Use flash devices and magnetic disks based on the requirements for Virtual SAN.

Verify that the cluster has the capacity to accommodate anticipated virtual machine consumption and the Number of failures to tolerate in the storage policy for the virtual machines.

The storage devices must meet the following requirements so that Virtual SAN can claim them:

  • The storage devices are local to the ESXi hosts. Virtual SAN cannot claim remote devices.
  • The storage devices do not have any preexisting partition information.
  • On the same host, you cannot have both all-flash and hybrid disk groups.

Prepare Devices for Disk Groups

Each disk group provides one flash caching device and at least one magnetic disk or one flash capacity device. The flash caching device must provide at least 10 percent of the anticipated storage consumed on the capacity devices, not including replicas such as mirrors.

Virtual SAN requires at least one disk group on a host that contributes storage to a cluster that consists of at least three hosts. Use hosts that have uniform configuration for best performance of Virtual SAN.

Raw and Usable Capacity

Provide raw storage capacity that is greater than the capacity for virtual machines to handle certain cases.

  • Do not include the size of the flash caching devices as capacity. These devices do not contribute storage and are used as cache unless you have added flash devices for storage.
  • Provide enough space to handle the Number of failures to tolerate value in a virtual machine storage policy. A Number of failures to tolerate that is greater than 0 extends the device footprint. If the Number of failures to tolerate is set to 1, the footprint is double. If the Number of failures to tolerate is set to 2, the footprint is triple, and so on.
  • Verify whether the Virtual SAN datastore has enough space for an operation by examining the space on the individual hosts rather than on the consolidated Virtual SAN datastore object. For example, when you evacuate a host, all free space in the datastore might be on the host that you are evacuating and the cluster is not able to accommodate the evacuation to another host.
  • Provide enough space to prevent the datastore from running out of capacity, if workloads that have thinly provisioned storage start consuming a large amount of storage.
  • Verify that the physical storage can accommodate the reprotection and maintenance mode of the hosts in the Virtual SAN cluster.
  • Consider the Virtual SAN overhead to the usable storage space.
    • On-disk format version 1.0 adds an additional overhead of approximately 1 GB per capacity device.
    • On-disk format version 2.0 adds an additional overhead, typically no more than 1-2 percent capacity per device.
    • On-disk format version 3.0 and later adds an additional overhead, typically no more than 1-2 percent capacity per device. Deduplication and compression with software checksum enabled require additional overhead of approximately 6.2 percent capacity per device.

For more information about planning the capacity of Virtual SAN datastores, see the VMware Virtual SAN Design and Sizing Guide.

Virtual SAN Policy Impact on Capacity

The Virtual SAN storage policy for virtual machines affects the capacity devices in several ways.

Table 1. Virtual SAN VM Policy and Raw Capacity
Aspects of Policy Influence Description
Policy changes
  • The Number of failures to tolerate influences the physical storage space that you must supply for virtual machines. The greater the Number of failures to tolerate is for higher availability, the more space you must provide.

    When a Number of failures to tolerate set to 1, it imposes two replicas of the VMDK file of a virtual machine. If the Number of failures to tolerate is set to 1, a VMDK file that is 50 GB requires 100 GB space on different hosts. If the Number of failures to tolerate is changed to 2, you must have enough space to support three replicas of the VMDK across the hosts in the cluster, or 150 GB.

  • Some policy changes, such as a new number of disk stripes per object, require temporary resources. Virtual SAN recreates the new objects that are affected by the change and for a certain time, the physical storage must accommodate the old and new objects.
Available space for reprotecting or maintenance mode When you place a host in maintenance mode or you clone a virtual machine, although the Virtual SAN datastore indicates that enough space is available, the datastore might not be able to evacuate the virtual machine objects because the free space is on the host that is being placed in maintenance mode.