I successfully installed FC4 on a PII-333smp machine (Intel mboard), using two mirrored RAID1 18GB SCSI drives aic7xxx. I also have two Promise Ultra 100 TX2 controllers, with 4 160GB PATA hard drives connected as master to each of four IDE channels, partitioned as RAID autodetect and recognized as hde, hdg, hdi, hdk. I created a RAID5 array md4 with these 4 drives using mdadm, was able to format it (jfs filesystem in this case) and mount it (I created a mount point called /data, and modified fstab appropriately). I was able to copy data to it, unmount and remount the volume, and inspection of /proc/mdstat showed the array to be normal. When I rebooted the machine the array disappeared, no trace of md4, and using mdadm --scan resulting in an error message saying the array did not exist. I have now discovered that the partion tables for all four drives have been erased. This is repeatable, I did it twice. The RAID1 array on aic7xxx is unaffected, but the system crashed the first reboot while it tried to mount a non-existant RAID array (this is another serious bug! the system should gracefully recover if it can't mount an external filesystem!). The second time I refrained from modifying fstab until I verified the problem. I followed the mdadm guidance explicitly, and cannot make this work, there are no existing bug reports on this subject. I can do this task in Windows 2000/XP/2003 within five minutes, why is it so difficult to create a RAID5 array in Linux? Chris Cantwell
I don't see any reference to Amanda in this bug. What amanda configuration are you using, and what does your /etc/amanda/{config}/amanda.conf contain? Also, what backup medium are you using? Tape? Disk? CD?
This turned out not to be a bug in Amanda, I incorrectly used the mdadm command. I tried to configure the RAID5 using the device naem /dev/hde etc. instead of the partition names /dev/hde1 etc. Worked perfectly after I configured the RAID correctly. Closed.