Site Recovery Manager determines the composition of a datastore group by the set of virtual machines that have files on the datastores in the group, and by the devices on which those datastores are stored.

When you use array-based replication, each storage array supports a set of replicated datastores. On storage area network (SAN) arrays that use connection protocols such as Fibre Channel and iSCSI, these datastores are called logical storage units (LUN) and are composed of one or more physical datastores. On network file system (NFS) arrays, the replicated datastores are typically referred to as volumes. In every pair of replicated storage devices, one datastore is the replication source and the other is the replication target. Data written to the source datastore is replicated to the target datastore on a schedule controlled by the replication software of the array. When you configure Site Recovery Manager to work with a storage replication adapter (SRA), the replication source is at the protected site and the replication target is at the recovery site.

A datastore provides storage for virtual machine files. By hiding the details of physical storage devices, datastores simplify the allocation of storage capacity and provide a uniform model for meeting the storage needs of virtual machines. Because any datastore can span multiple devices, Site Recovery Manager must ensure that all devices backing the datastore are replicated before it can protect the virtual machines that use that datastore. Site Recovery Manager must ensure that all datastores containing protected virtual machine files are replicated. During a recovery or test, Site Recovery Manager must handle all such datastores together.

To achieve this goal, Site Recovery Manager aggregates datastores into datastore groups to accommodate virtual machines that span multiple datastores. Site Recovery Manager regularly checks and ensures that datastore groups contain all necessary datastores to provide protection for the appropriate virtual machines. When necessary, Site Recovery Manager recalculates datastore groups. For example, this can occur when you add new devices to a virtual machine, and you store those devices on a datastore that was not previously a part of the datastore group.

A datastore group consists of the smallest set of datastores required to ensure that if any of a virtual machine's files is stored on a datastore in the group, all of the virtual machine's files are stored on datastores that are part of the same group. For example, if a virtual machine has disks on two different datastores, then Site Recovery Manager combines both datastores into a datastore group. Site Recovery Manager combines devices into datastore groups according to set criteria.

Two different datastores contain files that belong to the same virtual machine.

Datastores that belong to two virtual machines share a raw disk mapping (RDM) device on a SAN array, as in the case of a Microsoft cluster server (MSCS) cluster.

Two datastores span extents corresponding to different partitions of the same device.

A single datastore spans two extents corresponding to partitions of two different devices. The two extents must be in a single consistency group and the SRA must report consistency group information from the array in the device discovery stage. Otherwise, the creation of protection groups based on this datastore is not possible even though the SRA reports that the extents that make up this datastore are replicated.

Multiple datastores belong to a consistency group. A consistency group is a collection of replicated datastores where every state of the target set of datastores existed at a specific time as the state of the source set of datastores. Informally, the datastores are replicated together such that when recovery happens using those datastores, software accessing the targets does not see the data in a state that the software is not prepared to deal with.

Not all SRAs report consistency group information from the storage array, because not all storage arrays support consistency groups. If an SRA reports consistency group information from the array following a datastore discovery command, the LUNs that constitute a multi-extent VMFS datastore must be in the same storage array consistency group. If the array does not support consistency groups and the SRA does not report any consistency group information, Site Recovery Manager cannot protect virtual machines located on the multi-extent datastore.