[SOLVED] NPIV (vHBA) on proxmox

[alj]

Hello All,

Before buying hardware and support for proxmox I am testing proxmox with a 3-node HA cluster using Ceph. Using hand me down hardware, this is working VERY nicely. I look forward to working with it on better hardware.

I now have a vm that will be running an application (Hitachi CCI) that needs access to a command device (LUN) presented from the Hitachi storage controller on our SAN to the vm.

My preference is to use NPIV as my background is AIX on Power and NPIV works flawlessly in that environment. Unfortunately from what I read on Proxmox and Qemu, NPIV does not work. I've seen a few forum posts asking about it, and links in those posts that reference other sites with documentation that refers to virsh commands. This does not exist on proxmox and from another dev post in the forum I see libvirt is not used with proxmox. I thought I saw one that said it's incompatible but I can't find it now.

I was able to present the command device to the 3 nodes and using block passthrough presented the command device to the guest. This works but unfortunately guest migration does not work with the local devices mapped.

This is a show stopper for deploying proxmox in my environment. Which would be a shame because for things that don't need something special like Brocade Network Advisor, running on proxmox is working beautifully!

Is there another way to allow a vm raw device access to either lun or block and allow vm migration between hosts that I am missing?

Is NPIV/vHBA anywhere on the roadmap for proxmox?

Thanks in advance,

Al

 

[wolfgang]

Hi,

Is NPIV/vHBA anywhere on the roadmap for proxmox?

No this is not on the roadmap and it is not necessary.
You have to use the lun as local blockdev as you did,
Then install an LVM on to of this dev mark it as shared and then you can do a migration.

 

[alj]

Hello Wolfgang,

Thank you, I will try this a little later this morning.

Have a good day,

Al

 

[alj]

I removed the block device mapping from the vm and created the LVM configuration. All nodes can now see the storage. I'm confused now as to how to present a block device from this backend LUN. Do I create a disk on the LVM pool and present it to the node?

 

[alj]

I think I've got it figured out, I'm on the last step, which is to create a hard disk for the vm. The problem is that the LUN can't be larger than 1G, and I can't specify a disk size in the GUI of MB.

 

[wolfgang]

When you create the LVM on the disk you must add the LVM as Storage to PVE and mark it as shared.
You do not use it as iSCSI LUN.

 

[alj]

Understood on iSCSI. I was able to add the LVM to PVE and mark it as shared, it showed up on all nodes. The problem is that I can't allocated a disk of less than 1GB in the GUI. And I can't make the Hitachi raidcom command device more than 1GB in size.

Is there a way in the command line to create the disk/LV and then add it to the vm?

 

[wolfgang]

Then you can create the image with
"pvesm alloc"
and create the VM with
"qm create"
and add the manually created image in the qm create call

 

[alj]

I was able to create a larger LUN on the Hitachi controller, and then created a 1G LVM/Disk assigned to the VM, which migrates between hosts without any problem at all.

Too bad NPIV isn't supported, it would have been so much simpler

Thank you for the help, it is much appreciated. Hopefully I'll be able to sell it to my manager as a useful tool rather than getting pushed to VMWare.

Have a good day!