Blog Archives

Converting an entire Citrix & VDI platform from ESXi to AHV... in 30 days.

4/12/2023

The goal: move an entire production Citrix & VDI platform from ESXi to AHV. Below is our before (left) and after (right)

What is delivered:
•EMR sessions daily: 10000+
•VDI sessions daily: 4800+
•delivers 900+ general apps
•13000+ Citrix users
What makes it up:
•Published Apps (1155 servers)
•Infrastructure (69 Server)
•primary data center hosts- 110
•secondary data center hosts- 72

The POC

First thing we needed to do was a pilot POC. To accomplish this we initially build a 4-node AHV test cluster running a variety of AHV & test workloads.
We had a few hundred users running a variety of non-prod apps, the Citrix team used this cluster to iron out wrinkles in the Citrix deployment process.
Worked through “power-on” actions in our images (the way VMWare and Nutanix handle VM power reset options differs, so the images had to be changed to honor the AHV power action)
Worked through tweaking Cisco ACI/EPG configs (because ACI has a plugin for VMWare, but not Nutanix, there was a minor tweak in our config needed for AHV)

Cluster sizing

Citrix: On ESXi, our default cluster size had been 24 nodes, but this led to LCM patching process being quite lengthy. We decided to modify this to 12 node clusters for this deployment, both to decrease the failure domain and for manageability for upgrades.
VDI: since we have 2 different flavors of VDI (server-based and client-based) we decided on 2 different 24 node VDI clusters for the solution.

Workload & support challenges

EMR golden images from VMWare would not work in AHV. Worked w/ Nutanix Support, but couldn't get them working. These had to be recreated from scratch, and that worked.
Our backup solution does not work with AHV, so we had to set up DR Sync to back up our images.
During migrations - we were seeing higher than expected exception rates in the EMR, which required working with the vendor & Nutanix engineering -- eventually solved by forcing memory locality for the VM's on each host w/ vNUMA commands (not available in the GUI) 1-shot fix, once it's set on the host, it's done. This is slated to be fixed in a future release.
iTLB multihit – we disabled this flag– due to performance impacts of the mitigation code (similar to VMWare’s handling of the issue).

The Citrix migration

Drain old ESXi cluster
Convert 3 hosts to initially seed the new AHV cluster
Foundationed the remainder of the hosts and added them to the AHV clusters
Redeploy Citrix servers (power off on ESXi, redeploy on AHV via Citrix)

Entire process was done in less than a month

The VDI migration

Created initial 3-node AHV seed cluster
conversion process for personal/persistent VDI’s - used Nutanix Move to migrate those VDI's from ESXi->AHV. Moved 200+ VDI's that way with minimal errors.
Also used Move to migrate VM templates to AHV.

The migration was fairly painless and we actually saw improved density on the clusters running AHV instead of ESXi on top of AOS -- about a 10% improvement. The transition for our Citrix support staff was seamless.

The success of this project paved the way for our next project; converting our general production workloads from ESXi to AHV.

0 Comments

when vCenter boots into emergency mode

4/11/2023

0 Comments

"Entering emergency mode. Exit the shell to continue."

There's already a VMWare KB article about this: "Failed to start file system check on /dev/disk..." error on Photon OS based virtual appliances (2149838) but in our case this did not resolve the issue.

What we did to fix the issue:
Restarted vCenter appliance and entered GRUB menu.

Added "fsck.repair=yes" after "consoleblank=0" and booted.

All vCenter services started and were working as expected.

0 Comments

Converting an entire Citrix & VDI platform from ESXi to AHV... in 30 days.

when vCenter boots into emergency mode

Chad Dorr

Archives

Categories