iSCSI shared volume 2019 hypervizors

[tjarb]

Dear readers,

We're having a TS-435XeU storage for three hyper-v systems (10GBe).
The TS-435XeU has an iSCSI volume which holds the Virtual machines and the iSCSI volume is made available in all 3 hosts (MS iSCSI initiator). Every VM will be be running ONCE by one of the three 2019 server standard hypervisors.

I see that the VMs get corrupted, in a very short term. The hypervisors aren't in a cluster configuration yet.

I know it's a difference with the SMB shares, but whats the reason of this failure? Is it the partition administration (three writers) that makes the VMs corrupt?

Howto prevent this corruption? Should I use a different kind of share (SMB3)? Or should I just build up the cluster and tie the iSCSI volume up in the cluster-configuration (storage pool)?

Best regards,

[FSC830]

Access from 3 hyper-V hosts to same LUN without a cluster!??
Afaik in ESXi this works, but in Windows? Never seen such a configuration without cluster.
I guess thats the reason VMs get corrupted, but if you create a cluster the LUN can be accessed by only one host at a time.
So you may need to create 3 LUNs in total (one for each VM) and map all Luns to all hosts.

Regards

[tjarb]

Hi FSC830,

When configuirating the LUNs there's a warning that the cluster can be shared between multiple hosts in a cluster aware environment.
Is one cluster member then operating as proxy?

If I would make 3 LUNs, and there's a failover / move of the virtual machine, how does that work? Does the cluster do a sync from the active host to the other available hosts?

Best regards

[FSC830]

Its some time ago when I was playing with Hyper-V, I am using ESXi.
But there are some good setup guides at Microsoft.com, as far as I remember you need to create 3 LUNs at storage and map them to a Hyper-V host Next you need to setup the Hyper-V cluster and install VMs in cluster I cant remember ist the LUN needs to be shared as "cluster shared volume", ut setup guide should provide this information.
As long as all 3 HV-hosts are running each host can have one VM, multiple or all VMs. If all VMs running at one host, the others are idle. In case of HV-host crashes the VMs and cluster ressoureces will be taken over.
The information about the cluster is shared among all HV-hosts.

Why 3 LUNs? M$ cluster can share storage at a "per-disk" basis, so partitioning a disk is not useful for dividing storage space. If all 3 VMs at same disk, you always forced to switch all 3 VMs at same time to a host.
If you want use load balancing (each HVhost serves a VM) you need to divide the VM so that each VM has its own disk ressource.
In that case you can move each VM to each HV-host.

For an example see i.e. https://www.altaro.com/hyper-v/failover ... r-manager/

Regards