Re: lost xfs filesystem on md0

[Simon Matter <simon.matter@ch.sauter-bc.com>]

james rich schrieb:
> 
> Hi folks,
>         I've got two old SCSI disks configured in a raid0 setup on
> /dev/md0.  I had an unplanned power loss and now I get the following error
> message:
> 
> scsi0: ERROR on channel 0, id 6, lun 0, CDB: Read (10) 00 00 00 00 38 00
> 00 08 00
> Info fld=0x3f, Current sd08:21: sense key Medium Error
>  I/O error: dev 08:21, sector 31

This seems to be a hardware error.

> I/O error in filesystem ("md(9,0)") meta-data dev 0x900 block 0x30:
>     xfs_trans_read_buf
> 
> The filesystem completes the mount process but no files show up
> (/var/spool is gone now :( ).  Does the above error indicate an error with
> hardware or with xfs?  In either case, is there some way to recover the
> contents of /dev/md0?

Since raid0 is in fact not redundant, you can most likely not recover.
What you could try is to dd the broken disk to a new one and then
startup the raid. Usually when XFS finds a hardware problem it shuts
down. With the new drive there may be some corrupt sectors which could
not be read from the old drive but you can then try to repair the fs
with xfs_repair. The problem will we to dd the old to the new disk
because when dd reaches a bad sector it will most likely abort. If so
you have to avoid that dd tries to read the bad sectors. It can be
tricky...

Simon

> 
> James Rich
> james.rich@m.cc.utah.edu