partitioning errors inside VMs

[Maxwelldoug]

Latest Installer as of July 5th 2024 results in frequent install failures. I have now verified this across multiple hypervisor platforms on multiple hosts including multiple host OSes.

these failures seem to run down to two errors with what is likely a shared cause.

Install on a blank disk starts as such

but quickly devolves to one of the following errors depending how the machine is feeling (or other times works completely fine with seemingly no difference.)

While I wanted to create an issue for this, I cannot give much more detail as the logs are rather unhelpful and I know very little of what the script is doing. If there are any further debugging steps I can take, I am glad to help.

[Maxwelldoug]

Forgot to mention, this is not a case of the hypervisor presenting an invalid disk as the same hypervisor with the same disk worked with no configuration changes or changes to the installer version at this time yesterday, and I am reinstalling due to an unrelated failure on my part.

[Maxwelldoug]

Update to this, I found a distinction: disabling VM tools fixes seems to fix this, since issue creation I have done 4 attempts with VM tools enabled and 4 without, switching regularly. All attempts with VM tools failed, all attempts without it succeeded.

[taukakao]

How are you enabling these tools?
Is it a setting on your host or are you installing something in the live image?

[Maxwelldoug]

How are you enabling these tools? Is it a setting on your host or are you installing something in the live image?

Immediately after the time zone Albius asks if I want VM tools and I click yes.

[taukakao]

Then I'm pretty sure it's just been by pure chance.
The VM Tools page just influences which image gets downloaded. Albius fails before it even reaches the download here.

[Maxwelldoug]

It'd be a pretty big statistical anomaly as I'm now up to 6/6 and pattern maintains, but you'd know more than I do about the codebase, and if the choice only affects the image selected then I'm out of ideas.

[taukakao]

6/6 sounds incredibly unlikely
But I just checked and I absolutely can't find anything else that changes with that.
Maybe I'm missing something tho.

[Maxwelldoug]

This persists in installer 130

[Akryum]

It happens to me too with or without selecting to install the VM tools, using VirtualBox

[Maxwelldoug]

That confirms this made it to stable. I've heard second hand a few accounts of it on real hardware, although much rarer. Possibly a good time to start investigating.

[PyroDevil]

I encountered this issue and reported it here: Vanilla-OS/vanilla-installer#414

So I suspect that it is a timing issue, the partition gets created, but the partition device node doesn't exist yet. A blockdev --rereadpt /dev/$disk after the partitions are create might resolve this, it causes the kernel to reread the partition table. Otherwise adding a small sleep after it wasn't found immediately and retry would also do it.

[paullenhardt]

I have this issue occurring on real hardware. I have two laptops which I have been testing on and both have the same "No device found for /dev/sda4" error occur. Both laptops are older 8-12 years old (not sure if it matters).

This seems to reproduce every time I attempt to install on either of the laptops (currently on the 8-24-2024 build, but it happened on older builds as well). Let me know if there is any testing or extra steps to carry out.