Automating DRS rule creation for Metro Availability

  • 8 June 2016
  • 4 replies

Badge +5
We're in the process of deploying our new VDI environment, using View 6 & UniDesk ontop of our Nutanix MetroAvailability Clusters.
We will be balancing the number of production desktops running at both sites to spread the load around (we'll have some containers active in site A, some active in Site 😎, and want to ensure VM's do not get vmotioned to the 'standby' site.
Although we do have significant bandwidth (10Gbe via dark fibre) & low latency (~1-2ms) between sites, we'd like to avoid the additional overheads where possible.

Is there a way for us to easily create DRS rules within vSphere to ensure VM's on Active Containers at Site A should always run on those hosts & vice-versa for Site B? I vaguely remember coming across a PowerCLI script somewhere that did this (or something similar), but have been unable to find it again.

Ofcourse it would be amazing if this could be natively handled by Nutanix, but for now i'd be happy with a script we can run on a schedule to update the DRS rules.

Best answer by stephaneb 9 June 2016, 00:34

I wrote a script that does just that. It's available at the end of this blog post:

View original

4 replies

Userlevel 4
Badge +18

Nutanix has an NCC check to find VMs running against standby containers but we don't have anything formal to maintain DRS rules today, although this has been discussed. Regarding a script, maybe this blog post (from  ) can help:

Badge +2
I wrote a script that does just that. It's available at the end of this blog post:

Userlevel 7
Badge +30
Check out this post from Stephane on our Services team

This should be what you're looking for.

That all said, I'm curious, what's driving the requirement for Metro with a VDI setup?

Most shops that we interact with leverage the various in-built mechanisms in products like Citrix XenDesktop and Horizon View to do site surviveability (just as a general statement, Nutanix or otherwise), and only use Metro for Server Virtualization, where those apps may not have site availability solutions.

Also, as you likely know, even with a 10g, low latency link, in any vendor's sync-replication (again, Nutanix or some other solution), any sort of latency, at all, will be directly in the write path, which usually doesn't align well with VDI performance (server apps are a bit more forgiving on this front).

Anyways, food for thought.

Badge +5
Awesome, that post from Stephane is exactly the one I had come across previously. Thanks!

As for the driver behind Metro for VDI; its a direction from our CIO following previous instances of hardware failure taking out our production site. The goal is to move to a full Active/Active solution for all infrastructure; reducing RPO & RTO across the board and having provisions for seemlessly migrating all processing to either site without any downtime.