Back to blog
November 9, 20269 min read

Hardened Repository vs Object Lock: What Each Stops in Veeam 13

Veeam hardened repository vs S3 Object Lock in VBR 13.1.2: who enforces each lock, which threats each one survives, SOBR copy timing and the limits of both.

by BackupSec Team

Hardened Repository vs Object Lock: What Each Stops in Veeam 13

Hardened repository or Object Lock

A hardened repository locks backup files with the Linux immutable attribute on a server the backup team runs. Object Lock locks object versions in S3 storage, and the storage provider enforces the lock. Each protects a different failure domain. Where recovery must survive the loss of the primary site, use both: a hardened repository for fast local restores and an offsite Object Lock copy.

Note: This article describes Veeam Backup & Replication 13.1.2 (build 13.1.2.57), the build the cited User Guide pages apply to. Behaviour differed in v11 and v12.

How each one locks

Hardened repositoryObject Lock on Amazon S3
MechanismLinux immutable attribute, set and removed by the Veeam Immutability ServiceS3 Object Lock with Versioning, in Compliance mode
Enforced byThe host's Linux kernel and clock; root on the host can remove itThe storage provider
Period7 to 9,999 days1 to 999 days, plus Block Generation
ExtensionFiles of the active chain, from its last restore pointBlocks still in use, in active and inactive chains
Backup server accessSingle-use credentials that are not stored, or appliance certificatesStored keys that write and read the bucket
Clock defenceTime-shift detection blocks retention operationsThe provider's clock

Hardened repository boundary

Deployment credentials for a hardened repository are used once and not stored, so a compromised backup server cannot use them to connect. The Veeam Immutability Service (veeamimmureposvc) runs with root permissions as a child process of the non-root Data Mover and checks attributes every 20 minutes.

The boundary is the Linux host: anything with root on it, or below it, can remove the lock. The best practice guide's immutability checklist names out-of-band management (IPMI, iDRAC, iLO), which "can override any protection the OS enforces, including immutability", plus SSH and broad sudoers entries. Backup Data Hardening adds the hypervisor under a virtual repository.

  • Deploy on a physical server with local storage, as the requirements recommend. The Security & Compliance Analyzer flags a hardened repository in a virtual machine.
  • Prefer the Veeam Infrastructure Appliance: certificate-based authentication, no SSH, MFA that cannot be disabled without Security Officer accounts, and no domain join by default. Disable its Host Management console Web UI. Keep the domain restriction: lifting it requires Application Backup Repository mode, and combining the two roles "increases the attack surface on its immutability".
  • On a self-built server, disable SSH after deployment, lock root, review sudoers as carefully as root and automate patching. On either kind, restrict out-of-band management with firewall rules and MFA where supported, and use authenticated time (NTS).
  • Enable four-eyes authorization. Since 13.1 it covers reducing or disabling the period on Linux and hardened repositories, though it cannot protect a compromised backup server.
  • Do not count a governance-mode Linux repository, new in 13.1, as hardened: anyone with root on it can overwrite or delete locked files. The same applies to a Linux backup server acting as its own immutable repository, which Veeam says increases the attack surface.

Clock and session behaviour

A time shift above 86,400 seconds, or a shutdown longer than 24 hours, blocks retention operations, and the service "will not set or remove the immutable attribute for existing and new backup files". KB4482 makes the threshold configurable from 60 to 604,800 seconds and can disable the check. Keep the default, because raising or disabling it weakens the clock defence. Reset the protection only after confirming that the shift was expected and the current time is correct: in the Host Management console TUI on the appliance, with root on a manual server.

The flag is set when the backup session completes. If a session fails on an environmental issue and the next run creates a new full, the failed run's files stay mutable until set with the Set-VBRImmutabilityLockExpirationDate cmdlet. .VBM metadata files are never immutable. Verify the session result and the actual lock before counting a restore point, as described in how to prove immutable backups are immutable.

