mount-by-label finds drive not array ????

mount-by-label finds drive not array ????

[Tom Oehser]

Hi,

I'm accustomed to doing something like:

LABEL=tomsvol2  /mnt/foo   ext3      defaults        0 2

But if I do that on my raid-5 md arrays, it tries to mount
disk 1 of the 4 disks that make up the array!  Apparently,
the first stripe has enough of the filesystem that it must
be thinking that is what to mount!  Even with type "FD"!!!

So, with an array /dev/md0 made of hda1 hdb1 hdc1 hdd1, I
am having to do a literal:

/dev/md0        /mnt/bar      ext3      defaults        0 4

Since if I do it by "LABEL=", it tries to mount hda1, not md0!

Is this just a known/accepted brain-deadness of the mount code?

-Tom

Re: mount-by-label finds drive not array ????

[Evan Felix]

I've always found the mount by Label system to be a little confused
about things.  My specific issues cam from haveing devfs turned on in
the kernel, but not mounting it.

Basically the mount code scans /proc/partitions and examines each one in
turn,  which usually lists devices as the were 'discovered' or in the
order device drivers were loaded.  so it sounds like the IDE drives come
up first, a valid label is found, and it tries to mount it.  if you
could magically load the md code before your ide code it would find it
first.(build the raid code into your kernel?)

I would suggest avoiding mount by label like the plauge, but thats just
me.  It just consuses things, doesent always work and makes reading
/etc/fstab confusing.

Evan

On Thu, 2004-07-01 at 04:41, Tom Oehser wrote:
> Hi,
> 
> I'm accustomed to doing something like:
> 
> LABEL=tomsvol2  /mnt/foo   ext3      defaults        0 2
> 
> But if I do that on my raid-5 md arrays, it tries to mount
> disk 1 of the 4 disks that make up the array!  Apparently,
> the first stripe has enough of the filesystem that it must
> be thinking that is what to mount!  Even with type "FD"!!!
> 
> So, with an array /dev/md0 made of hda1 hdb1 hdc1 hdd1, I
> am having to do a literal:
> 
> /dev/md0        /mnt/bar      ext3      defaults        0 4
> 
> Since if I do it by "LABEL=", it tries to mount hda1, not md0!
> 
> Is this just a known/accepted brain-deadness of the mount code?
> 
> -Tom
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: mount-by-label finds drive not array ????

[robin-lists]

> -----Original Message-----
> From: linux-raid-owner@vger.kernel.org 
> [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Tom Oehser
> Sent: 01 July 2004 11:41
> To: linux-raid@vger.kernel.org
> Subject: mount-by-label finds drive not array ????
> 
> 
> Hi,
> 
> I'm accustomed to doing something like:
> 
> LABEL=tomsvol2  /mnt/foo   ext3      defaults        0 2
> 
> But if I do that on my raid-5 md arrays, it tries to mount 
> disk 1 of the 4 disks that make up the array!  Apparently, 
> the first stripe has enough of the filesystem that it must be 
> thinking that is what to mount!  Even with type "FD"!!!
> 
> So, with an array /dev/md0 made of hda1 hdb1 hdc1 hdd1, I am 
> having to do a literal:
> 
> /dev/md0        /mnt/bar      ext3      defaults        0 4
> 
> Since if I do it by "LABEL=", it tries to mount hda1, not md0!
> 
> Is this just a known/accepted brain-deadness of the mount code?

Tom,

I'm just considering doing something like this.

Did you ever get any resolution to this issue?

Thanks,

R.
--
http://robinbowes.com  

Re: mount-by-label finds drive not array ????

[Luca Berra]

On Thu, Jul 22, 2004 at 05:33:13PM +0100, robin-lists@robinbowes.com wrote:
>> 
>> /dev/md0        /mnt/bar      ext3      defaults        0 4
>> 
>> Since if I do it by "LABEL=", it tries to mount hda1, not md0!
>> 
>> Is this just a known/accepted brain-deadness of the mount code?
>
>Tom,
>
>I'm just considering doing something like this.
>
>Did you ever get any resolution to this issue?
is there any need to mount a md device by label, since you can already
identify the md components by uuid when you assemble the raid.

L.

-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \

Re: mount-by-label finds drive not array ????

[TJ Harrell]

Check the archives for the list.. There was a discussion on labels within
the last two months, I believe. The only solution was to avoid mounting by
label.


----- Original Message ----- 
From: <robin-lists@robinbowes.com>
To: "'Tom Oehser'" <tom@toms.net>; <linux-raid@vger.kernel.org>
Sent: Thursday, July 22, 2004 12:33 PM
Subject: RE: mount-by-label finds drive not array ????


