VMFS3 vs VMFS5 vs VMFS6: What’s the Difference?

VMware Virtual Machine File System (VMFS) is a clustered filesystem used by VMware ESXi to store virtual machine files on shared storage. Over time VMware introduced several VMFS versions with improvements in scalability, performance, and storage efficiency.

The most common versions encountered during data recovery are VMFS3, VMFS5, and VMFS6.

Quick takeaway VMFS3 is legacy and restrictive, VMFS5 modernized datastore management, and VMFS6 added automatic space reclamation and better support for modern storage hardware.

VMFS3

VMFS3 was widely used in VMware ESX/ESXi 3.x and 4.x environments and is still encountered on legacy storage systems.

Key characteristics

Block size selectable at format time (1 MB, 2 MB, 4 MB, or 8 MB)
Maximum file size depended on the selected block size
Maximum datastore size: 64 TB
Uses SCSI reservations for locking
Metadata structure optimized for older ESX architectures

Typical issues with VMFS3

Block size restrictions

The maximum size of a VMDK depends on the block size selected during formatting. If the datastore was created with a small block size, administrators may not be able to create large virtual disks without recreating the datastore.

SCSI reservation contention

VMFS3 relies heavily on SCSI reservations for locking shared storage. In busy environments this may cause performance problems or temporary storage lockups when multiple hosts attempt metadata operations.

Legacy partition layout

VMFS3 volumes typically use older partitioning layouts, which complicates expansion and migration to modern storage systems.

Upgrade artifacts

When a VMFS3 datastore is upgraded to VMFS5, some legacy structures may remain. These hybrid layouts may complicate administration, migration, and data recovery.

VMFS5

VMFS5 was introduced with VMware vSphere 5 and significantly improved storage management.

Key improvements

Unified 1 MB block size
Maximum virtual disk size: 62 TB
Maximum datastore size: 64 TB
Improved ATS (Atomic Test & Set) locking
Support for GPT partitioning
Simplified datastore management

VMFS5 also removed the need to choose a block size during formatting.

Typical issues with VMFS5

Datastore upgrade inconsistencies

When VMFS3 datastores are upgraded in place to VMFS5, they may retain legacy alignment or metadata layout characteristics. These upgraded volumes may behave differently from newly created VMFS5 datastores.

ATS misconfiguration

VMFS5 relies on ATS locking when supported by the storage array. If the array firmware or configuration does not support ATS correctly, ESXi may fall back to SCSI reservations, causing performance degradation.

Extent-related complexity

Expanding VMFS5 datastores or using multiple extents can complicate management. In recovery scenarios, multi-extent datastores are also more difficult to reconstruct correctly.

Manual space reclamation

Thin-provisioned storage arrays require manual UNMAP operations in many VMFS5 environments. If this is not done periodically, freed blocks may remain allocated on the storage system.

VMFS6

VMFS6 was introduced with vSphere 6.5 and is the current generation VMFS filesystem.

Major enhancements

Automatic space reclamation

VMFS6 supports automatic UNMAP, allowing ESXi to return freed blocks to the storage array automatically.

better thin-provisioned storage efficiency
reduced administrative maintenance

Native 4K disk support

VMFS6 supports modern storage devices with 4K native sector sizes, improving compatibility with newer storage hardware.

Improved metadata handling

Metadata operations were optimized for modern ESXi workloads and larger environments.

Large virtual disks

VMFS6 continues support for virtual disks up to 62 TB.

Typical issues with VMFS6

Automatic UNMAP side effects

In some storage environments, automatic UNMAP can generate additional I/O load, especially on arrays with limited garbage collection performance.

4K sector compatibility

Older storage arrays or RAID controllers may not fully support 4K native disks, leading to datastore mounting problems or degraded performance.

Storage firmware incompatibility

VMFS6 relies more heavily on modern storage features. Firmware mismatches between ESXi and storage arrays can cause datastore accessibility problems.

Wider impact of corruption

Large modern datastores often contain many virtual machines. When storage hardware fails or metadata becomes damaged, the impact may affect multiple VMs at once.

Quick Comparison

Feature	VMFS3	VMFS5	VMFS6
Introduced in	ESX/ESXi 3	vSphere 5	vSphere 6.5
Block size	1–8 MB selectable	Fixed 1 MB	Fixed 1 MB
Max datastore size	64 TB	64 TB	64 TB
Max VMDK size	Up to 2 TB – 512 B (block dependent)	62 TB	62 TB
Automatic UNMAP	No	Limited / manual	Yes
4K disk support	No	Partial	Yes
Locking mechanism	SCSI reservations	ATS + fallback	ATS optimized

VMFS Versions and Data Recovery

When recovering data from a VMFS datastore, the filesystem version is important because metadata structures differ between versions, partition layouts may change, upgrade paths may leave hybrid layouts, and VMFS6 introduces automatic space reclamation behavior.

Recovery tools must correctly interpret these structures to reconstruct:

virtual disk files (VMDK)
virtual machine configuration files
snapshots
deleted or lost virtual machines

How to Identify the VMFS Version

The VMFS version can usually be determined from:

ESXi datastore properties
partition metadata on the disk
VMware management tools
specialized recovery software

Need Help Recovering VMFS Datastores?

If a datastore becomes inaccessible due to disk failure, controller issues, accidental deletion, or metadata corruption, it may still be possible to recover virtual machines directly from the VMFS volume.

Tools designed specifically for VMFS recovery can reconstruct datastore structures and extract virtual machine files even when the ESXi host can no longer mount the datastore.