NPIV VM Requirements

[heyyou82]

Unfortunately at this time I don't have NPIV capable HBAs to test this out, but we are beginning to plan for our next iteration of our environment and are looking into NPIV.

From the VMware "Fibre Channel SAN Configuration Guide" page 91

"NPIV can only be used for virtual machines with RDM disks. Virtual machines with regular virtual disks use the WWNs of the host’s physical HBAs. For more information on RDMs, see the ESX Server 3 Configuration Guide or ESX Server 3i Configuration Guide."

Does this mean that NPIV can only be used on a VM that exclusively uses RDM disks? I.E. can you not have a VM with regular disks stored on a VMFS and then an additional SCSI adapter attached to the VM using NPIV to present storage across the SAN directly to the VM?

Also, from what I am reading, with NPIV you still need to present the LUNs from the storage device to ALL ESX hosts as well as the VM's WWN as well. Am I reading this correctly?

Maybe I had unrealistic hopes for NPIV: I was hoping space from the storage appliance on the SAN could be presented directly to the VM, and only the VM, by attaching a virtual HBA to the VM, which then would participate on the SAN as any other HBA. This would help in cases where a VM has a standard System and Application Drive (low disk space, can make use of snapshots, cloning, vRanger backups, etc) and then a third very large drive that can be presented and maintained from the storage side directly. In these cases it isn't fun to manage large VMFS volumes to hold the virtual machine data drives, nor is it fun to manage a whole bunch of LUNs to be used as RDMs to the VMs.

Thoughts?

[hugop]

+Does this mean that NPIV can only be used on a VM that exclusively uses

RDM disks? I.E. can you not have a VM with regular disks stored on a

VMFS and then an additional SCSI adapter attached to the VM using NPIV

to present storage across the SAN directly to the VM?+

Your VM can still use a .vmdk for it's system drive and you can still give it an RDM for it's data disk and use NPIV on this second disk.

+Also, from what I am reading, with NPIV you still need to present the

LUNs from the storage device to ALL ESX hosts as well as the VM's WWN

as well. Am I reading this correctly?+

Yes. Once you enable NPIV on your ESX server, you can generate virtual WWPNs for use with NPIV.

[RussH]

I read it that although you can now use SAN tools to manage the RDM based on the NPIV addressv (i.e. zoning)- you cant directly present luns to it (u still need to do this via ESX and RDMs).

Theres a decent doc here on it -

[kjb007]

Does this mean that NPIV can only be used on a VM that exclusively uses RDM disks? I.E. can you not have a VM with regular disks stored on a VMFS and then an additional SCSI adapter attached to the VM using NPIV to present storage across the SAN directly to the VM?

If you are using NPIV, then you are assigning disks using the WWNN of the vm, so by definition, you are using an RDM. If you used regular disks via a VMFS, then RDM and NPIV is not in the picture, since you will use a datastore attached to the ESX host and not the vm.

Also, from what I am reading, with NPIV you still need to present the LUNs from the storage device to ALL ESX hosts as well as the VM's WWN as well. Am I reading this correctly?

This is not correct, since you are using NPIV, you zone to the WWNN of the vm virtual HBA, and not the ESX host.

-KjB

[heyyou82]

That document was very helpful. Based on what the document says and other posts in this thread, it does seem like there is still confusion in regards to Zoning and Presentation.

From the document, it looks like you must zone your ESX servers' HBAs into a SAN fabric Zone (soft or hard) with the storage device you wish to present LUNs from. From the storage device, if it is capable, you then only present the LUN to the WWPN of the virtual machine.

For instance, if you had a JBOD device on the SAN that not only hosts LUNs formatted as VMFS for ESX, but also hosted some LUNs to be presented directly to VMs, from what I understand the only difference would be presenting the LUN to the VM's WWPN, as the ESX servers are already zoned in with the storage device.