sense errors..

Ben Scott dragonhawk at gmail.com
Thu Jan 4 01:50:39 CST 2007


On 1/2/07, Kevin Foote <kevin.foote at gmail.com> wrote:
> hostX kernel: scsi0: ERROR on channel 0, id 0, lun 0, CDB: ...

  Error on controller (card) zero, channel (bus) zero, target zero,
logical unit zero.  "CDB" stands for "Command Descriptor Block" (or
something like that) and is only useful to people who know the SCSI
protocol in binary.

> hostX kernel: Info fld=0x88b8f17, Current sda: sense key Medium Error

  In SCSI-land, "sense" is the mechanism for reporting detailed
information.  The target (disk) just reports that it had an error; the
initiator (computer) has to then ask the target for the "sense data"
to find out the details about the error.

  "Medium Error" means the target (disk) reported a problem with the
physical medium (i.e., the disk itself).  Note that this is just what
the target reported, not necessarily what is actually wrong.  Hardware
can have bugs, too.

> hostX kernel: Additional sense: Data synchronization mark error

  "Data synchronization mark error" is the specific error code.  I've
never seen that before and have no idea what it means.  Google mostly
finds listings of that error in the SCSI spec documents.

  So that's a weird one.  It seems to happy a lot, though.  That's not good.

> hostX kernel: end_request: I/O error, dev sda, sector 143363863

  That's being reported by the higher levels of the kernel.  It
basically means that it tried to read block 143363863 and it failed
("I/O error").  The more specific problem has already been logged.

  What I find interesting is that location is very near the end of a
"73 GB" disk.  In fact, it may be the last block.  In fact, it may be
*beyond* the last block.  In other words, the system may be trying to
read beyond the end of the disk.  There's a few different reasons why
that might happen (e.g., corrupt file system, corrupt partition table,
faulty disk capacity identification), but it's something to keep in
mind.

  If you can afford to lose the contents of a disk, try a destructive
write test:

**** WARNING: THIS WILL DESTROY ALL DATA ON THE DISK ****

badblocks -s -v -w /dev/sda

**** WARNING: THAT WILL DESTROY ALL DATA ON THE DISK ****

> hostX kernel: Additional sense: Unrecovered read error

  That doesn't sound good either.

-- Ben



More information about the Linux-PowerEdge mailing list