5 Replies Latest reply on Aug 7, 2013 6:18 PM by Al911

    P2V Cluster

    Al911 Lurker

      Good Afternoon All,


      I have a question in regards to clustering of a Physical server with a Virtual one using ESXi VMFS RDM.


      I have a 3 node Veritas cluster with 3 SQL instances running on Windows 2008 R2 SP1 that will be on RAID-5 VSP owned LUNs. Each server is the fail over for the other one but the order is always physical to physical and lastly to VM.


      If the fail over of an instance occurs to the VM, I have a concern that the Windows NTFS bock size will not be the same from the VM size. As you know NTFS has a 4KB block size for up to 16TB where VMFS has 1MB+.


      My question is that when the LUN(s) are presented to VM via RDM, would VMFS force its blocksize or would OS be the determining factor since VMFS is only a mapping at that point to present LUN via Fiber directly to OS? In addition since Veritas has it abstraction layer as volume manager, wont it stop creation of any other block sizes?


      Thank you!

        • 1. Re: P2V Cluster
          weinstein5 Guru
          User Moderators

          Welcome to the Community - Using an RDM the VM will read the native NTFS drive and its block but your big problem is VMware does not support a 3 node MSCS cluster - it only supports a 2 node cluster.

          1 person found this helpful
          • 2. Re: P2V Cluster
            Al911 Lurker

            Thank you so much for your assistance.


            I am using a Veritas Cluster and not MS cluster. We did build this environment before and VCS can make it happen but can you point me to the direction that mentions 3.5 can only do 2 node clustering for reference.


            So once I present any LUN as RDM, the block size created will be native NTFS that is done by Windows and not ESXi itself where it can be 1MB+? Could you also point me to this within the user guides.



            • 3. Re: P2V Cluster
              weinstein5 Guru
              User Moderators

              http://www.vmware.com/pdf/vi3_35/esx_3/r35u2/vi3_35_25_u2_mscs.pdf only references 2 node clusters -


              RDMs allow the VMs to access the native files system - so if the NTFS disk was created with larger block size only the VM will see it - but there is another problem - since you are using VI-3 the largest LUN that can be presented to your ESX/ESXi host is  2 TB - 512 B which means that iwll be the largest RDM that your VM will be able to access

              1 person found this helpful
              • 4. Re: P2V Cluster
                Al911 Lurker

                Thanks again.


                I will check the VCS but I don't think that limitation is there for me since it is not MSCS. As for the LUN sizes being 2TB max, I am aware and fine with it.


                My biggest LUN is 1.8TB within the cluster. As long as all block sizes within the cluster follow the NTFS format of 4KB I am fine. don't want to run into a situation where 4KB is mixed with 1MB+ block sizes.


                So in reality once RDM is created that VM from SAN/SCSI perspective is OS presented fully. VMFS does nothing except to provide address.


                Also I am assuming best is to use Physical Pass-thru for the VM RDM?

                • 5. Re: P2V Cluster
                  Al911 Lurker

                  I have looked around a lot but no clear answer for rdm block size. I can even use vi 4 ... Would that make any difference.