Skip to main content
Anyone running Traditional SQL Clusters on a Nutanix / HyperV implementation?
Hello



Assuming you mean a Failover Cluster configuration, then I believe this is (currently) not supported as this requires sharing the data accross all nodes.



I believe Nutanix does not support a CSV nor sharing a vhd(x) file.



We ended up using a SQL Server AlwaysOn configuration which is fully supported by Nutanix. Down side is that it requires an Enterprise license and more storage.



Hope this helps!
Hi Tom, You're correct. We don't yet support SCSI 3 Presistent Reservations, which are required for the shared disks to work properly with failover clustering. This is likely to change at some point in the future.

What part of SCSI3 persistant reservations are not supported? Is it only if shared VMDK's are used, or if, in a VMware example a SCSI adapter is set to physical bus sharing mode? Reason I am asking is I see a really good future use case with Storage Replica in the next version of Windows Server.

This still requires SCSI3 persistant reservations to talk to the underlying VMDK in Windows Failover Cluster manager, but the disks arent actually "shared" at the VMDK level as replication happens in-guest over SMB3 to the other cluster node. Reservations are only used to control locks to the destination replica at the OS level. I wrote about this in my blog here http://www.cloudypolitics.com/virtualize-windows-storage-replica-cluster-on-vsphere/

Hopefully with this feature we can create windows clusters on Nutanix using SQL Server Standard edition.


SCSI3 PR isn't implemented in the Nutanix SCSI stack as yet. So anything that requires SCSI3 PR, including Pass Through RDM's for Failover Clustering, will be prevented from working. However in the case of failover clustering there is a solution that will work on Nutanix using third party tools. If failover clustering for SQL Server and other applications is required then SIOS DataKeeper and other similar tools can be used to enable the traditional failover clusters.

These third party tools are supported by their respective vendors. I'll update this thread when there is more news on SCSI3 PR and also any validation / joint papers with SIOS and other vendors of similar technology.


vcdxnz001 wrote:Hi Tom, You're correct. We don't yet support SCSI 3 Presistent Reservations, which are required for the shared disks to work properly with failover clustering. This is likely to change at some point in the future.

Hi, you indicated that SCSI 3 Persistent Reservations are not supported, but is CSV supported? Beginning in SQL Server 2014, AlwaysOn Failover Cluster Instances supports Clustered Shared Volumes (CSV) in both Windows Server 2008 R2 and Windows Server 2012. // 



Thanks,



Ian
CSV aka Cluster Shared Volumes also requires SCSI 3 Persistent Reservations. This may work with NOS 4.1.5, which has just been released as SCSI 3 PR is now tech preview. But we haven't tested it.



Also SQL FCI aka MSCS aka Failover Cluster is now in tech preview with the use of in-guest iSCSI as of NOS 4.1.5.
Hi 



Has this been tested yet?Jason
Hi Jason,



We have tested it internally in our QA and performance lab environments across the world under various different conditions for quite some time now (across all phases of testing). Performance may not be as good at the moment as it will be when the feature is GA. Tech Preview does not mean not tested. But it should be used in non-production environment until we fully GA the feature. It will become GA in a future release.
Hi 



Is it going to be supported to synchronously replicate a pair of Windows Cluster VMs (VMWare) using Nutanix replication to another Nutanix cluster?



Is it even supported to have local Nutanix snapshotting of cluster pair of VMs?



Put another way, what options are there for protecting/DR of any cluster pair? Would prefer to keep away from SQL mirroring/log shipping.
HI Dazza,



I haven’t tested it, I’ll find out. But for SQL Server we would recommend FCI local site (clustering) with Availability Groups (replication) between sites. This will be application consistent, as anything we do at the storage level would only be crash consistent.
Thanks  would love to know.

