Deploying Horizon Cloud Pod Architecture

[D81]

Hi everybody

I have a customer with two datacenters on two different sites. Both sites have the same hardware specifications in terms of hosts and vcenter.

Im currently deploying a Horizon 8 enviroment and App volumes in one site and the test are working fine.

The idea is to deploy the same Horizon and app volumes architecture on the second site tol work as DR. So one site wil be the main one and in case any resource of the site may go down the users will be able to switch to the secondary site until the services on the main site is restored.

In case of App volumes I have seen that you can create two sites and link them with an intermediary storage that allow the main site (App Vol instance 1) to upload the packages there and the secondary site (App Vol instance 2)  will copy those packages to its own datastores. So in case that the main App Vol instance 1 is down you can use the resources from the secondary site.

However in case of Horizon view I see that there is another kind of architecture that is called "Cloud Pod Architecture" which is a way to federate several sites to a single horizon console to manage the resources between sites.

https://docs.omnissa.com/es-ES/bundle/Horizon-Cloud-Pod-Architecture/page/CloudPodArchitectureinHorizon8.html

Im reading the documentation but I have some initial questions:

GOLD IMAGES

If an horizon administrator creates one golden image on site A, can he deploy pools on both sites with the same gold image or does he have to manualy move or copy the gold image from site A to datastore on site B in order to use it on site B?


NETWORK

Regarding the network configuration on both sites, this customer has different network range ips on each site. So as an example on site A the users have the 192.168.1.0/24 however in site B the users have 192.168.2.0/24 

So in case a user is conecting to the site A, he will have visibility to all the machines on that site, but what happens if he wants to connect to a machine on site B (cause site A is down)? is it as simple as give him conectivity throuw the firewall between sites or does he perfom any other changes?

LICENSING

The users will only use one site or the other, so if the customer has 200 concurrent users, they will be switched to one site at time. Can they use the same horizon licensing on both sites? Or they must buy a new pack of licenses for the second site?


DR SCENARIOS

Im not sure how DR will be managed on this architecture. Lets mention some scenarios:

The customer will use Instant cones pools for the main part of his users and also about 20 full clones for specific users. The idea is to have all of them deployed on site A, so site B will only act in case of a failure of the main service.

In case the Horizon view servers from site A are down, I assume that the users can switch to connect to the Horizon servers from site B and they will have access to the resources of the site A without any problem. Is that correct? In case it is correct is there any way to automaticaly balance this user connection or it may be manualy done by the user?

In case whole site A is down (lets say there is a flood and the building is off), the users will only have access to the resources on site B. Is there any way to automaticaly "clone" the golden images from site A to site B in order to launch the pools automaticaly and minimize the downtime for the users while site A is restored? Or should the horizon administrator keep a manual copy of the golden image on both sites?

Thanks in advance!

[Sean Massey-1]

So there is a lot to discuss here, and some of this goes beyond Horizon into your underlying or supporting infrastructure.

4 hours ago, D81 said:

GOLD IMAGES

If an horizon administrator creates one golden image on site A, can he deploy pools on both sites with the same gold image or does he have to manualy move or copy the gold image from site A to datastore on site B in order to use it on site B?

No.  There needs to be a process to either replicate the image to the second site or build a new image in the same site. Some customers do it manually by exporting the VM as an OVA and importing it.  Others use vSphere Content Libraries.  And I've seen other customers just automate their build process and build the image from scratch in each location.

4 hours ago, D81 said:

NETWORK

Regarding the network configuration on both sites, this customer has different network range ips on each site. So as an example on site A the users have the 192.168.1.0/24 however in site B the users have 192.168.2.0/24 

So in case a user is conecting to the site A, he will have visibility to all the machines on that site, but what happens if he wants to connect to a machine on site B (cause site A is down)? is it as simple as give him conectivity throuw the firewall between sites or does he perfom any other changes?

This depends on how the customer's networking between sites is configured.  You need to talk to your customer's networking team to understand what sort of connectivity they have between sites, how much bandwidth they have, and how they secure traffic between them. 

4 hours ago, D81 said:

LICENSING

The users will only use one site or the other, so if the customer has 200 concurrent users, they will be switched to one site at time. Can they use the same horizon licensing on both sites? Or they must buy a new pack of licenses for the second site?

They can use the same Horizon licensing in both sites.

4 hours ago, D81 said:

DR SCENARIOS

Im not sure how DR will be managed on this architecture.

So...this is the first problem. Before asking questions about potential DR scenarios here, you should be spending time to understand your customer's DR strategy, plans, and any requirements or expectations around desktops that they have. The DR plans and requirements will impact how you build and operate the environment.  It could also impact things like file server design.

4 hours ago, D81 said:

The customer will use Instant cones pools for the main part of his users and also about 20 full clones for specific users. The idea is to have all of them deployed on site A, so site B will only act in case of a failure of the main service.

In case the Horizon view servers from site A are down, I assume that the users can switch to connect to the Horizon servers from site B and they will have access to the resources of the site A without any problem. Is that correct? In case it is correct is there any way to automaticaly balance this user connection or it may be manualy done by the user?

What do mean by "access to resources of site A?" Is that desktops in Site A? Or applications/services/servers in site A?

4 hours ago, D81 said:

In case whole site A is down (lets say there is a flood and the building is off), the users will only have access to the resources on site B. Is there any way to automaticaly "clone" the golden images from site A to site B in order to launch the pools automaticaly and minimize the downtime for the users while site A is restored? Or should the horizon administrator keep a manual copy of the golden image on both sites?

Horizon has no automatic method of replicating the image between sites.  It relies on the infrastructure or admins to perform that operation. 

[D81]

Quote

What do mean by "access to resources of site A?" Is that desktops in Site A? Or applications/services/servers in site A?

Many thanks for the explanations!

When I mentioned the posibility of accessing the resources of site A I was meaning to the pool sessions (instant clone VDI).

So imagine that there is a faulty update of horizon servers from site A and they become unabailable during some time. Would the users be able to connect to the horizon servers from site B and access to the pools already deployed on site A?

 

[Robin Harmsen]

2 hours ago, D81 said:

Many thanks for the explanations!

When I mentioned the posibility of accessing the resources of site A I was meaning to the pool sessions (instant clone VDI).

So imagine that there is a faulty update of horizon servers from site A and they become unabailable during some time. Would the users be able to connect to the horizon servers from site B and access to the pools already deployed on site A?

 

They will not be able to access the sessions in site B when all horizon connection servers in site A are down...


I also read something about persistend desktops, those will also not replicate / be available at the other site.

[D81]

OK, thanks for the answer...

But at this point, if golden images must  be manualy copied between sites, if the pools of the site A are not gonna be available on site B, etc. 

Then is there any advantage of having a horizon cloud pod architecture in case of DR?