Code review comment for ~mwhudson/curtin:lp-1893818

> > > > But in groovy+, the udev rule from multipath-tools that has
> > > > always attempted to remove the devices nodes for the partitions
> > > > of a disk that is a multipath member actually succeeds, and
> > >
> > > Won't this break curtin on older multipaths that don't have this rule?
> >
> > No, if the device node for a partition of a disk that is a multipath member
> is still present, this branch ignores it (or at least that's the intent...)
> >
> > > Or will this work with older multipath output since we're only referring
> to
> > > the mpath dev (rather than a path of the mpath device)?
> >
> > Er, I think so. Not completely sure about the distinction you're making
> here.
>
> In the description, in groovy+ we will no longer have the partitions on the
> path members IIUC. For example (say a 2 path disk with 1 partition) we'd see:
>
> /dev/sda
> /dev/sda1
> /dev/sdb
> /dev/sdb1
> /dev/dm-0 (mpatha)
> /dev/dm-1 (mpatha-part1)
>
> And on Groovy+ we see:
>
> /dev/sda
> /dev/sdb
> /dev/dm-0 (mpatha)
> /dev/dm-1 (mpatha-part1)
>
> Is that correct?

Yes.

> My concern was if the single-path partitions are present (as they are on
> Focal and older) does the new code still handle things OK?

Oh I see, you're worried about the downstream impact of this: whether the storage configs produced by curtin after this change can actually be installed. That's a good question!

> I *think* it's yes due to the fact that curtin knows how to create partitions
> on MP devices since 20.04.

I think you're probably right but I should also do some testing of this situation.

> I've added some replies in-line, awkwardly, you have to select your previous
> commit to see those.

Replied.