On Thu, 10 Jun 1999, Duncan Grove wrote:
Hi,
> I was running 4x9GB SCSI disks as a raid-0 /dev/md0 under redhat-5.2
> (I was using kernel 2.2.3). I upgraded the sytem to redhat-6.0 (with the
> standard modular 2.2.5 kernel), and things fell over. I tried running
> mkraid --upgrade but it barfed because of some problem with chunk-size.
> Never mind, I had a backup :) so I recreated the raid with mkraid and
> mke2fs and copied everything there. Rebooted the machine, and it fell
> over during startup while it was trying to detect the raid again.
> Anyway, it seems that if you mkraid --really-force (which I would expect
> to kill the fs) and then try and mount things (ie without mke2fs on it
> first) I get my old file system back.
mkraid only touchs the raid SuperBlock, it doesn't have anything
to do with filesystems. I'm not sure if one can expect a raid device fs to
be intact after a mkraid, though...
> I have a feeling this is maybe
> something to do with autoloading the raid-0 module... /proc/mdstat
> doesn't show anything unless I've run mkraid. Maybe I should build a
> kernel with raid-0 in built (not a module)?I know addraid and raidstart
Sure! either build it as a module, and use a initrd image, or just
into the kernel...
You have to mark the raid-involved partitions as type 'fd' while
the raid is not running, also... check
http://ostenfeld.dk/~jakob/Software-RAID.HOWTO/Software-RAID.HOWTO.html.
> are deprecated, but I tried them anyway to no avail. Has anyone else had
> this problem, or even better know of a solution? It's kind of urgent, or
> a bunch of chemists are going to kill me (it runs their home dirs for a
> Beowulf cluster). I can't keep bringing the machine up in single user
> mode to kludge bringing the disk up!
Hope it helps :)
*****---(*)---**********************************************---------->
Francisco J. Montilla Systems & Network administrator
[EMAIL PROTECTED] irc: pukka Seville Spain
INSFLUG (LiNUX) Coordinator. www.insflug.org - ftp.insflug.org
