4.3 vm-pool hard shutdown susceptibility?

User Support General
1

I purchased a new laptop for use with Qubes 4.3 (having used Qubes as my daily driver for years now) however I’m still running on my old 4.2 install machine due to the new one regularly corrupting the vm-pool on hard shutdown (specifically power loss during suspend).

I wholly admit this is a “me problem” and I simply shouldn’t let it occur.

With that said, the laptop running out of power whilst suspended is a fairly common occurrence for me on a laptop OS. I’ve never lost the pool on 4.2 (or earlier) across various different laptops, including on other types of hard shutdowns such as freeze ups, and I’m interested as to what may have changed between 4.2 and 4.3 that seemingly makes corruption more likely.

I see various issues on Github to do with pool issues during upgrades, but this is on clean install(s) of 4.3.

I would suspect bad hardware on the new machine, or perhaps the fact it is from a different manufacturer, however it does appear consistently due to the lack of a proper shutdown with the machine being otherwise fine. I’m tempted now to install 4.2 on it to see if it behaves any better, though I believe we’re now coming up on the EOL for 4.2 so that won’t do much good apart from providing a comparison.

Interested what others experiences are, e.g. if you’re merrily on 4.3 with similar risky behavior and having no problems.

2

Do you use LVM or btrfs?

3

iirc the installer default was LVM

4

I do wonder if btrfs would serve you better. My experience with LVM has been that it’s very powerful but not greatly robust against power failures.

1 Like
5

Actually, I should have followed with experience: when using btrfs on r4.3, I’ve had over 20-30 hard shutdowns and I’ve only experienced corruption once, when the shutdown happened during a dom0 update.

6

I’ve been noticing other posts where folk have LVM problems like this.
One is here, and @marcos-morar comments with links to older ones, even going back to a case with forced power off in july 2025 (maybe 4.2).

Some ideas/questions:

  • suspend+hard shutdown is a frequent cause?
  • specific hardware? - like some disk or controller write cache that tells lies about when data has been fully written, so strict ordering is not possible…?
  • or only tired/worn disks?
  • unlikely for LVM to have a bug giving such problems? - it seems such a problem should also show in normal Fedora and other Linuxes that use LVM. I looked for such reports, but did not find any big number yet…
    • except if there is any special option or feature in Qubes use of LVM, or Xen underneath, that could cause it?

Maybe it is useful for people to tell what hardware they have, if they are comfortable - it could give a clue. Especially disk type and system/motherboard, but who knows what else might contribute?

And this ^^^ sounds useful.

What does everyone else think?

I am worrying about turning on suspend, because of it, and I need to decrease heat generation ! Btrfs is not tempting.

1 Like