[CLUE-Tech] Software RAID stupidity

[Dave Hahn]

Hey Jim:

I've seen this twice before.  Both times the failure of the RAID to load was
related to partition information; once when I replaced a drive in the RAID
and the partition was flagged as type 0x83 instead of 0xfd.  The second time
was when the super block on the raid device was invalid.

I always found some hints in dmesg.

The bad part of the lesson was that neither raid worked correctly afterward.
However, if you can still mount the raid under 2.2.14 it sounds like you are
better off than I was.

-----Original Message-----
From: clue-tech-admin at clue.denver.co.us
[mailto:clue-tech-admin at clue.denver.co.us]On Behalf Of Jim Ockers
Sent: Wednesday, October 25, 2000 3:17 PM
To: clue-tech at clue.denver.co.us
Subject: [CLUE-Tech] Software RAID stupidity


Hello y'all,

I am temporarily stumped by this and thought I'd ask the group, in case
there is anyone who has any suggestions.

I have a Red Hat 6.2 system with Linux 2.2.14 on it.  I wanted some device
stuff that is available on the 2.2.1[67] kernels from a 2.4 backport, so I
thought I'd install kernel 2.2.17 .

Well, everything's fine except that my RAID filesystem will not start
under 2.2.17 .  Everything else, including sound, networking, etc., works
fine.  (I haven't gotten to test the new device drivers though because
the RAID filesystem has my /home directory on it, which is important to me.)

The RAID personality I'm using is RAID0, striped.  The raid0.o module is
loaded.  The "/sbin/raidstart" command says this:

/dev/md0: Invalid argument

For more technical information, here is the full set of parameters etc.
under 2.2.14 (under which the RAID works fine):

[15] root at niamey.ockers.net:/tmp > cat /etc/raidtab
raiddev             /dev/md0
raid-level                  0
nr-raid-disks               2
chunk-size                  64k
persistent-superblock       1
#nr-spare-disks     0
    device          /dev/hdc1
    raid-disk     0
    device          /dev/hdd1
    raid-disk     1

[16] root at niamey.ockers.net:/tmp > cat /proc/mdstat
Personalities : [raid0]
read_ahead 1024 sectors
md0 : active raid0 hdd1[1] hdc1[0] 80035584 blocks 64k chunks
unused devices: <none>
[17] root at niamey.ockers.net:/tmp > lsmod
Module                  Size  Used by
raid0                   3036   1
<deletia..>

Now, under 2.2.17, the /etc/raidtab hasn't changed, but /proc/mdstat says
this:

[19] root at niamey.ockers.net:/tmp > cat /proc/mdstat
Personalities : [2 raid0]
read_ahead not set
md0 : inactive
md1 : inactive
md2 : inactive
md3 : inactive

[22] root at niamey.ockers.net:/tmp > cat lsmod.2217
Module                  Size  Used by
raid0                   2104   0  (unused)
<deletia..>

The "size" as reported by lsmod is smaller than it was under 2.2.14!!  The
following are the actual file sizes (and the 2.2.17 one is smaller):

[24] root at niamey.ockers.net:/lib/modules > find . -name 'raid0.o' -exec
ls -l {} \;
-rw-r--r--    1 root     root         5140 Mar  7  2000
./2.2.14-5.0/block/raid0.o
-rw-r--r--    1 root     root         3684 Oct 25 13:03
./2.2.17/block/raid0.o

The module loads under 2.2.17 with no complains.  Nothing whatsoever is
logged to syslog.

That "invalid argument" thing is getting me down.  The only time I've seen
it before is under Linux 2.0.x, when the actual RAID devices refused to
start themselves for mdrun, due to a corrupted RAID-5 array.  I ran ckraid
to rebuild the RAID array, and then mdrun would start the RAID stuff OK.

--
Jim Ockers (ockers at ockers.net)                     Ask me about Linux!
Contact info: please see http://www.ockers.net/

Fight Spam! Join CAUCE (Coalition Against Unsolicited Commercial Email)
at http://www.cauce.org/ .

_______________________________________________
CLUE-Tech mailing list
CLUE-Tech at clue.denver.co.us
http://clue.denver.co.us/mailman/listinfo/clue-tech