Can’t get past the UEFI shell.
Yes, it’s set to Linux and using VirtIO.
I’ve retried many times with different ZVOLs and have used both ones created by using the VM creation GUI and ones I create both before and after with the exact same result. Tried 32GB exact, larger, smaller, no change.
I’ve used different versions of HAOS qcow2 files as well (up to the latest 10rc2 files).
Always boots into the UEFI shell and never shows any file systems when selecting boot from file either.
Tried forcing the file conversion using -f qcow2 -O raw with no luck either.
sudo doesn’t help.
Tried it on multiple vdevs and in different locations as well with no luck. Other VMs work just fine.
Various configurations of other VM settings make no difference.
TrueNAS Scale - Bluefin [release]
Primary VDEV for the ZVOL is a pair of mirrored 2TB NVME M.2 drives.
Been using HAOS for years, hypervisors, including KVM, and other stuff for years, and I’m stumped.
I’ve tried converting from other image types even and I get the same result.
That’s one of the first things I tried, having had the experience in the past before myself as well.
There aren’t even and files listed. It’s like it doesn’t even see the drives.
I’ve continued to work on this since posting and have also tried direct dd image copy after converting the qcow2 file, also without luck and with no files visible.
Right, so I gave up and went the supervised route.
Even Debian gave me trouble booting after an install and I had to edit Grub but I got it working (Using the link troy shared works great for anyone else who stumbles across this).
Super weird given that Debian itself has also never had issues in the past - I must have borked something somewhere else and just cant find it.
Other OSs work fine - Windows, Ubuntu, etc.
Anyway - Got my Coral M.2 via PCIe riser working via passthrough with Frigate just fine now.
At least now I’m officially a community member instead of a lurker
Hey there, I had the shell prompt. What I did wront was use the wrong location for the ZVOL. I used the mnt path. I changed the qemu-img convert to use the /dev/{{pool-name}}/{{otherdir if app}}/{{yourZVOLName}}
Many thanks to @troy for this guide. It has worked flawlessly.
As TrueNAS Scale is evolving, there is a new option, WM min memory. I chose 1GB, with 2GB as the max / nominal value. In that way, the WM is only using 1GB at the moment, with half of that taken by homeassistant. Docs say that RAM amount will be increased up to the max if neeced. For me, a double win, as my modest NUC only has 8GB.
I have a MQTT broker as an App. To be able to communicate the VM with apps I had to create a bridge as explained above.
I am still sweating to have bluetooth working. There is an integrated Intel BT device, which works in TrueNAS after installing extra firmware and even scans nearby devices, but fails completely in HAOS:
It’s not a typical install… Here we are just downloading an image and writing it too disc. The disc in our case is the zvol. I guess these steps would be considered the installation. After that it’s just a matter of adding the zvol to a VM and booting up.
I wonder if your first download was either corrupt or did not complete successfully.
In your screenshot I see subsequent downloads are appending a number suffix and unxz is complaining about that.
Looking at the last three lines in the screenshot you shared the suffix is .3
Unxz ignores the suffix so I think it’s trying to unzip the first download but it says unexpected end of input.
Please try removing all downloaded copies of the hass ova file and try again. If it fails again, please share the full output of download and attempt to unzip again.
