vMotion with Raw Device Mappings ESX 4.1?

[HughBorg707]

Is this possible yet?   I have a customer running a management VM on an HBA equipped server that needs to talk to an EMC backend. There are two ESX hosts in the cluster with HBA cards.

Is this a good candidate for FT?

Thanks

Hugh

[a_p_]

I think we are talking about different things here.

Is this possible yet?

Yes.

I have a customer running a management VM on an HBA equipped server that needs to talk to an EMC backend

In this case - if I'm not mistaken - we are talking about vmDirectPath.

André

[DSTAVERT]

vMotion with RDM

Fault Tollerance requirements

Note that only Virtual RDM is supported.

[HughBorg707]

Thanks for the replies. I am going to try this per an engineer at EMC's recommendation:

Here’s a VMware KB article about another issue that can cause Vmotion to fail with RDMs:

http://kb.vmware.com/selfservice/microsites/microsite.do -> Article ID 1016210

Symptoms

• Unable to vMotion virtual machines with RDMs
  
• vMotion compatibility check between certain hosts may fail even though all hosts see every single LUN
  
• vMotion of virtual machines with RDMs fails verification of the RDM disks with the error:
  
  "Virtual Disk 'hard disk 0' is a mapped direct access LUN and its not accessible"
  
  
  
• A virtual machine with RDMs running on host A cannot vMotion to host B. If you shut down the virtual machine on host A and power it up on host B, vMotion still fails with the above error.
  
• If you shutdown the virtual machine and remove and recreate the RDM pointer files, vMotion still fails the verification test with the above error when you try to vMotion to another host
  

Resolution

vMotion compatibility for RDM LUNs is dependent on the vml identifier instead of the unique identifier (such as the NAA, EUI, or T10) for the LUN. If the LUN presentation for that LUN is not consistent across all ESX hosts, vMotion compatibility may fail. Cold migration of that virtual machine does not fail.

A vml mismatch can occur, for example, if you present the LUNs as different LUN numbers to the ESX hosts or by presenting the LUN to the second HBA only instead of both HBAs.

Here is an example where the vml's between two hosts do not match:

• How RDM LUNs for virtual machine RDMTest1 are visible on esxhost1:
  
  vml.02000500006006048000019010406353303035344553594d4d4554 -> naa.60060480000190104063533030353445
  vml.02000600006006048000019010406353303035344653594d4d4554 -> naa.60060480000190104063533030353446
  vml.02000100006006048000019010406353303035353053594d4d4554 -> naa.60060480000190104063533030353530
  vml.02000200006006048000019010406353303035353153594d4d4554 -> naa.60060480000190104063533030353531
  
  
  
• How RDM LUNs for the same virtual machine are visible on esxhost2:
  
  vml.02004d00006006048000019010406353303035344553594d4d4554 -> naa.60060480000190104063533030353445
  vml.02004e00006006048000019010406353303035344653594d4d4554 -> naa.60060480000190104063533030353446
  vml.02004f00006006048000019010406353303035353053594d4d4554 -> naa.60060480000190104063533030353530
  vml.02005000006006048000019010406353303035353153594d4d4554 -> naa.60060480000190104063533030353531
  

Note: You can view this information under /vmfs/devices/disks with the command: ls -la

As you can see from the above example, the red characters indicate the 5th and 6th characters of the vml do not match between the hosts.

Ensure that you present the RDM LUNs by the same LUN number and to all HBAs in a single ESX host.

To resolve this issue:

1. Shut down the virtual machine.
   
2. Remove the RDMs from the virtual machine.
   
3. Unpresent the LUNs from the SAN.
   
4. Rescan the storage adapters on each ESX host.
   
5. Represent the LUNs.
   
   
   Note: Ensure that the RDM LUN is presented to all ESX hosts to which the virtual machine is being vMotioned.
   
   
   
   
6. Rescan the storage adapters on each ESX host.
   
7. Re-add the RDMs to the virtual machine.