Hi Members
I tested a lot more things with the udev and by-uuid...
Miro Dietiker schrieb:
I am running sid on a PPC (pegasos). After deciding to change my fstab
to use UUID (but then on second thought to labels instead) I read many
tutorials claiming that there should be a /dev/disk/by-uuid/ directory.
My system has only a by-path and by-id.
That's exactly what happened to me right now too with an amd64 system.
Did you find a solution?
Looks like a md-device (showing up in /sys/block/mdX) has no UUID in udev.
Also lvm volumes (showing up in /sys/block/dm-X) have no UUID in udev.
Since lvm-paths from wrapper are persistent, UUIDs are not needed to
access them.
Since md devices store participants and latest mounting information in
superblock to restore, mdX numbers are persistent too and we don't need
any UUID.
But: when adding new disks inside a system with conflicting md
superblocks (say 2 disks tell to be /dev/md0 raid1 and another 2 disks
tell to be /dev/md0 raid1 from an earlyer production...) things go
wrong. Possibly ancient disks are mounted as md0 where the newer
(possibly original root array) from system is assembled elsewhere.
I don't understand why the UUID of the md superblock could be used for
accessing the md device and getting some distance from the strict mdX
persistency (which could not be guaranteed in any case)...
Are there any inputs on this?
== udevinfo START
# udevinfo -a -p $(udevinfo -q path -n /dev/md2)
looking at device '/block/md2':
KERNEL=="md2"
SUBSYSTEM=="block"
DRIVER==""
ATTR{stat}==" 43 0 296 0 2
0 4 0 0 0 0"
ATTR{size}=="2007808"
ATTR{removable}=="0"
ATTR{range}=="1"
ATTR{dev}=="9:2"
== udevinfo END
Using the UUID for /boot/grub/menu.lst like:
# kopt=root=UUID=xxxxxxxxxx ro
system boot fails (no uuid of md found since it tries to access via
/dev/disk/by_uuid/xxxxxx implicitly)
What's wrong with the default system setup? How to use UUID of md on boot?!
Is it even possible to get uuid in that environment?
ThX for all inputs!
--
Miro Dietiker
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]