Long time since the last update, but better late than never…
When VMware released version 6.0 of vSphere, they announced support for VMs running Microsoft Clusters and SCSI Bus Sharing enabled RDMs.
This was a big feature for us, running several MSCS Clusters in our environment.
Pre vSphere 6.0, you needed to shut down all VM’s engaged in SCSI Bus sharing to be able to set the ESXi host in maintenance mode, which was a real downside for these VMs.
So we reinstalled our ESXi servers to ESXi 6.0 Update2, started one test cluster and ran a vMotion. The vMotion tasks itself worked as expected but the clustering service inside the VM’s failed instantly when the vMotion was done, resulting in corrupt data on the data disks.
We tried this several times, and got the same result.
After some testing it seems like you need to set the multipathing policy to FIXED for all RDM disks to be able to get this working, even though VMware states stat Round RobinÂ should be supported. We have tried with both EMC VPLEX, EMC VMAX and HP 3PAR disk, and we are getting the same result no matter what storage you are using.
We have created a SR with VMware, and engineering have confirmed that Round Robin does not work(even though it should).
Yesterday, VMware delivered a sandbox build of ESXi 6.0, build 7153901, that should fix this issue.
We tried it out, and the vMotion actually works with Round Robin on this release.
Now we just need to wait for a hotfix that we can apply in our production environment.
To be continued…
Just got this info from VMware:
Resolved in 6.0 P3 – due for release on July 28th 2016