I'm trying to solve a corrupted file system problem on a cluster front end.

The filesytem is (or was ) 500Gb ext3.
We had a small electrical power failure yesterday, the server do not stop but the disk array (SCSI Raid 5 disk system) restarted. This morning, the filesystem was not available (I/O error) so I reboot the front end. The fsck failed with the message:
root inode is not a directory

There are nearly 400Gb of data on this filesystem. Any idea to solve the problem ? Google always point to a commercial software or windows software...

Thanks

Patrick

If it is a SCSI Raid 5 disk system, you probably want to look there for your answer. You may have to wait until it can rebuild itself. Then again, it may be too corrupted, but that's probably where you'll want to look to find out. That's the downside of raid 5. You get your redundancy and higher capacity than raid 1, but if you have a disk problem, you have to rebuild a drive w/ parity calculations rather than simply copying the drive. Hopefully, the error you are getting is the disk array attempting to prevent you from writing to it until it can finish all of it's calculations and get your array back online.

HTH

Forrest

I don't think it is related to the Raid because I haven't any failure on the disks. It look like if the ext3 filesystems was corrupted. Runing fsck says:

Missing '..' in directory inode 48922819.
Fix<y>? yes

Entry '..' in ... (48922819) points to inode (2) located in a bad block.
Clear<y>? yes

for many directories and then:


Pass 3: Checking directory connectivity
Root inode is not a directory; aborting.
e2fsck: aborted