MSCS in Nutanix

  • 2 May 2015
  • 4 replies

Userlevel 1
Badge +8
Hello -

Is MSCS (clustering VMs across physical hosts) supported on Nutanix? What ESXi/NOS versions support MSCS?

How would I then present the shared disks (quorum, etc...) to the VMs?


4 replies

Userlevel 2
Badge +10
Failover clustering is not supported on NOS with ESXi due to the lack of support of RDM's and SCSI 3 persistent reservations.This is targeted to be addressed in a not so far future release.
Theoretically it should be possible with shared VHD on SMB3 with Hyper-V, but no testing has been done on that (AFAIK).
SQL Server Always On and other non-shared disk clustering techniques are the only way of doing clustering on NOS with ESXi.At a stretch you could create a Windows VM as an iSCSI target and use that to form the cluster. Not an elegant solution however.
Badge +2
The lack of MSCS support is a biggy for us now, and we are a prospective customer deciding whether to adopt Nutanix. Do you have ANY more informaiton about when this will be supported? Even if it's guidance? The problem with AAGs (the alternative) is the overhead on storage acapacity of having multiple copies of the same DB.

Badge +4
It's been in tech preview since June, so you can test it now.
Userlevel 4
Badge +19
Hi Dazza,

As others have pointed to a blog article I'll let you read that. But I wanted to let you know that there are other options that will allow a traditional failover cluster on Nutanix. Such as SIOS Technologies Datakeeper. We have not tested or validated this option ourselves, but it does allow a stroage independent failover cluster to be created, even across sites.

As for the problem around multiple copies of the data. We've found that compression is very effective with SQL Server workloads, even if they natively use page compression, and savings can be 30% or more, depending on the environment. So this somewhat reduces the overheads.

Failover clustering has its limitations, including the options that are available for availability. In future versions we will be reducing the capcity overheads even further, so customers get more usable storage space. Having that duplicated data can be a benefit when it comes to failure domain isolation and data protection. But there are always trade offs.

Anyway, ultimately we want to allow our customers to decide the best protection approach for their workloads and so we will be releasing an update to NOS that allows for in guest iSCSI access with support for SCSI 3 Persistent Reservations to allow for failover clusters to be created.

Kind regards,