Cluster SQL 2014 with vsphere 6.0U2 3Host and 4 VM(Windows Server 2012 R2), Is it possible?

[omegape]

Good afternoon.

Is it possible to make this proposal? Cluster SQL 2014 with vsphere 6.0U2 3 Host and 4 VM´s (Windows Server 2012 R2).
I am currently testing with 2 disks (iSCSI) assigned by RDM in 1VM and shared for the rest of VM´s. It is 2 disks look good in the 4VM´s.

The problem is to validate the cluster, ERROR (Validate ISCSI-3 Persistent Reservation).
The error appears to have 2 virtual machines on the same Host. If I separate virtual machines on different host validation it is ok.

My problem is that I need 4 virtual machines and have 3 Host. 2 virtual machine must be on the same host and can not advance through the Error indicated.
Can you help me.

Thank you.

[akarydas2]

Here are some links about MSCS:

VMware KB: Microsoft Clustering on VMware vSphere: Guidelines for supported configurations

VMware KB: Microsoft Cluster Service (MSCS) support on ESXi/ESX

And based on the table below is not supported the pRDM when both nodes are on same Host:

Even if you trick it and configure your 4 nodes are you sure you want a configuration that is not supported by vendor?

Maybe try direct iSCSI through the node initiators to your storage and avoid the pRDM?

Best Regards

[omegape]

Hello akarydas2.

I understand that according to the table shown could not do what I want do.

My disk are RDM, according to the table, I can not have a cluster with multiple virtual machines on the same host. is it ok?

My final park want to build is:

Host 1 to 1 virtual machine

Host 2 to 1 virtual machine

Host 3 to 2 virtual machines

1 Cluster with these four virtual machines.

Thanks

[cesprov]

You can't, or at least shouldn't, do what you are trying to do with RDMs.  You'll need to configure the disks as VMDKs using bus sharing.

But question...you have SQL 2014, why are you wanting to set up a traditional old-style cluster?  You should really be looking at setting up an AlwaysOn Availability Group which requires no shared disks.  In my experience, AGs are a lot less flaky in VMs.  With an old-style cluster and bus sharing, you have to configure it exactly correct or you'll get txlog corruption.  I've also seen traditional clusters split-brain where both nodes thought they were active and as you would guess, that totally jacks up your DBs.  Yes that scenario is rare but possible if not configured correctly.

[akarydas2]

Since seems like your configuration is not supported, I would suggest to keep your VMs as they are and instead of using physical or virtual bus sharing to dedicate a storage for them and connect them using the traditional iSCSI through their iSCSI initiators.

I am not sure what exactly you will have on top of this vSphere cluster and if you will share it with other services but keep in mind that the pRDM are not controlled by the i/o storage control also. So if you are not sure about the storage utilization that the MS cluster will need, sound like the better approach to use a dedicated storage and have a supported configuration.

Best Regards