to an isolated network so that duplicate computer names, IP addresses or a second domain controller do not cause conflicts. SureBackup does this via the virtual lab.

How do I test the cloud copy without downloading everything?

Restore individual files and at least one representative system from the cloud copy, measure the throughput and extrapolate to your total volume. Not every type of restore is available directly from a cloud repository: according to the Veeam documentation, Instant VM Recovery on the tenant side is not supported for all platforms.

Conclusion

A backup is only worth something once the restore has been proven to work. Define RPO and RTO for each system, automate what can be automated, and schedule the remaining tests as fixed tasks - including the off-site copy and an annual disaster recovery run. Document every result and close any gaps you find. This gives you a solid basis for IT-Grundschutz, the GDPR and, where applicable, NIS-2.

With Veeam Cloud Connect, you transfer your backup copies in encrypted form to our cloud repository in an ISO/IEC 27001 certified data centre in Germany; additional encryption with your own key, immutability and Insider Protection are available, and we provide a data processing agreement (DPA) under Art. 28 GDPR. During the free trial period (up to 30 days, always ending at the end of the month), you can also try out restoring from the cloud repository and measure the transfer time over your line. With the Veeam Managed Service, we monitor your backup servers on site at your premises, with individually agreed response times. We are happy to answer your questions at hello@myveeam.cloud or +49 8106 928 9493.

Sources

Request a quote

Briefly describe what you need. Our specialists will get back to you with a no-obligation quote.

Fields marked * are required.

Helpful details include, for example, the approximate data volume in TB, the number of Microsoft 365 users or your Veeam version.

We only use your details to respond to your enquiry. Find out more in our Privacy Policy.

Prefer to contact us directly? hello@myveeam.cloud or +49 8106 928 9493