Instant Clones High Availability on a Streched vSan Cluster

[FredericLOUKA]

Hi,

We have a streched Cluster on vSAN 6.7 U3 with a Horizon Infrastructure 7.12.

We use Instant Clones.

The streched cluster is streched through 2 rooms.

The fact is that I'm trying to test the disaster recovery plan, shuting down all the ESXi of my first room.

I saw different results and I'm not sure about the rules I have to follow with Instant Clones and HA.

In case of disaster recovery, do we have to enable HA on the cluster ? In my opinion, instant clones shoud be recreated by the horizon infrastructure on the second room, shoudn't it ? So we should not need to enable HA on the cluster ?

We enabled HA on our cluster and we saw that the VDI rebooted on the others ESXi but we observed strange behaviours, like some VMs in error state in Horizon Console....

What is the best way to have a disaster recovery plan in my case with my instant clones ? What should be the "normal" behaviour ? Recreation ? Reboot from HA ?

Thanks for your help

Fred

[nburton935]

Fred,

- Any reason you don't have separated vSAN-enabled clusters, with each pool behind a single Global Entitlement?

- Furthermore, you could have completely separate pods and join them to a CPA, which also helps from a DR perspective if the Connection Servers in a single pod were to fail.

vSAN stretched cluster is primarily for persistent desktops from a DR perspective. With instant clones, stretched vSAN adds an unnecessary layer of complexity. You're better off to go with 2 vSAN-enabled clusters with 2 pools behind a single Global Entitlement.

-Nick

[FredericLOUKA]

Hello Nick,

Does this mean that I need to have 2 copy of the master for one pool ? (one copy on each cluster ?)

And what about the AppVolume ? If I have 2 clusters, I need to find a way of replicate them ?

So it will consume twice the storage space instead of the streched cluster's configuration ?

Thanks again for your help Nick,

Fred

[sjesse]

Take a look at this

VMware Workspace ONE and VMware Horizon Reference Architecture | VMware

In explains your questions and more. Its not something you should or have to follow 100%(nothing wrong with doing that either).

[nburton935]

Fred,

- You're likely already doubling your storage consumption due to the stretched cluster anyway.

- You can simply do all of your changes on a single golden image. If both clusters remain within same vCenter, Horizon can automatically use the same VM from your source cluster for your 2nd pool - you'll see it on your list when creating / updating pools after selecting the vCenter. If clusters are in separate vCenters, you can either use a Content Library or script the copy to the 2nd cluster. There's a good fling out there by William Lam for cross-vCenter copying.

- App Volumes has Storage Groups when dealing with multiple clusters / datastores. It allows you to keep multiple datastores in sync with each other for your AppStacks. See here: Configure Storage Groups

I noticed that sjesse recently shared the Reference Architecture - this is the official guide that goes into multi-site scenarios.

Hope this helps,

-Nick

[FredericLOUKA]

Thanks to both of you for your answers. I will give a look at the documentation but it seems that having 2 clusters with one vCenter seems to be the best solution for using disaster recovery with instant clones on vSAN.

Fred