Dear Maintainer,

   * What led up to the situation?
	Upgrading a system from Wheezy to Jessie. The system uses a split /usr on LVM and uses LABEL in fstab to mount this.
   * What exactly did you do (or not do) that was effective (or
     ineffective)?
	fstab looks like this:
	...
	LABEL=usr		/usr	xfs	...
	...

	blkid's output looks like this:
	...
	/dev/mapper/vg00-usr: LABEL="usr" ...
	...
   * What was the outcome of this action?
	Initramfs loops in "Begin: Running /scripts/local-block ... done." until it gives up and drops to the shell (or panics when panic=xx is given).
   * What outcome did you expect instead?
	Activate volume group, mount /usr, continue to boot.

Possible workarounds beforehand:
- Changing fstab from "LABEL=usr /usr ..." to "/dev/mapper/vg00-usr /usr ..." *before* rebooting works around this bug.
(Note that doing the reverse change - at any point after the installation - would also trigger this bug. Basically, UUID or LABEL must not be used for /usr when using Jessie with systemd)

Possible workarounds afterwards:
- Running "vgchange -ay; exit" in the initramfs recovery shell allows the boot to continue
- Using init=/lib/sysvinit/init also works (because then initramfs will not attempts to mount /usr)

Without those workarounds, which all require console access, the system will be unable to boot after the upgrade.

I originally added this to Bug #776409, but it looks like a different bug, after all.

Regards,
Marc

Control: reassign -1 lvm2
Control: forcemerge 612402 -1

No it doesn't.

Yes it is different.

Use of UUIDs to identify logical volumes is not recommended, for the
reasons explained in <https://bugs.debian.org/612402#85>.  The same
reasoning applies to labels.

Ben.