> > -----Original Message-----
> > From: linux-raid-owner@vger.kernel.org
> > [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Tom Oehser
> > Sent: 01 July 2004 11:41
> > To: linux-raid@vger.kernel.org
> > Subject: mount-by-label finds drive not array ????
> >
> >
> > Hi,
> >
> > I'm accustomed to doing something like:
> >
> > LABEL=tomsvol2  /mnt/foo   ext3      defaults        0 2
> >
> > But if I do that on my raid-5 md arrays, it tries to mount
> > disk 1 of the 4 disks that make up the array!  Apparently,
> > the first stripe has enough of the filesystem that it must be
> > thinking that is what to mount!  Even with type "FD"!!!
> >
> > So, with an array /dev/md0 made of hda1 hdb1 hdc1 hdd1, I am
> > having to do a literal:
> >
> > /dev/md0        /mnt/bar      ext3      defaults        0 4
> >
> > Since if I do it by "LABEL=", it tries to mount hda1, not md0!
> >
> > Is this just a known/accepted brain-deadness of the mount code?
>
> Tom,
>
> I'm just considering doing something like this.
>
> Did you ever get any resolution to this issue?
>
> Thanks,
>
> R.
> --
> http://robinbowes.com
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

RE: mount-by-label finds drive not array ????

[robin-lists]

> -----Original Message-----
> From: linux-raid-owner@vger.kernel.org 
> [mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Luca Berra
> Sent: 22 July 2004 17:54
> To: linux-raid@vger.kernel.org
> Subject: Re: mount-by-label finds drive not array ????
> 
> On Thu, Jul 22, 2004 at 05:33:13PM +0100, 
> robin-lists@robinbowes.com wrote:
> >> 
> >> /dev/md0        /mnt/bar      ext3      defaults        0 4
> >> 
> >> Since if I do it by "LABEL=", it tries to mount hda1, not md0!
> >> 
> >> Is this just a known/accepted brain-deadness of the mount code?
> >
> >Tom,
> >
> >I'm just considering doing something like this.
> >
> >Did you ever get any resolution to this issue?
> is there any need to mount a md device by label, since you 
> can already identify the md components by uuid when you 
> assemble the raid.

Probably not. It's just something I'm trying.

Incidentally, it seems to work OK for me.

I'm using Fedora Core 2. Here's my /etc/fstab (only labelled partitions
listed):

LABEL=dude_root         /                       ext3    defaults        1 1
LABEL=hda_music         /home/slimserver/music  ext3    defaults        0 0
LABEL=/                 /mnt/hda                ext3    defaults        1 1
LABEL=/boot             /mnt/hda/boot           ext3    defaults        1 2
LABEL=/home             /mnt/hda/home           ext3    defaults        1 2
LABEL=/opt              /mnt/hda/opt            ext3    defaults        1 2
LABEL=/tmp              /mnt/hda/tmp            ext3    defaults        1 2
LABEL=/usr              /mnt/hda/usr            ext3    defaults        1 2
LABEL=/usr/local        /mnt/hda/usr/local      ext3    defaults        1 2
LABEL=/var              /mnt/hda/var            ext3    defaults        1 2

And here's what get's mounted:

/dev/md0 on / type ext3 (rw)
/dev/hda11 on /home/slimserver/music type ext3 (rw)
/dev/hda7 on /mnt/hda type ext3 (rw)
/dev/hda1 on /mnt/hda/boot type ext3 (rw)
/dev/hda3 on /mnt/hda/home type ext3 (rw)
/dev/hda5 on /mnt/hda/opt type ext3 (rw)
/dev/hda10 on /mnt/hda/tmp type ext3 (rw)
/dev/hda2 on /mnt/hda/usr type ext3 (rw)
/dev/hda6 on /mnt/hda/usr/local type ext3 (rw)
/dev/hda9 on /mnt/hda/var type ext3 (rw)

The reason I decided I'd like to use labels is because I found I was
spending far too much time trying to find file systems on old disks to copy
data across in a new system.

Take the example above: /dev/hda is from another system and I wanted to
mount all partitions to check some config settings. I found I was doing
stuff like:

# cat /proc/partitions
major minor  #blocks  name

   3     0  120060864 hda
   3     1     104391 hda1
   3     2    3068415 hda2
   3     3    2048287 hda3
   3     4          1 hda4
etc.

# mkdir /mnt/hda1 ; mount /dev/hda1 /mnt/hda1
# ls /mnt/hda1
(ah, /dev/hda1 must me the old /boot partition)
# umount /dev/hda1
etc.

Eventually I'd find the root filesystem, mount it, then mount all the other
filesystems on the old root fs.

With labels I can just do this:

# findfs LABEL=hostname_root
/dev/hda7

# mount /dev/hda7 /mnt/hda

Or even:

# mount `findfs LABEL=hostname_root` /mnt/hda

Or even;

# for d in usr usr/local tmp home var ; do mount `findfs LABEL=hostname_$d`
/dev/hda/$d ; done

R.