I have a customer who wants to do network maintenance in one of the data centres let's call DC1. How can we ensure that all new users are redirected to DC2 and that users with an existing session in DC1 can reconnect to DC1?

Goal:
Drain 1 pod in the data centre without user impact.

Customer environment:

• 2 datacenters
• 2 pods ( 1 pod per dc)
• 6 connections servers per pod (2 CS are dedicated for SAML authentication with WS1 Access)
• 4 UAGs only for outside connections

You might just want to disable provisoning for the pools in DC1.

Assuming they are using instant clone desktops, that way no new VM's will be created.
And when users logout the 'old' VMs will get destroyed. Eventually the POD/pools will be empty.

When the pod is drained ou might also want to disable the connection servers and UAG's and perhaps change loadbalancing for that DC, in orde to make sure it is not handeling any traffic before doing network maintenance.

Edited October 18, 2024Oct 18 by Robin Harmsen

2 minutes ago, Robin Harmsen said:

You might just want to disable provisoning for the pools in DC1.

Assuming they are using instant clone desktops, that way no new VM's will be created.
And when users logout the 'old' VMs will get destroyed. Eventually the POD/pools will be empty.

When the pod is drained ou might also want to disable the connection servers and UAG's and perhaps change loadbalancing for that DC, in orde to make sure it is not handeling any traffic before doing network maintenance.

That might take too long for the customer's maintenance window or cause other issues depending on how home sites are configured.  And it wouldn't prevent Horizon from sending users to that site as long as spare desktops were available.

It's been a while since I've done a CPA upgrade where we had to prevent users from logging into a site.  I think we had to disable the pools themselves in the pod that we wanted to perform maintenance on so that users would not get assigned desktops in that site, even if spare desktops were available.  We had to do other things too like disable load balancing for the UAGs in that site.

When you disable the entire pool, existing sessions cannot reconnect. To speed things up you can manually delete unused machines.. so you only have to deal with active sessions.

4 minutes ago, Robin Harmsen said:

When you disable the entire pool, existing sessions cannot reconnect. To speed things up you can manually delete unused machines.. so you only have to deal with active sessions.

Ah, yeah.  I missed that requirement.  I think your approach would work better...

If RDSH farms, you have to disable the RDSHs themselves to drain properly.  its a PITA w\o scripting:   FARM>> RDS hosts>>Selctall >>Disable  repeat

Sorry for the late reply.

The environment only consists of instant clone VDIs.

• Disabling the pool is not an option because reconnecting to a disconnected VDI is not possible as already mentioned.
• Disabling pool provisioning is the best option because you will always have a minimum number of VDIs (2) left in each pool. We currently have 7 pools, so 7 VDI will always be available.

So, ideally, I would like to put on POD in maintenance on the highest level and disable the pool for new connections, but reconnection to a disconnected session should be possible.

If you disable the provisioning in the pools no new VDI's machines will be created anymore, so no new VDI's will become avaiable...