invalid root filesystem error

[Maaz]

OS: SLES10 SP2
filesystem : reiserfs

the root(/) file system is a software RAID-5 consisting upon /dev/{sda3, sdb1, sdc1}

/dev/sda1 is /boot
/dev/sda2 is swap

I dont know how, but this sles10sp2 does not boot properly, please help me properly boot this machine.

following are the console output, for your kind consideration

SCSI device sdc: drive cache: write back
SCSI device sdd: 1953525168 512-byte hdwr sectors (1000205 MB)
sdd: Write Protect is off
SCSI device sdd: drive cache: write back
md: bind
md: bind
mdadm: /dev/md0 assembled from 2 drives - not enough to start the array while not clean - consider -- force.
attempting manual resume
Waiting for device /dev/md0 to appear: ok
rootfs: major=9 minor=0 devn=2304
/dev/md0: unknown volume type
invalid root filesystem -- exiting to /bin/sh
$

please help me boot this machine properly

Regards
Maaz

[Steven E. Protter]

Shalom Maaz,

One of your three disks, sdc1 is not a part of the disk array any more.

You will need to boot the system off install media and see about rebuilding the arrray.

I find your set up, booting the system off a three disk raid 5 array very problematic. There are not enough disks in the array to make it reliable.

SEP

[Maaz]

Hi SEP

Thanks for providing help.

I tried booting the system in rescue via installation media, and tried following commands, without luck

# mdadm --assemble --force /dev/md0
mdadm: /dev/md0 not identified in config file.

# mdadm -A /dev/md0 --run
mdadm: /dev/md0 not identified in config file.

# mdadm -A /dev/md0 --run --force
mdadm: /dev/md0 not identified in config file.

now I have rebuild the server.

Regards
Maaz

[Matti_Kurkela]

It may be too late now, but the error message "mdadm: /dev/md0 not identified in config file" is trying to tell you there is no information about md0 in the config file, and you haven't given mdadm permission to figure it out automatically.

I think you should have tried:
mdadm --assemble --scan --force

When the system is starting, it does something like "mdadm --assemble --scan", without the force.

MK

[TwoProc]

Maaz,

It's not what you're asking, but I can tell you how to avoid this situation in the future.

Before you rebuild that system - go out and purchase a "Compaq Smart Array" hd controller, which provides a very robust hardware mirroring solution, and doesn't depend on software mirroring or software tools to config/restore the mirrors (a la Perstor needing a Windows application to restore a hardware mirror). The are many models of the Smart Array controller to choose from, and in general, the more $$ you pay the bigger the cache. And don't forget to spring the extra $$ for the optional battery backed write through cache card. It makes the thing amazing for throughput on top of the amazing robustness of the hardware mirroring and software contained in rom for restores, etc. Another advantage, it makes cloning to another system, or from a saved harddrive (cold standby) of a pulled mirror drive a cinch. Can't recommend them enough. They are standard faire in the ProLiant server series.

Of course, I use them all the time for SCSI drives, so you'll need to check to see if they support other drive types, but I assume that by now, they do.

[Maaz]

Hi Thanks Matti Kurkela
>I think you should have tried:
>mdadm --assemble --scan --force
Oh, I missed ;(. Anyways thanks a million for informing.

Hi TwoProc
Thanks for suggestions, this is the only system having Software RAID, rest are on using HW RAID.

Regards
Maaz