Raw device mapping (RDM)

<yambe:breadcrumb>VMWare_vSphere_or_ESXi|VMWare vSphere or ESXi</yambe:breadcrumb>

Raw device mapping (RDM)

Ideally we can map LUNs to ESXi hosts and create VMFS filesystem. The immediate advantage is that same LUN can be accessed in parallel for read/write from multiple ESXi hosts. On top of this VMFS filesystem we can create .vmdk files or virtual hard-disks which can be used for persistent storage in VMs. However, if it is required to map a hardware device directly with a VM to expose low-level device APIs then we can use Raw Device Mapping (RDM).

Note that RDM is of two types:

• Virtual compatibility mode: In this case RDM acts like a virtual-disk file. Most of the low-level access are masked and managed by ESXi. This mode supports snapshots.
• Physical compatibility mode: Most of the commands are pass-through to the actual device as it is.

There are special considerations while doing vMotion or while expanding RDM devices. Use references below for more information

Refer:

About RDM, advantages and limitations

https://docs.vmware.com/en/VMware-vSphere/6.5/com.vmware.vsphere.storage.doc/GUID-9E206B41-4B2D-48F0-85A3-B8715D78E846.html

Migration VMs with RDM

https://kb.vmware.com/s/article/1005241

Expanding size of RDM

https://kb.vmware.com/s/article/1007021

ESXi installation or boot gets stuck at vmw_satp_alua

It is possible that during installation or boot ESXi gets stuck after displaying

vmw_satp_alua loaded successfully

This can happen if:

• We are trying to install ESXi on a node which was earlier part of Hyper-V cluster and has disks mapped to the host via storage (FC/iSCSI) with persistent reservation enabled
• We have mapped RDM to a ESXi host so that it can be further mapped with a windows VM.

In either case the solution is either to ask ESXi that the disk is perennially reserved and hence it should not scan it via:

• https://byteben.com/bb/esxi-host-slow-boot-stuck-vmw_satp_alua/
• https://kb.vmware.com/s/article/1016106

OR

Unmap the storage from Hyper-V cluster (Remove reservation) and then unmap the LUN from host and then proceed further.

• https://docs.microsoft.com/en-us/powershell/module/failoverclusters/clear-clusterdiskreservation?view=windowsserver2019-ps

If the hosts have been formatted or hyper-V cluster is down then deletion / unmap of such LUNs fails due to persistent reservation. In case of Dell MD3800F found that:

• There is SMCLI command to remove such reservations
  • https://www.dell.com/community/PowerVault/Need-SMcli-command-to-remove-persistent-reservation-MD3000i/td-p/4381316
  • SMCLI.exe is available at "C:\Program Files\Dell\MD Storage Software\MD Storage Mager\Client" and can be executed via administrator cmd.
  • Ran:

    smcli <md3800-raid-controller-ip>
    clear allVirtualDisks reservations;
    show allVirtualDisks reservations;

    Not sure if this helped or not

  • Other SMLI commands are available at https://downloads.dell.com/manuals/all-products/esuprt_ser_stor_net/esuprt_powervault/powervault-md3000i_reference%20guide2_en-us.pdf
• Finally had to clear entire storage configuration (There was no useful data on entire storage) via Dell storage manager for MD3800F using "Right click -> Configuration -> Clear -> Virtual disk"
• This LED to some other storage warnings / alerts but removed all LUNs, raid groups and reservations, etc. to allow installation of ESXi on hosts that were earlier part of Hyper-V cluster.

<yambe:breadcrumb>VMWare_vSphere_or_ESXi|VMWare vSphere or ESXi</yambe:breadcrumb>