Live Browse and Block-Level Browse

To browse and recover files, you must have information about the files and folders included in a backup (metadata). This file metadata can be provided in the following ways:

  • Discovering file and folder information dynamically during the browse operation. This capability is known as Live Browse. Live Browse is used when file and folder metadata is not available for a backup.


    This method was originally introduced as a feature called Live File Recovery for VMware restores. It is now the default method for guest file and folder restores for all hypervisors except Docker.

  • Collecting file and folder information during the backup (not recommended).

Live browse with block-level reads is replacing metadata collection as the default mechanism for browse and restore operations.

Live browse can be used for both Windows and Linux virtual machines. To read data for advanced Linux file systems, a File Recovery Enabler for Linux can be used to browse and restore data from backups of Linux guest VMs.

If antivirus software is installed on VSA proxies or MediaAgents, specify exclusions for Commvault files and folders. Otherwise, the time required to browse and recover files might increase significantly.


The Multipath Input Output (MPIO) must be disabled for CVLT disks. For more information, see Disable the Multipath Input Output (MPIO) for Commvault disks.

Block-Level Browse

Block-level browse enables users to browse and recover files without requiring metadata collection during backup. Block-level browse is supported for streaming backups for all hypervisors, and for IntelliSnap backups for VMware and Nutanix AHV. As noted below, some hypervisors still provide the option to collect metadata during backups to enable file browsing and recovery.

Block-level browse uses a block-level driver to mount a pseudo disk on the MediaAgent being used for browse and restore operations. The pseudo disk is used to get file system information, enabling browse and restore operations to read directly from stored backup data, without relying on content indexing.

Block-level browse replaces 3DFS Live Browse and is used as the underlying mechanism for the following features:

  • Application aware backups

  • Live browse

  • Live file recovery

For Windows guest VMs, the Windows MediaAgent that is used for browse and restore operations must have the Virtual Server Agent installed.

For Linux guest VMs, a File Recovery Enabler for Linux can be used to browse and restore data from backups of Linux guest VMs.

To stage extents for block-level live browse operations, the block-level browse feature uses a Least Recently Used (LRU)-based pseudomount cache in the job results directory. The pseudomount cache is pruned periodically to free up extents. At least 20 GB of free space is required.

When restoring large files, you can improve the performance by increasing the size of the pseudomount cache so that it does not need to be pruned frequently during restores. For optimal performance, the free space on the pseudomount cache should be at least 10% of the overall disk size for the cache plus the size of the data to be restored. For example:

Disk size for pseudomount cache

Size of data to restore

Free space required

500 GB

10 GB

At least 60 GB

1 TB

100 GB

At least 200 GB

100 GB

5 GB

20 GB (minimum requirement)

For block-level restores, in addition to the restore job, the Job Controller launches a persistent recovery job that opens a common pipeline, enabling multiple extent recall requests to be submitted as a group. The default timeout for a persistent recovery job is 7 days. For block-level restores using the Virtual Server Agent, the persistent recovery job remains open for 7 days and can be used for subsequent block-level restores that use the same proxy.


  • Unless otherwise noted for a particular feature, block-level browse (also called live browse) and block-level restores are not supported from backups to tape libraries or virtual tape libraries.

  • Block-level browse is not supported on VM disks that have a mount point instead of a drive letter.

  • Due to a Microsoft limitation, block-level browse is not supported on ReFS volumes if the MediaAgent used for the browse is running on a Microsoft 2008 R2 server or earlier version.

Block-Level Browse for Large Disks

To perform a block-level browse for disks that are 16 TB or larger, take the following actions:

  • Create a new disk on the MediaAgent and format it with an NTFS cluster size of 8 KB (to live browse a disk of 16 TB to 32 TB) or 16 KB (to live browse a disk of 32 TB to 64 TB).

  • Change the path of the Job Results folder to use the new disk.

During a live browse, a sparse file with size equal to the source volume is created to stage the recalled extents. NTFS volumes with allocation unit size less than 64 KB do not support the creation of sparse files greater than 16TB. For more information on supported volume size and allocation unit size, see Microsoft KB article Default cluster size for NTFS, FAT, and exFAT.

Automatic Installation of Virtual Server Agent or MediaAgent

Commvault automatically installs the Virtual Server Agent or the MediaAgent when those packages are required for an operation:

  • When a live browse is performed, to access file and folder information for a virtual machine backup that does not contain metadata, the Virtual Server Agent is automatically installed to the MediaAgent that is used for the browse operation.

  • When a live browse for a Windows guest VM uses a Linux MediaAgent and the VSA proxy that is used for the live browse does not have the MediaAgent package, Commvault automatically installs the MediaAgent on the VSA proxy.

The remote installation restarts CVD services on the MediaAgent or VSA proxy and does not check for running jobs being handled by the MediaAgent or VSA proxy. As a result, jobs that were running might go Pending. After the install software job completes, any affected jobs restart automatically.

To disable automatic installations, configure the DisableAutomaticPushInstall additional setting.


If the automatic installation of packages to clients is disabled, some on-demand features may not work as expected.

Cloud Library Support for Live Browse

For all VSA hypervisors that support live browse, you can perform live browse operations with the following cloud libraries:

  • Amazon S3: S3 Standard storage class

  • Google Cloud Storage: Region or multi-region storage class

  • Microsoft Azure Storage: Default container storage using hot or cold access tiers with General Purpose v1 (GPv1) or General Purpose v2 (GPv2) storage account