Root mount failed:5

Matthew Dillon dillon at apollo.backplane.com
Thu Oct 11 08:08:59 PDT 2012


:
:On Thu, Oct 11, 2012 at 5:16 PM, Sascha Wildner <saw at online.de> wrote:
:>> #mount /dev/ad8s1d /mnt
:>> hammer:mount   on /mnt: Input/Output error
:>
:>
:> That should surely be mount_hammer there?
:>
:
:mount_hammer gives same error :-(

    The (recent) mount code looks at the filesystem type field in the
    partition now so just 'mount' should work.

    I'm a bit perplexed but there are a couple of places where HAMMER
    doesn't kprintf() the problem when returning an EIO.

    Your only choice here might be to run a hammer recover scan to recover
    the data.  The 'hammer recover' directive is not like fsck... it isn't
    run in-place.  It scans the media and regenerates the filesystem to a
    target directory.   You need to have a second drive mounted to run it.

    How large is this drive?  If there were a way for me to obtain an image
    of it (might not be possible if its hundreds of gigabytes)... then I
    could debug it here.

    If not then perhaps you could do a 'hammer show' dump.  This will
    generate a huge text output file.  e.g. something like:

	hammer -vvv -f <blockdevice> show > hugefile

    That won't fix anything but might help me debug the issue.

    --

    * 'hammer recover' might be your only way to recover the data, requires
      a separately mounted filesystem because it doesn't run in-place.

    * 'hammer show' output might help me figure out what happened, but
      won't help you recover the data.

    * Are you using any special features such as live dedup?  Live dedup
      is known to have issues, but background (hammer cleanup based) dedup
      should be fine.  I run background dedup on all my boxes.

    * How old is the filesystem?

    * No AHCI on this system?  It's mounted via /dev/adXX instead of
      /dev/daXX.  IDE or SATA disk?

					-Matt
					Matthew Dillon 
					<dillon at backplane.com>



More information about the Users mailing list