- Make sure to have the RDM in physical mode
- For VMDK's the disk must be "Thick provision Eager Zeroed" and "Independent - Persistent"
Tip: for the sharing SCSI node I use SCSI1:X so I know that everything from SCSI controller 1 are sharing disks.
- Change the .ctkEnabled setting from the corresponding SCSI controller to false (this will disable change block tracking; so all changes are directly written to the disk. This prevents that the CBT file is locked and the other VM can't access that.)
Add a new row with the option SCSI1:0 (or whatever SCSI ID your shared RDM will be) and give it the value "multi-writer" so it is possible for multiple VM's to connect to the disk. It removes the write protection
VMware KB: Disabling simultaneous write protection provided by VMFS using the multi-writer flag
If you don't do this, the first powered on VM protects the disk, and the second VM will not properly start and give errors about file locks etc.
- Make sure you set these settings on BOTH VM's because these are VM settings. Also make sure you document it properly and test it before using it in production. Try things like extending disks etc.
You also could use powerCLI to set the settings above for example
$vm = "LINTST"
new-advancedsetting -entity $vm -name "scsi1:$X.sharing" -value "multi-writer" -confirm:$false
new-advancedsetting -entity $vm -name "scsi1:$X.ctkEnabled" -value "false" -confirm:$false
Or set the setting for the whole SCSI1: node with a simple loop
$vm = "LINTST"
foreach($x in (0..15)) {
new-advancedsetting -entity $vm -name "scsi1:$X.sharing" -value "multi-writer" -confirm:$false
new-advancedsetting -entity $vm -name "scsi1:$X.ctkEnabled" -value "false" -confirm:$false
}
- DRS rules: Don't forget to make a DRS rule to seperate both VM's when using it in the same cluster. (In my case I have a cluster with hosts across 2 datacenters)
- Linux fencing : I created a separate fencing user, the only privileges the Fence user need are "Interaction - PowerOff / PowerOn"
Hope this will get you going a bit