Object Lock boundary

In Compliance mode on Amazon S3, "a protected object version can't be overwritten or deleted by any user, including the root user in your AWS account", and its retention period can't be shortened. The one exception AWS names is deleting the AWS account. On Amazon S3, Veeam uses Compliance mode for each uploaded object, and the Security & Compliance Analyzer checks that immutable object storage repositories use it. S3-compatible providers implement Object Lock themselves, so the AWS guarantee does not transfer automatically; the S3-compatible immutability matrix lists what each one supports.

The boundary is the account and the provider. Nothing at the backup site can shorten an existing Compliance lock, but the backup server's stored keys can still read data, stop future backups, add delete markers on top of locked versions and extend retention, which nobody can shorten afterwards.

  • Create the bucket with Object Lock and Versioning, and disable default retention, which "may result in an unpredictable system behavior and data loss". Do not toggle Versioning or Object Lock after adding the bucket to Veeam, and do not add lifecycle rules.
  • Place the bucket in a separate account and region, administered by identities that do not exist in production. Scope the backup server's key to that bucket.
  • Set the period deliberately. Since 13.0.1 P1 the default mode follows the job's entire retention with a 30-day minimum, and the actual retention adds Block Generation: by default 10 days, or 30 for Amazon S3, IBM Cloud, Google Cloud and 11:11. Every locked day is billed, and a period longer than job retention "will result in extra charges".

Threats compared

ThreatHardened repositoryOffsite Object Lock (Amazon S3, Compliance)
Backup server compromisedExisting locks hold; the server can still read backups, stop jobs and lower the period for new filesExisting versions hold; stored keys can still read, add delete markers, extend locks and stop jobs
Production domain compromisedHolds if the repository, its out-of-band management and any hypervisor under it are outside that domainHolds if cloud administration uses separate identities
Root, sudoers or out-of-band access on the repository hostLocks can be removedNot affected
Hypervisor compromiseA virtual repository's disk can be deletedNot affected
Site lossLost with the siteSurvives in another region
Cloud account compromiseNot affectedLocked versions stay unless the account is deleted; access may be lost

This table is our reading of the documented mechanisms. Neither Veeam nor AWS has certified it.

SOBR topology and copy timing

Veeam allows immutability on any tier of a scale-out backup repository. During an incident the question is when the second immutable copy exists.

  • Copy mode starts its copy session after the backup or backup copy job finishes, outside prohibited hours. Bandwidth and backlog set the delay.
  • Move mode runs an offload session every 4 hours for inactive chains outside the operational restore window, so the offsite copy lags by at least that window. Linux-based performance extents need direct internet access for offload, without an HTTP(S) proxy. On a hardened performance extent, still-immutable files are copied and deleted only after their lock expires.
  • Direct to object storage has no transfer stage, but a written object is not a finished backup. Confirm job success first.

The capacity tier locks backups only for its object storage repository's period, "even if their retention policy allows for longer storage", and an entire-retention repository switches to the minimum period there. That includes GFS backups, and with a move policy a hardened performance extent ignores GFS retention too. A yearly GFS point can end up locked for the repository period only, so check its observed expiry.

Note: For direct-backup object storage used as a performance extent with a move policy, Veeam shortens the immutability period to fit the move. Its example has 28-day job retention, a 7-day move policy and a weekly full: 7 plus 7 plus 3 days for the offload gives 17 days. That algorithm is documented for object storage extents only.

Our order of preference:

  1. Hardened repository as the performance tier, with copy to an immutable capacity tier in another region. Monitor every offload session and the lock expiry of offsite GFS points.
  2. Direct to immutable object storage, where network restore throughput meets the RTO.
  3. Move only, where the measured offsite lag meets the recovery requirement.

Limits of both controls

  • An attacker who controls the backup server can disable jobs or change the immutability period. Existing locks never drop below their original expiry, but they expire, so alert on disabled jobs and failed sessions well inside the lock period. Four-eyes approval does not cover raising the period, which fills a hardened repository or bills Object Lock storage until expiry.
  • Immutability preserves corrupted or already encrypted data as faithfully as good data, and it does not protect a lost encryption key. Test restores from each copy; since 13.1 a second backup server can attach to an immutable bucket read-only for that.
  • The timing above covers image-level VM and physical machine backups. Log backups, plug-ins, NFS and SMB WORM and deduplication appliances follow their own rules. A backup copy job to a hardened repository is immutable only with GFS retention enabled.

