If you are reading this article, then presumably you already know what a metaLUN is. But then again, you could be an inquisitive type who is determined to know everything there is to know about ESX and may only have an inkling of an idea about the concept of metaLUNs. So I ask the pardon of the metaExperienced, but I must take a moment to describe metaLUNs and why they are useful for those not "in the know."
Although the prefix "meta" is often attached to constructs in order to denote their ethereal complexity, metaLUNs are incredibly easy to understand. A metaLUN combines two or more LUNs into a single logical unit. This single logical unit is not a new LUN, but rather an extension of the first LUN in the metaLUN.
Suppose that a Windows 2003 file server uses a 200 GB LUN (LUN A) and now the server needs additional storage. The SAN administrator carves a 100 GB LUN (LUN B) for the file server. Instead of presenting a new LUN to the file server, LUN A and LUN B are metaLUNed. The file server now sees LUN A with 100 GB of free space at the end of the disk. The file server administrator can easily claim this free space by extending the existing NTFS partition to the end of the disk, utilizing the additional 100 GB of free space.
Although a SAN incurs a performance penalty for using metaLUNs, some SAN administrators opt to use this technology to divorce administrators of attached servers from the hassle of having to integrate new disks into their system
While administrators of Windows and Linux servers can grow their partitions into free space, ESX administrators cannot. ESX sees the free space at the end of a LUN after a metaLUN operation, but any existing VMFS-2 volume on that LUN cannot be grown into that free space.
In the VMware world, a volume is similar to a Windows or Linux partition. A VMFS-2 volume can be extended onto a new SCSI device (LUN) with the VMFS tools command line utility, vmkfstools. The syntax for this operation is
vmkfstools -Z <extension-SCSIDevice> <target-SCSIDevice>. In this case the SCSI devices are the LUNs presented to the ESX server. This command will extend the VMFS-2 volume on the target LUN onto the extension LUN. It appears that you would be able to extend a VMFS-2 volume into free space at the end of a LUN simply by specifying the same LUN for the two arguments that the command takes. But if you attempt this, an error occurs,
FSS_Extend: Invalid argument. As I stated earlier, a VMFS-2 volume cannot can be grown into the free space created at the end of a disk that is created by a metaLUN operation.
Although it may be disappointing to administrators who are used to growing partitions to the end of a disk, there are other options. In lieu of growing a VMFS-2 volume to the end of the disk, you simply can create a new VMFS-2 volume in the new free space. This is a perfectly valid option and by far the easiest way for ESX to leverage the free space created by a metaLUN.
Another option is to not use metaLUNs at all, but instead to present a new LUN to the ESX server and then use the vmkfstools command to extend an existing VMFS-2 volume onto the new LUN. But VMware recommends strongly against extends to begin with, and they are common topics of heated conversation on the VMTN discussion forums. Because of their controversy, it may be best to avoid VMFS-2 extents all together and simply create a separate VMFS-2 volume on the new LUN.
In conclusion, we discussed what metaLUNs are and why they are useful. Although ESX 2.5 and VMFS-2 cannot take advantage of metaLUNs with the transparency of Windows or Linux, I hope that I have shown how to make ESX 2.5, VMFS-2, and metaLUNs work for you.
Note: The tests in this document are based on a Dell/EMC CX300 SAN with FLARE 19.
Andrew Kutz has been professionally involved in the technology sector for 11 years. For the last six of them, he has worked with the latest technologies while employed by the University of Texas at Austin. He is an avid fan of VMware, .NET, open source, Terminal Services, coding and comics. He graduated from the University of Texas at Austin, BA in Ancient History and Classical Civilization. He currently lives in Austin, Tex., with his wife Mandy and their two puppies, Lucy and CJ.
This was first published in January 2007