Cluster Share Volumes (CSV)
Abstract- This article will deal with the CSVs, how they work and some things to take into account when implementing them.
Index Terms- CSV, LUN,
I. INTRODUCTION
This document presents an idea about CSV, a unique application of Windows Server 2008 R2, which is optimized by Hyper-V, for a better understanding of its handling and to be able to interpret it in a better way.
II. PROCEDURE
A. Cluster Shared Volumes (CSV)
Cluster Shared Volumes (CSV) works as a distributed access file system optimized for Hyper-V. A comparison that we could easily make would be a clustered file system, in addition to taking the view of storage, a CSV is a LUN that has been used in a Microsoft Cluster Failover for life.
CSVs are not enabled by default in the cluster, simply go to the properties of the cluster and activate them, when doing this process we will realize that as I said above has only been developed to support Hyper-V.
Cluster Shared Volume (CSV) allows fast migration of Virtual Machines between different nodes of a Cluster what we will call Live Migration and facilitates storage management. Before the introduction of Cluster Shared Volume (CSV), a Virtual Machine configured in High Availability, required its own or own LUNs, this also implied that the more Virtual Machines we had in High Availability, the more LUNs we would need to present to the Hyper Host -V; This really affected a lot, but thanks to that they had to add this technology to Windows Server 2008 R2, facilitating many things.
Cluster Shared Volumes (CSV) provides many advantages, including easier storage management, greater resistance to failures, the ability to store many virtual machines on a single LUN, and individually fail over; In particular, CSV provides the infrastructure to support and enhance live migration of Hyper-V virtual machines.
B. Operation
CSV operates by reorienting the I / O operations from the node with the ownership of the virtual machine to the node with the ownership of the LUNs. I / O operations are passed between the nodes through Server Message Block. All operations of all active nodes must be passed through SMB to the node co-coordinator, then passed back to shared storage via shared SCSI, iSCSI, Fiber Channel or multi-method storage initiator Others.
The following figure shows how CSV works from Windows Server 2008 R2 by associating a LUN to more than one server (cluster node) either via Fiber Chanel or iSCSI by configuring an active / passive scenario.
C. Frequent Problems
Network I / O redirection has another interesting application and is the ability for a node in the cluster to keep moving a virtual machine even though the paths of that node with the storage are lost. In other words, if we cut the fiber or Ethernet cables that connect a node to the shared storage, it will route the I / O through the network to the CSV coordinator so that it is the one that writes it. It is not a desirable situation for a long time, but it will give us time to put that node in maintenance mode and evacuate the virtual machines by Live Migration to other nodes that work properly to repair the problem.
One consequence is that if we need to copy files to a CSV, for example to provision a new VM, we must do so from the node that is its coordinator at that moment; So we get the best performance and, above all, we will avoid saturating the internal network with the extra traffic produced by the copy. Of course, it is best to delegate this type of consideration to System Center Virtual Machine Manager.