The VMware Cloud Foundation (VCF) Holodeck Toolkit has become quite popular amongst our field teams for quickly spinning up a fully self-contained VCF environment, running on just a single physical ESXi host for educational and hands on experience.
Internally, we have a Holodeck Google Space which has been getting quite busy where folks can ask questions but also share their own experiences and setup using Holodeck. I had noticed there were some common questions and themes that were coming up along with some neat tips and tricks from various folks, some of which I also found quite insightful.
I know some of these tips and tricks will also benefit our customers and partners looking to deploy VCF using Holodeck, so I figure I put together a consolidated list that would benefit everyone. If you have other Holodeck tips and tricks, feel free to leave a comment and happy to keep updating this page.
Disclaimer: Modifications to the default Holodeck deployment and configuration is not officially supported by the Holodeck team, please use at your own risk.
Deploying Holodeck with old or unsupported CPU
If you are interested in deploying the latest VCF 5.1 release, which officially adds support for vSphere 8.0 Update 2, then you definitely will want to review some of the hardware considerations with the latest vSphere 8.x release, which I have shared HERE and HERE. Since Holodeck deploys Nested ESXi using ESXi Kickstart, if you are using an older or unsupported CPU, you could run into some deployment issues.
Luckily, there is a workaround which I had blogged about HERE, where you can add several kernel options to override the errors within the ESXi kickstart. Thanks to Tim Sommer for sharing this Holodeck tweak, you can manually update the VLCGui.ps1 script and append the following to the kscfg variable as shown below:
$kscfg+="install --firstdisk --novmfsondisk --ignoreprereqwarnings --ignoreprereqerrors --forceunsupportedinstall`n"
Unable to power on NSX Edge Node for old or unsupported CPU
Even with the workaround above, you could still face more challenges with using an old or unsupported CPU. As mentioned in VMware KB 87765, NSX 3.x now requires support for 1GB Hugepages and this may prevent your NSX Edge Nodes from powering on and failing the deployment.
Fortunately, there is another workaround, as mentioned in the VMware KB where you can apply the following VM Advanced Setting to the NSX Edge Node VM, which should then allow it to be powered on.
featMask.vm.cpuid.PDPE1GB = Val:1
NSX Edge Node running on AMD Ryzen Threadripper
While the latest NSX 3.x release supports AMD EPYC, if you are looking to deploy VCF using AMD consumer CPUs like Ryzen, you will still run into some problems. Luckily, the workaround that I had shared in this blog post HERE can be still be used to workaround this problem and glad to hear it also works for the latest AMD Ryzen Threadripper CPUs.
Holdeck with Multi-Physical ESXi Host
Holodeck has been designed and tested to work on only a single physical ESXi host, however you may not have a single host with the minimum resource requirements but you might be able to meet that using another physical host. It certainly would be nice if Holodeck could provide that additional flexibility, but no worries, James Kilby recently shared on his blog here on how he was able to setup a multi-host configuration using Holodeck with just a handful of tweaks! Nice work James, I may also need to try this out at some point ...
"Unable to power on NSX Edge Node for old or unsupported CPU" fix above did not not work for me. you don't have time to change the edge node when deployed from Holodeck, and even so, it did not work. what DID work was to set the settings in the virtual ESXi hosts running the virtual edge. I added featMask.vm.cpuid.pdpe1gb = Val:1
monitor_control.enable_fullcpuid = "TRUE"
That worked 🙂
Have you seen an issue where RDP suddenly stops. I have tried two different computers, rebooting the console server, rebooting the Holo router, and redeploying the Holo routers. I can get to the Internet and ping the gateway so that part seems to be working.