דילוג לתוכן

How does CBT work? The guide to change tracking and virtual machine backup.

How tracking changed blocks reduces backup work, what to check in VMware and Hyper-V, and how to ensure that acceleration does not replace proper recovery.

23.10.2026 · 8 minutes read

Backing up a virtual machine does not have to reread the entire disk in every run. CBT, an acronym for Changed Block Tracking, allows the backup software to ask the virtualization layer which disk areas have changed since the last tracking point. The goal is to reduce read work in incremental backups. The guide explains the principle, use in VMware, and the differences from Hyper-V, along with checks to perform before relying on it.

The change map
The virtual disk
did not changechanged
What do we focus on when reading it?

The areas that changedAccording to the tracking map, when it is correct and supported

We do not start searching everything again: the change map indicates the necessary areas. The number of blocks and their color are for illustration only.

What is a block, and why is it not enough to look at files?

A virtual disk contains data areas that the backup software can process at the block level. A small change within a large file may affect only some of its disk areas. Instead of assuming that the entire large file needs to be read again, an appropriate tracking mechanism can indicate the areas that changed. A block is not necessarily a file, and the size and accuracy of tracking units depend on implementation.

The distinction is also important in interpreting reports: disk size, the volume of data read, and the volume of data sent to storage are different metrics. Do not expect them all to show the same number.

How CBT integrates into the backup run

  1. A baseline is created: The backup program saves an appropriate base copy and information that allows identifying the tracking state.
  2. The system continues to work: The virtualization layer records which areas have changed.
  3. Another run begins: The backup software requests the map of changes between the appropriate checkpoints.
  4. The required areas are called: When tracking is intact and supported, the incremental backup can focus on the changes rather than scanning the entire data set.
  5. A restore point is saved: The backup software manages the data and the relationship to previous copies according to the backup format and its retention policy.
This is how a incremental backup run proceeds
01Base pointPrevious copy and valid tracking identifier
02Change mapThe virtualization layer marks what has changed
03Reading the changesThe backup software reads the required areas
04Restore pointSaved by the backup software, alongside the reference to the base
The map helps select what to read. Data storage and restore point management remain the responsibility of the backup software.

In VMware, tracking occurs at the virtualization layer, not by searching files from the guest operating system. The backup software uses vSphere interfaces to request the relevant information. Documentation of tracking and its use in a managed backup plan.

Numerical example: what can be saved?

Suppose the machine has a 1TB virtual disk, and between two runs, areas totaling about 10GB have changed. When tracking is supported and functioning properly, reading can be limited to the areas indicated by the map. This is an illustration of the principle, not a promise that exactly 10GB will be transferred or that the backup will be a hundred times faster.

The same disk. A different volume of changes.
The entire disk1 TB
The areas that changed in the example10 GB
About 1% of the disk's volume in the exampleLess data for testing does not guarantee the same proportion of backup time reduction.
For illustration purposes: 1TB = 1,000GB. The actual volume read and transferred depends on tracking units and processing; the diagram is not a performance guarantee.

Actual time also depends on storage speed, host load, network, and processing by the backup software. Compression, encryption, and the tracking unit may change the ratio between data read and data sent. Therefore, several similar runs are compared, rather than relying on a single run.

Before running in VMware

According to VMware documentation, virtual hardware version 7 or higher, CBT activation, and a supported disk configuration are required. An RDM disk in physical compatibility mode and disks not affected by snapshots are not suitable for the described mechanism. It is also advisable to check the backup solution's support for your host version and configuration; platform-level support alone does not constitute product compatibility approval.

During initial activation, existing snapshots should be handled according to the documentation and a valid starting point ensured. This does not imply that temporary snapshots may not be used as part of a backup: their creation may be part of the normal process. Disk files or tracking files should not be deleted manually based on a general guide. CBT Requirements and Limitations in VMware Documentation.

Running the backup plan in BackupONE

  1. Open or edit the appropriate VMware backup plan.
  2. Locate the advanced settings for machine selection or for the plan, according to the interface version.
  3. Select the option Use Changed Block Tracking.
  4. Save, run the plan, and check the log to ensure that tracking has indeed been enabled, and not just whether the checkbox remained selected.
  5. Also check the restore point that was created and the business outcome of a test restore.

The option’s location and screen names may differ between browser-based and local management. The system may try to enable tracking on every run and use an alternative path when tracking is unavailable. The documentation explains activation and behavior when CBT is unavailable.

CBT does not guarantee application consistency

A change map is not proof that a database or application was saved in a consistent state. In VMware, application-consistent processing is a separate setting that uses VMware Tools and the relevant mechanisms within the guest OS. VMware Tools is required for guest-consistent processing, not for the CBT mechanism itself.

For a critical plan, decide what to do if application processing fails and check what actually happened during the run. Technically successful block transfer does not replace checking that the application starts and the required information is available. Distinguish change tracking from application processing in VMware backup.

What is different in Hyper-V and other platforms?

In Hyper-V, the corresponding mechanism is RCT — Resilient Change Tracking. The backup documentation specifies operating system and VM configuration version requirements. In the documented versions, the option may appear as a beta feature and be associated with backing up the current checkpoint. Check the installed version before enabling it. Consult the requirements and instructions for enabling change tracking in Hyper-V.

Do not copy VMware settings to another platform just because it also has incremental backup. In Proxmox or any other environment, it is worth separately checking the backup method, the compatibility, and the limitations of the relevant solution.

What to check when the incremental backup suddenly becomes slow?

  • Were CBT or RCT actually enabled in the last run?
  • Was there a change in the machine configuration, disks, snapshots, or checkpoints?
  • Is there a prior valid recovery point, or does the log report that it cannot be used?
  • Is the delay in reading, transferring, or at the storage target?
  • Was a valid restore point created despite switching to another read path?

In Hyper-V, the message Changed Block Tracking was not applied with code 1934 refers to a situation in which a previous tracking point could not be obtained or verified. This message alone does not establish that the entire backup failed; check the run’s outcome and the affected VM list. The documentation explains code 1934 and a missing tracking point.

Restore, retention, and protection: what remains the responsibility of the plan?

CBT does not encrypt the backup, does not prevent deletion, and does not provide immunity to ransomware. Retention policies, permissions, separate storage targets, and protections such as Object Lock are handled at other layers. Before changing the backup chain or resetting tracking, it is important to understand how recovery capability is preserved; automatic resets should not be performed in the hope of improving speed.

A useful test is a documented recovery in an appropriate environment: selecting a point in time, opening a file, or launching a machine and checking the application. This verifies the purpose of the backup, beyond a green light in the report.

Frequently asked questions

Does CBT replace a full backup?

No. It helps identify changes. Backup software still requires a valid database and recovery points that can be used.

Does a backup without CBT necessarily fail?

No. Depending on the product and configuration, a broader read may be performed and the backup may be slower. Check the log and the run result.

Can a fixed acceleration rate be guaranteed?

No. The scope of changes and bottlenecks varies. Measure read, transfer, and runtime under similar conditions, and also check recovery.

Back to BackupONE virtual machines backup solution →

Continue reading