So here lies the problem with using AAGs for DR form a business PoV…….. it’s up against current SAN technologies like HP 3Par Cluster Extensions. This gives you multi-site synchronous stretched cluster capabilities under a SQL Standard license model. That deals with HA and DR. So, when we look at our current SQL project the SAN looks very appealing because it could make the difference between our SQL project costing £150k in SQL licenses or £600k. If Nutanix requires an Enterprise SQL license to achieve the same HA/DR capabilities via AAG, then the reduced TCO of any move from SAN to Nutanix could be negated by a need to go Enterprise, assuming of course there's no other driver to go Enterprise.
The biggest problem with stretched cluster is the miantenance of the cluster itself. Unless you have two of them you can't guarantee non-disruptive operations for production and they are very complex to implement and maintain. So whatever you save up front you end up paying for over time in increased opex. Nutanix offers Metro availability that can be implemented with a few clicks and removes a lot of the problems with a traditional SAN based stretched cluster. However this doesn't work with SQL Clusters at this time, which require in-guest iSCSI access from Volume Groups.
Hi 



With the release of our v4.5 OS today we now support Windows Failover Cluster with in-guest iSCSI access with MPIO. This will help with your project and for other customers that need to support traditional failover clustering instead of or in addition to Always On Availability Groups.
That's good news. Now just waiting for the functionality to replicate that cluster 😃
hi,



so now nutanix support scsi right?.i am new to windows .can you provide some links about the traditional sql clutering?



rgds

ritchie
I know this is an old thread but it popped up in a search so I thought I would add. You can run failover SQL clusters in Nutanix with no issue. I've built them over the last few months and it's a solid, high performance, high available solution. Each drive is created as a Volume Group that both nodes can access. I have one Server 2012 running SQL 2016 Standard now and am planning another running Server 2016 and SQL Enterprise this year (2019).

Hi David,

SCSI3 PR isn't implemented in the Nutanix SCSI stack as yet. So anything that requires SCSI3 PR, including Pass Through RDM's for Failover Clustering, will be prevented from working. However in the case of failover clustering there is a solution that will work on Nutanix using third party tools. If failover clustering for SQL Server and other applications is required then SIOS DataKeeper and other similar tools can be used to enable the traditional failover clusters. These third party tools are supported by their respective vendors. I'll update this thread when there is more news on SCSI3 PR and also any validation / joint papers with SIOS and other vendors of similar technology.

Kind regards,

Michael

Hi @vcdxnz001 - Time for that update on the thread? 🙂 . Since SCSI-3 PR support was added with  AOS 5.17. How it affect WSFC configurations and related  on each mayor hypervisor (AHV, ESXi and HyperV)  running on Nutanix HCI. Thanks for sharing and commenting.

Here is a good reference: https://next.nutanix.com/archive-44/shared-storage-for-windows-server-failover-clusters-on-nutanix-27364


Hi Hugo, 

 

Now SCSI 3 PR is natively supported it works across hypervisors. For AHV if you connect vDisks to multiple VM’s and use WSFC it will natively work without any special configuration. Usually you would use a shared volume group between the VM’s for the cluster storage. This also works with RedHat and Linux clusters that use SCSI fencing.  This is a major win and makes setting up failover clusters very simple.

For ESXi and Hyper-V you may need to use iSCSI connections to Volumes on AOS. Although I think you can pass the SCSI 3 PR command via SMB3 in Hyper-V, so it may also work natively with normal shared vDisks. Recommend using LTS release of 5.20 or above if possible so you get the latest features. 

Here is a good article - http://longwhiteclouds.com/2020/11/09/simple-guest-os-clustering-without-complex-config/


Doing a WSFC configuration with WS2019 | ESXi 6.7 | AOS 5.20.0.1, all going well until we add a 5th disk. We are presenting using Nutanix Volumes. Then found this comment https://docs.microsoft.com/en-us/troubleshoot/windows-server/high-availability/scsi-3-persistent-reservation-test-fail. A sort of configured limit could be the reason?. Any advice? - thank you!

 


There is no configured limit. Are you using iSCSI connections to those volumes? I would suggest opening a support request for that. 


Yes, using iSCSI connections. Already working with Nutanix Support.