A passed Security & Compliance Analyzer check does not prove that every restore point is locked or restorable.

How BackupSec approaches this

ZeroVault is managed S3-compatible object storage built to hold the offsite copy this article argues for. It offers S3 Object Lock with governance and compliance modes and legal hold; Veeam writes compliance-mode locks by default, which nobody, including our own staff, can shorten or remove before they expire. The period follows your repository and job settings, which we agree with you in writing during onboarding. You choose one of 15 regions in North America, Europe and Asia, which lets you keep the copy in a different failure domain from your primary site. There are no API fees, pricing and egress terms are on request, and support is 24/7 in Turkish and English. A cloud copy is still exposed to an account compromise, which is why the administration has to stay separate from your production identities.

Sources · 26

Veeam · AWS

Frequently asked questions

Should I use a hardened repository or Object Lock?

Use both where recovery must survive the loss of the primary site. A Linux hardened repository gives fast local restores and keeps its deployment credentials off the backup server, but root on the host, out-of-band management or a hypervisor below it can remove the lock. An offsite Object Lock copy in Compliance mode survives site loss and an on-premises compromise, and depends on the cloud account and the provider instead. [Veeam documents the hardened repository's authentication boundary](https://helpcenter.veeam.com/docs/vbr/userguide/hardened_repository.html). Neither control replaces tested restores.

Can a Veeam hardened repository be joined to a domain?

Not if it was deployed from the Veeam Infrastructure Appliance. In VBR 13.1.2 (build 13.1.2.57) that hardened repository cannot join a domain by default. Lifting the restriction requires enabling Application Backup Repository mode and disabling strict FIPS compliance mode, and [Veeam notes](https://helpcenter.veeam.com/docs/vbr/userguide/hardened_repository_limitations.html) that combining the roles increases the attack surface on its immutability. Veeam documents no such restriction for a self-built Linux hardened repository. The risk is whether compromised domain identities gain privileged access to the repository, its out-of-band management or its hypervisor, so keep all of them outside the production domain.

Does SOBR copy mode guarantee an offsite copy within minutes?

No. In VBR 13.1.2, Veeam [starts the copy session after the backup or backup copy job finishes](https://helpcenter.veeam.com/docs/vbr/userguide/capacity_tier_copy.html), and it does not run during prohibited hours. Transfer time depends on data size, bandwidth, backlog and errors. Count the offsite copy only after the transfer and its lock have completed. Move-only mode waits until the chain is inactive and outside the operational restore window, then runs on a 4-hour offload schedule. On a hardened performance extent, still-immutable files are copied rather than moved. Monitor the age of the newest restore point that reached the capacity tier.

Does the documented 17-day move example apply to Linux hardened repositories?

Not as documented. The [VBR 13.1.2 performance-tier example](https://helpcenter.veeam.com/docs/vbr/userguide/immutability_performance_tier.html) sits under the heading for direct-backup object storage repositories used as performance extents. Its 28-day job retention, 7-day move policy and weekly full produce the documented 17-day period. A Linux hardened repository keeps its own period and active-chain rules, but with a move policy Veeam ignores GFS retention for its lock, so GFS fulls are immutable only for the repository's period. Check the actual extent type and the observed expiry on files or objects before relying on either number.

Is a governance-mode Linux repository the same as a hardened repository?

No. Veeam Backup & Replication 13.1 added [governance-mode immutability for standard Linux repositories](https://helpcenter.veeam.com/docs/vbr/wn/backup_repositories.html). It uses the same immutability engine as the hardened repository, but [anyone with root on the repository](https://helpcenter.veeam.com/docs/vbr/userguide/linux_server.html) can overwrite or delete locked files, and it can host roles a hardened repository cannot. Veeam positions it for flexibility of deletion at the system administrator level. Against an attacker with root, keep a hardened repository or a Compliance-mode Object Lock copy.