VMFS recovery

VMFS is a filesystem designed by VMware Company to be used in ESX servers and vSphere. Because VMFS is designed to mostly store disk images for ESX virtual machines, it is tuned for small number of large files. Disk space for a VMFS volume can be supplied either as a single disk or as several disks (or hardware RAIDs). The VMFS driver, in turn, combines available disk space extents into a JBOD.

VMFS metadata

In VMFS, metadata is located in the beginning of the volume or, in case of several extents, in the beginning of the first extent. This metadata includes inodes, directories, and repository for small files.

To store disk images of virtual machines, VMFS uses 1 MB clusters, which is larger than any other common filesystem uses. Thus, a file could not occupy less than 1 MB if there were no special approaches used for small files. Starting with VMFS 5, files smaller than 1 KB are stored directly in their corresponding inodes, much like resident files in NTFS. For files larger than 1 KB, VMFS creates one large file and cuts it into 8 KB clusters which are then used to store small files. All these approaches allow to reduce disk space usage for small files. It seems that directories, no matter the size, are stored neither inside inodes nor in the small files repositories. A directory always occupies at least 1 MB of disk space.

All VMFS metadata is typically stored within first 3 to 5 GB of the VMFS volume meaning that the full VMFS structure can be recovered just by scanning only first 5 GB of the VMFS volume.

VMFS recovery approaches

There are two steps in VMFS data recovery — assembling a VMFS volume from the extents and filesystem recovery on the assembled VMFS volume. The VMFS recovery capability in ReclaiMe Pro allows to recover data from a VMFS volume both in simple and complex cases like those requiring preliminary RAID recovery.

Since VMFS is often used on one or several hardware RAIDs, the first thing in VMFS recovery is to recover the correct VMFS storage configuration — detect configuration of one or several RAID arrays composing a VMFS volume using the RAID recovery feature of ReclaiMe Pro.

Once VMFS extents are assembled correctly, ReclaiMe Pro automatically starts searching the assembled storage for VMFS volumes and displays all the found volumes along with all their components in the Disk and partition list under the VMFS section. Then you need to select the needed VMFS volume and click Start scan.

If the VMFS volume cannot be assembled in automatic mode, you can assemble it manually from the segments. To do this, use Find partitions for each component of the VMFS volume:

  • If a volume was created over separate disks, use Find partitions for each disk and add the found VMFS partitions into a list of available devices.
  • In case of RAID, use Find partitions on the RAID rather than on separate disks. Once all the VMFS extents are found, select them, click Manual RAID, sort all the extents according to their first segment numbers (which are provided during Find partitions process) in ascending order, and combine them into a JBOD. After that, you can start filesystem recovery by clicking Start scan.