Hi,
I wrote in yesterday about a serious file system corruption incident. This
issue is related, but separate, so I want to create a separate post.
To summarize yesterday's issue: A 1.75 TB xfs file system got corrupted.
Upon running xfs_repair, I ended up with about 3000 files in a lost+found
directory. They are listed by INODE number. The file names seem to have gotten
lost.
I don't know if this is normal or not for xfs_repair.
I am wondering if my particular setup method could have contributed to a
failure of xfs_repair. When I create a RAID-5 array with a 3ware 9000 SATA
card, I
get a 1.75 TB device -- let's call it /dev/sda. On the advice of some Linux
experts who should know about these things, I DO NOT create any partitions on
the device /dev/sda. I do not run fdisk on the device. When I put the xfs
filesystem on it, I use:
"mkfs.xfs /dev/sda"
xfs doesn't complain about not being on a partition. I haven't had any issues
until this one -- after many months and many machines that have been set up
this way. The only thing that worries me is that if a user opens up a disk
management utility such as the Mandrake "DiskDrake" program, he will get a
message
saying "I don't understand this device. It doesn't have any information that
I can understand. Do you want to create a partition on it?" I don't use
DiskDrake, so I don't really care about this message. And I have made it
difficult
for my users to run DiskDrake out in the field.
But does the fact that I don't run fdisk on the device, and the lack of at
least a single partition, in any way making the xfs file system vulnerable?
I would appreciate your speedy and thoughtful comments.
Regards,
Andy Liebman
|