Guest MPIO Support?

[hutchingsp]

Can anyone point me to vmware's (official documented) position on using MPIO within guests?

I'm looking at P4000 and HP have said they don't support their DSM MPIO within a VM (this would be Windows 2003/2008 R2 x64) because vmware don't support it, but apparently Windows inbuilt MPIO is supported - can anyone confirm this?

Thanks very much.

[FranckRookie]

Hi Hutchingsp,

There was a thread about that question a few month ago (you can fin it here). Apparently, VMware do not support any third-party MPIO in a virtual machine. But we were not able to find any official document on that point.

Sorry.

Good luck.

Regards

Franck

[thakala]

There is absolutely no need to use MPIO within guests, guests do not see multiple paths to storage. ESX will handle all your MPIO needs, either with native multipathing plugin or with 3rd party plugin such as EMC PowerPath/VE.

http://v-reality.info

[hutchingsp]

There is absolutely no need to use MPIO within guests, guests do not see multiple paths to storage. ESX will handle all your MPIO needs, either with native multipathing plugin or with 3rd party plugin such as EMC PowerPath/VE.

I don't understand what you mean.

If I have a File Server that is a VM, and I want to use the MS iSCSI initiator to mount volumes from the VM, it's a potential disadvantage not to be able to use Windows MPIO.

[thakala]

If I have a File Server that is a VM, and I want to use the MS iSCSI initiator to mount volumes from the VM, it's a potential disadvantage not to be able to use Windows MPIO.

Is there any particular reason why you need to mount iSCSI target directly to a VM? I would mount iSCSI target on ESX host and present storage to a VM either as VMDK or RDM, this way MPIO will be handled by ESX.

Tomi

http://v-reality.info

[hutchingsp]

If I have a File Server that is a VM, and I want to use the MS iSCSI initiator to mount volumes from the VM, it's a potential disadvantage not to be able to use Windows MPIO.

Is there any particular reason why you need to mount iSCSI target directly to a VM? I would mount iSCSI target on ESX host and present storage to a VM either as VMDK or RDM, this way MPIO will be handled by ESX.

Many SAN's these days take application aware snapshots (Exchange/SQL etc.) and to do that they need to quiesce the application.

You can't do this if the databases/logs are on a VMDK.

[FranckRookie]

In this case, either you design your VM storage network for redundancy as recommended by VMware, or you use MPIO inside your VM but you are not in a supported configuration...

[thakala]

For that there is RDM, raw device mapping, which you can use to present native LUNs from storage array through ESX to a VM.

Tomi

http://v-reality.info

[hutchingsp]

For that there is RDM, raw device mapping, which you can use to present native LUNs from storage array through ESX to a VM.

Tomi

http://v-reality.info

Thank you! I use RDM now but on a very "dumb" SAN.

So with vSphere 4.1 I could use vSphere's iSCSI initiator and allocate LUNs to VM's as RDM and get the benefit of vSphere's MPIO and still use vendor specific VSS provider to take application aware snapshots of things such as Exchange?

[thakala]

I have no first-hand experience on this issue but if you provision RDM in physical mode guest sees original storage device type as is, I would guess that vendor specific VSS provider should work fine in that situation. If you use physical mode RDM you cannot use vSphere snapshots or create a clone of that VM.

Tomi

http://v-reality.info

[FranckRookie]

If you want redundancy for the link to your storage, the supported solution looks to be the following one:

- have an iSCSI LUN presented to the hosts via several paths

- create a virtual disk on that LUN

- attach that disk to your virtual machine

This way, you rely on the VMware layer to insure multipathing and redundancy...

Franck