Bug#727708: systemd status when using multiple block device layers (MD/LVM/dm-crypt) below the root-fs
Christoph Anton Mitterer <calestyo@scientia.net> writes:
> Right now there is a rather fixed order in which things work (i.e.
> phys->MD->LVM->dm-crypt->rootfs) out of the box (well in some cases at
> least) and IIRC, due to some "obscure" code in the cryptsetup initramfs
> scripts it works also with (dm-crypt->LVM)... but ideally all this
> should be handled more generally...
> Looking over the bugs from the systemd package in Debian give quite some
> concern here,... many things seem to not work yet (e.g. support for
> cryptsetup key scripts)... and many other bugs seem to be quite old and
> not having been worked on for very long.
For whatever it's worth, I've been using systemd on a system with LVM and
dm-crypt (with LUKS) for about a month now, in the dm-crypt -> LVM ->
filesystem mode, and haven't had any trouble.
The page you linked to makes several assertions that I find curious. That
doesn't mean that they're wrong, but they seem somewhat contrary to my
experience. For example, I'm not sure where "however, for we really
should should try to avoid forcing users to use initramfs images, for
setups where they're not strictly needed" comes from; I've been using
initramfs images for every Debian system I run, and every Debian system
Stanford runs, for so long that I can't remember when we originally
changed. I don't understand why this would be something one would want to
avoid.
Similarly, I'm not sure why the focus on only adding necessary tools to
the initramfs image. Surely this doesn't matter much if the tools are
harmless when unneeded?
Hm. I'm also not sure that this whole digression really belongs on this
thread; maybe we should move it to debian-devel.
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: