Unable to boot appvm due to qubes-relabel-rw

[Rune Philosof]

After updating from `fedora-38` to `fedora-39`, I have some AppVMs that I am unable to boot because `qubes-relabel-rw` takes more than 60s.

The VMs have large `/rw/bind-dirs/var/lib/docker`.

I can clean my docker state before making the template switch, but it would be nice to avoid having to recreate the state.

It would be good with a more user friendly process for this, both for explaining the problem and the solution.

[Rune Philosof]

Also, if this was a rarely used appvm, then I might not have noticed that the problem started due to a template vm upgrade.

I also might not have the old template vm present on my system to switch back to for cleaning up the bind dir.

So it would be nice with another solution for this.

The simplest to implement is probably to remove the time out, and have the user kill the vm if it takes too long.

But that might be problematic in the case of service vms that start at boot? If so, the time out could be conditional.

--
You received this message because you are subscribed to a topic in the Google Groups "qubes-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/qubes-users/HejWPZX102Q/unsubscribe.
To unsubscribe from this group and all its topics, send an email to qubes-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/qubes-users/8c63b924-605f-4bbe-8ead-7236abc7764en%40googlegroups.com.

--

Med venlig hilsen / Best regards

Rune Philosof
Software developer

+45 28 45 64 08
r...@abtion.com


Vesterbrogade 15, 3
1620 København V

Sverigesgade 18
5000 Odense C

https://abtion.com

[unman]

On Mon, Jun 17, 2024 at 12:02:25PM +0200, 'Rune Philosof' via qubes-users wrote:
> Also, if this was a rarely used appvm, then I might not have noticed that
> the problem started due to a template vm upgrade.
> I also might not have the old template vm present on my system to switch
> back to for cleaning up the bind dir.
>
> So it would be nice with another solution for this.
>
> The simplest to implement is probably to remove the time out, and have the
> user kill the vm if it takes too long.
> But that might be problematic in the case of service vms that start at
> boot? If so, the time out could be conditional.
>

> On Mon, Jun 17, 2024 at 11:23???AM 'Rune Philosof' via qubes-users <

> qubes...@googlegroups.com> wrote:
>
> > After updating from `fedora-38` to `fedora-39`, I have some AppVMs that I
> > am unable to boot because `qubes-relabel-rw` takes more than 60s.
> >
> > The VMs have large `/rw/bind-dirs/var/lib/docker`.
> >
> > I can clean my docker state before making the template switch, but it
> > would be nice to avoid having to recreate the state.
> >
> > It would be good with a more user friendly process for this, both for
> > explaining the problem and the solution.
> >

Rune
Thanks for suggesting this - I'll have a think, and see what might be
best suggestion for most users.
unman