Re: [gentoo-user] LVM, /usr and really really bad thoughts.

[Canek Peláez Valdés]

On Fri, Mar 9, 2012 at 8:48 PM, Dale <rdalek1967@×××××.com> wrote:
> Howdy,
>
> Well, this is what I am thinking about jumping into.  Ya'll ready for
> this?  I'm thinking about redoing my partition layout.  I'm wanting to
> keep / (root) on a normal ext4 file system.  I want to put /usr, /var,
> /home, and such on LVM.  I been using that dracut thingy to build the
> init thingy.  Sorry, I'm full of thingys tonight.  Maybe I need my meds?
>  Anyway, the init thingy seems to be working, I think.  I asked a while
> back how to tell for sure but it didn't get any replies so I am not real
> sure it is.  I do get this tho:
>
> root@fireball / # dmesg | grep init
> [    0.000000] Command line: root=/dev/sda3 init=/sbin/init
> [    0.000000] initial memory mapped : 0 - 20000000
> [    0.000000] init_memory_mapping: 0000000000000000-00000000bfc91000
> [    0.000000] init_memory_mapping: 0000000100000000-0000000440000000
> [    0.000000] Kernel command line: root=/dev/sda3 init=/sbin/init
> [    0.000000] Memory: 16387452k/17825792k available (6262k kernel code,
> 1052572k absent, 385768k reserved, 6647k data, 4852k init)
> [    0.003045] Security Framework initialized
> [    0.388120] SCSI subsystem initialized
> [    0.410739] pnp: PnP ACPI init
> [    0.787822] Trying to unpack rootfs image as initramfs...
> [    0.867787] Freeing initrd memory: 5084k freed
> [    0.880111] audit: initializing netlink socket (disabled)
> [    0.880439] type=2000 audit(1331081750.879:1): initialized
> [    0.912626] fuse init (API version 7.17)
> [    1.258561] ehci_hcd 0000:00:12.2: init command 0010005 (park)=0
> ithresh=1 period=512 RUN
> [    1.270152] ehci_hcd 0000:00:13.2: init command 0010005 (park)=0
> ithresh=1 period=512 RUN
> [    1.583458] device-mapper: ioctl: 4.22.0-ioctl (2011-10-19)
> initialised: dm-devel@××××××.com
> [    4.258421] init-early.sh used greatest stack depth: 3696 bytes left
> [    4.503735] init.sh used greatest stack depth: 3576 bytes left
> root@fireball / # dmesg | grep dracut
> [    3.018189] dracut: Checking reiserfs: /dev/sda3
> [    3.018531] dracut: issuing reiserfsck -a  /dev/sda3
> [    3.033879] dracut: Reiserfs super block in block 16 on 0x803 of
> format 3.6 with standard journal
> [    3.034463] dracut: Blocks (total/free): 4883760/2502678 by 4096 bytes
> [    3.034781] dracut: Filesystem is clean
> [    3.035210] dracut: Remounting /dev/sda3 with -o ro
> [    3.082413] dracut: Mounted root filesystem /dev/sda3
> [    3.158322] dracut: Switching root
> root@fireball / #
>
> And grub looks like this:
>
> title=Initramfs-new_kernel
> root (hd0,0)
> kernel /boot/bzImage-3.2.2-1 root=/dev/sda3 init=/sbin/init
> initrd /initramfs-3.2.2-1.img
>
> Does anyone think dracut is not working?  I need to make certain before
> diving into the next step.
>
> I have a second drive that is plenty large enough.  Thanks Kashani.  I
> plan to move everything currently to the larger drive then just sort of
> do a fresh install on my regular OS drive.
>
> One question I have right off the bat, how do I tell dracut to mount
> /usr?  I think it used to have a usr USE flag but that seems to have
> disappeared during a upgrade.  Is it magic?  Does it need to mount /var
> as well for logging?
>
> Just for the record, dracut is the only way I could get a init thingy to
> build and let me boot.  I tried different ways and they just didn't
> work.  At least I think dracut is working which is a good start.  ;-)
>
> I hope there is a few dracut users on here that have at least /usr on a
> separate partition.

I keep my /usr partition in /, but seeing the modules from dracut, the
"magic" happens at:

/usr/lib/dracut/modules.d/98usrmount/mount-usr.sh

Basically, it seems that if /usr is specified in /etc/fstab, then
dracut will mount it. It says nothing about LVM, but that is taken
care of in the scripts at:

/usr/lib/dracut/modules.d/90lvm

I'm not familiar with LVM, but it seems simple enough. And you can
create and modify your own dracut modules, of course.

Regards.
-- 
Canek Peláez Valdés
Posgrado en Ciencia e Ingeniería de la Computación
Universidad Nacional Autónoma de México