LVM: How to access a foreign volume group

[Xen]

Volker Wysk schreef op 04-12-2017 2:03:

> I'm about to set up an LVM cache (for my encrypted root file system).

I suggest the following in /etc/initramfs-tools/hooks/dmcache:

-----------
#!/bin/sh

if [ "$1" = "prereqs" ]; then
     exit 0
fi

. /usr/share/initramfs-tools/hook-functions

# pdata_tools is the cache_check executable
copy_exec /usr/sbin/pdata_tools

# you have to use relative symlinks here:
for f in /usr/sbin/cache*; do ln -sr ${DESTDIR}/usr/sbin/pdata_tools 
${DESTDIR}$f; done

# safest is just to copy all of the device mapper modules:
copy_modules_dir kernel/drivers/md
-----------

But I assume you have already done this?

> It would be bad, if I had to restore it from a backup, because I have
> installed and configured a lot of things.

The greatest risk is that the logical volume for root is not activated 
because of a missing module or executable in your initrd.

> How to instruct LVM to operate on another installation? When I start
> the rescue system, which I've installed on an USB stick, I can access
> only the native LVM installation, like it should be.

That's because your main system is encrypted.

Otherwise LVM would start destroying your system (maybe).

> Also, both volume groups have the same name.

Yes you won't be able to activate both volume groups very well until you 
rename the one in the rescue system.

This isn't very hard.

BEFORE you do any cryptsetup open on the main system, run:

vgrename kubuntu-vg rescue-vg

Then you must verify that /etc/fstab in the RESCUE system contains no 
references to "kubuntu-vg" and you might also have to rerun 
"update-grub" if you want to keep it this way.

That is all you need to do to rename the volume group.

After that you can open the crypt and it won't conflict.

> Because I installed both
> from the Kubuntu 16.04 install ISO (from an USB stick), and this
> assigns "kubuntu-vg" to the VG name.

You are wise to say so, the LVM of 16.04 is not very good with 
conflicts.

(16.10 is better).


Make sure /etc/initramfs-tools/hooks/dmcache is executable:

chmod +x /etc/initramfs-tools/hooks/dmcache

Also rerun update-initramfs -u.

After, verify that the initramfs contains everything you need:

lsinitramfs /boot/initrd* | grep "pdata\|cache"

It has to contain at least:

usr/sbin/cache_restore
usr/sbin/cache_repair
usr/sbin/cache_metadata_size
usr/sbin/cache_dump
usr/sbin/cache_check
usr/sbin/pdata_tools
lib/modules/4.10.0-40-generic/kernel/drivers/md/dm-cache-cleaner.ko
lib/modules/4.10.0-40-generic/kernel/drivers/md/dm-cache-smq.ko
lib/modules/4.10.0-40-generic/kernel/drivers/md/dm-cache.ko

Or similar.

I assume you will be running this from your installed system. So you 
have only one chance to get it right before you need to reboot into the 
live or rescue environment.

If you end up on an initrd prompt (busybox) because your system doesn't 
boot, try to run:

   vgchange -ay
   vgchange -ay

   exit

To continue booting.