vSphere 4.1 - trouble to make NPIV working

[bobst_martin]

Hi,

TARGET : create a Win2003 R2 ENT CLUSTER with 2 nodes: 1 physical + 1 Virtual

it seems that something is wrong on my config, but I can't found what

4 ESX 4.1 with 2 QLE2460 each

cat /proc/scsi/qla2xxx/3 or cat /proc/scsi/qla2xxx/4 .... "NPIV Supported : YES" but no entry with "Virtual Port Information:"

from the brocade switch : switchshow from switch 1

ESX#1 is connected on port 19

ESX#3 is connected on port 17

17  17   162500   id    N4   Online      FC  F-Port  1 N Port + 0 NPIV public
18  18   162300   id    N2   Online      FC  F-Port  21:00:00:e0:8b:0f:da:d3
19  19   162100   id    N4   Online      FC  F-Port  1 N Port + 0 NPIV public

I can't found any WWN VMWARE on my switch

reading many post, I discover a workaround: removed NPIV, and then Generate new NPIV

oh very surprising, while I poweronthe VM, I run many times (every second) "cat /proc/scsi/qla2xxx/3"

and suddenly I can read :

Virtual Port Information:

Virtual Port WWNN:WWPN:ID =282a000c2900000b:282a000c2900000c:162501;
Virtual Port 1:VP State = <ACTIVE>, Vp Flags = 0x0

and It deseapeared one second later.

How do I create my VM :

1/ create VM with 1 disk SCSI0:0 for Win2003 R2 entreprise system

2/ create LUN 250MB on Clariion VNX5500 storage

3/ presenting LUN 250MB to StorageGroup ESX with 4 Hosts : ESX#1 to ESX#4

  ( StorageGroup ESX is already carrying 7 VMFS LUN)

3bis/ presenting LUN 250MB to an other StorageGroup TEST with 1 host : physical server running Win2003 R2 entreprise system on a local disk

4/ power off VM

5/ EDIT VM : enable NPIV and add RDM disk as Virtual with SCSI1:0

5bis/ EDIT VM : edit SCSI Controller #2 : change BUS SHARING to Physical (needed for MSCS)

REM:

MSCS cluster is working well, using CLUADMIN  I can MOVE my QUORUM LUN 250MB from physical node to VM node

I can't VMOTION : SCSI controller 1 is engaged in bus sharing

VM powered OFF, I can poweron on any host ESX#1 or ... ESX#4

more questions :

my ESX host is holding 2 HBAs, HBA#1 is connected to FABRIC#1, HBA#2 is connected to FABRIC#2.

what do I need to do with NPIV ?

generate only 1 WWn, it could be visible from 2 fabrics ?

generate 2 WWn, and attached WWn#1 to physical HBA#1 .... #2 to #2 ? how to do that ?

any help appreciate, it's my first NPIV

[Texiwill]

Hello,

You do not need to use NPIV actually, NPIV in vSphere is backed by an RDM so you had to create an RDM to even use NPIV. Interesting note, NPIV is only there so that you know to which VM an RDM is mapped when looking at things from the storage side. There is no real impact when using NPIV from the vSphere side. For my MSCS clusters I just use RDMs.

NPIV does not give you a virtual HBA capability or a way to directly connect a VM to an HBA, it is a software construct to allow refined Zoning and Presentation from Storage at least in vSphere.

Best regards,

Edward L. Haletky

Communities Moderator, VMware vExpert,

Author: VMware vSphere and Virtual Infrastructure Security,VMware ESX and ESXi in the Enterprise 2nd Edition

Podcast: The Virtualization Security Podcast Resources: The Virtualization Bookshelf

[bobst_martin]

Hi Texiwill,

you said : For my MSCS clusters I just use RDMs.

can you use Vmotion with your RDM ? because I can't , VMWARE support answer me : Vmotion do not support VM with RDM.

can you describe how have you created your MSCS

THX

[a_p_]

VMware provides detailed documentations for setting up MSCS (http://kb.vmware.com/kb/1004617). Please take a look at the document according to your version of vSphere and make sure you pick the correct setup (cluster-across-boxes, cluster-in-a-box, ...)

Although it is possible to vMotion VMs with RDMs, this is not supported for MSCS nodes where the RDMs are used as shared storage (http://kb.vmware.com/kb/1037959)

André

[bobst_martin]

Thanks

I already found the first link "setup mscs". but without screenshot, It's not a very clear document to me.

anyway, according to the second link, I'm trying to create something no supported by VMWARE.

what a pity...

thx for all

Cordialement

Martin

Message was edited by: a.p. - removed personal data from mail reply

[bobst_martin]

----------  MODIFIED  -------------

sorry email send to forum instead of french vmware support

------------------------------------------

[bobst_martin]

Finaly, with the help of VMWARE support, we did it.

problem come from zoning / LUN declaration

as ESX has 2 HBA 1 by fabric, Clariion has 6 FC ports 3 per fabric, then each esx has 6 paths to clariion.

zoning for VM must have also 6 paths

On Clariion declaration must be similar as zoning (6 path to declare manualy with a specific parameter : FAILOVER MODE 1 (ACTIVE/PASSIVE)

for clariion, LUN MASKING need to put the VM in same STORAGE GROUP as ESX, all LUN (RDM, VMFS, ...) must be presented to both ESX and VM

and then, ESX is working as a "FILTER" to present the righr RDM thru NPIV to the right VM

my conclusion : we can not managed VM like physical server (from storage view) NPIV is not a virtual HBA inside a VM

I was expecting a Fiber channnel switch inside ESX.

thanks to all.