many systemd units failing in oracular LXD containers

[Nick Rosbrook]

On Fri, Jul 26, 2024 at 12:33 PM Nick Rosbrook
<nick.rosbrook at canonical.com> wrote:
>
> On Fri, Jul 26, 2024 at 12:20 PM Robie Basak <robie.basak at ubuntu.com> wrote:
> >
> > On Fri, Jul 26, 2024 at 12:11:05PM -0400, Nick Rosbrook wrote:
> > > In short, this is not systemd's bug.
> >
> > I don't think that matters. The idea of the autopkgtest infrastructure
> > and "always being green" is that we hold back packaging updates if it
> > would regress behaviour, even if it's the "fault" of a different
> > package.
> >
> > It follows that we should revert if an update slips past our CI
> > infrastructure such that the behaviour regresses.
> >
> > Otherwise, new unfortunate interactions between packages cause everybody
> > else's development to grind to a halt.
> >
> > There is a trade-off here of course, in terms of minimising cost. It may
> > be appropriate to "push ahead" after a regression occurs, or indeed
> > deliberately bypass CI to land regression behaviour, if that is on
> > balance going to maximise progress from that point.
> >
> > But if such a trade-off to be made, I think it needs justification. Our
> > default position should be to minimise regression and "always be green".
> >
>
> That's all fair.
>
> In this particular case, the LXD team is already working hard on
> fixing it there, so I think reverting systemd at this point would be
> more trouble than it's worth. I will sync with them, and if it seems
> like it will be a while before their fixes land for users, I will
> upload a workaround to override the problematic unit settings for all
> services.

I have just checked with the LXD team, and this has already been fixed
in latest/edge for LXD. I confirmed the fix myself just now.

-Nick