What is a Cluster Shared Volume?

A Cluster Shared Volume (CSV) is a volume that is simultaneously available to directly read from and write to by all nodes in a Microsoft® Failover Cluster. This capability, referred to as Direct I/O, is made possible by providing a distributed access file system, which allows each node to utilize its storage interconnect—Internet SCSI (iSCSI), Fibre Channel, Serial Attached SCSI (SAS)—for communication with the volume on the shared storage array. In addition, CSVs provide the capability to reroute data over the network in the event that a node loses its path(s) to the shared storage array.

Why should I implement CSVs?

In addition to the increased availability (see the "How do CSVs work?" question), CSVs provide the ability for multiple virtual machines (VMs) to share a single shared storage volume while also allowing these VMs to migrate (live or quick) independently between hosts in the cluster. This ability addresses the limitation that existed in the initial release of Hyper-V in which it was recommended that each VM have its own dedicated LUN to allow for independent quick migration of each VM.

How do CSVs work?

There is a coordination node with each CSV that owns the shared storage volume and the NT Files System (NTFS) partition for all nodes in the cluster. This node can be thought of as the resource owner in the traditional Microsoft cluster model. In a fully functional state (no failed paths), a non-coordination node is capable of sending data directly to the volume over the storage interconnect (Fibre Channel, iSCSI, SAS) and the file system metadata is routed over the private network to the coordination node that will make any updates to the NTFS partition. If a non-coordination node loses its path(s) to storage, the data can be rerouted over the private network, and the coordination node will perform the read/write to volume through its path to the storage array. If a coordination node loses its path(s) to storage, a new coordination node is identified and the disk resource ownership is transferred to that node. These capabilities are made possible by a filter driver that sits above the NTFS partition on all nodes in the cluster. This filter driver is able to route both metadata and data over the network.

Does the CSV filter driver model impact any applications?

The CSV filter can impact applications outside of the VM that directly access files such as antivirus and data protection (backup/restore) applications. In addition, there have been changes to the VSS framework in R2 to support CSVs. Most data protection software that touts Hyper-V support provides that capability through VSS-based snapshots, and you want to ensure that those applications specifically provide CSV support.

Can I make my existing traditional volume a CSV?

Yes. However, the path will change, so any applications, services, and so on that rely on that path will be broken. For example, if you have a VM that resides on the volume, then you should export the VM, make the volume a CSV, and then import the VM.

What needs to be in place to support live migration and CSVs?

• All nodes are running one of the following OS versions:
  
  • Microsoft Windows Server® 2008 R2 x64 Enterprise Edition
  • Microsoft Windows Server 2008 R2 x64 Datacenter Edition
  • Microsoft Hyper-V Server 2008 R2
• All hardware and software are supported with 2008 R2
• A failover cluster must be configured, and the solution must pass all of the tests in the validation wizard. For more details, refer to: http://technet.microsoft.com/en-us/library/cc771404.aspx.
• All nodes in the cluster must use the same processor or processors from the same manufacturer.
• All nodes must have access to the access to the shared storage devices (LUNs that reside on iSCSI, Fibre Channel, and SAS storage arrays, such as the Dell™ PowerVault™ MD3000/3000i, Dell EqualLogic™ PS Series arrays, or Dell/EMC arrays).

Are CSVs required to support live migration?

No, CSVs are not required to support live migration; however, the capabilities provided with CSVs make it very compelling to implement them with any R2 implementation that is targeting to utilize live migration.

Where can I get more details on CSVs?

• You can ask questions by starting a thread on this page.
• In addition, Microsoft has an article on their Technet site: http://technet.microsoft.com/en-us/library/dd630633(WS.10).aspx