//////////////////////////////////////////////////////////////////////////////////////
// DriveStatusError: Finding the offsending disk on multi-disk computer
// 
// Bryan Smith, Monday June 2 2008
//////////////////////////////////////////////////////////////////////////////////////

One of our servers with four disks sporadically displayed the following message:

---------------------------------------------------------------------
ata2: status=0x51 { DriveReady SeekComplete Error }
ata2: error=0x84 { DriveStatusError BadCRC }
---------------------------------------------------------------------

This server acts as a distributed file system of sorts, balancing the data across multiple disks.

As far as I can tell, the error is a result of failed data transmission from the disk to the controller (see related forum discussion), meaning that either a disk (or all), a cable or a controller was causing the problem.

I wrote a shell script to help find the problem. The above errors come from disk reads, so I simply diff every file with another file. I'll include the script below at the end of this post.

Using the shell script, I quickly found the disk. I tried replacing the cable, but the problem persisted, so I'm going through the process of backing up the data and replacing the disk.

(Potentially unrelated: another disk is chirping when being read, so I'm replacing it while I have the computer down.)

In this case, the hardware is failing. Here's the script I mentioned.

---------------------------------------------------------------------
---------------------------------------------------------------------
#!/bin/sh
# Takes in list of mounted disks, and reads files to attempt to replicate disk errors
# ./try_find_disk_error.sh [/path/to/mounted/disk1] {[/path/to/mounted/disk2], ...}
# Bryan Smith - bryanesmith@gmail.com on Monday June 2 2008

# Make sure user specified parameters
if [ $# -ne 0 ]; then

# Iterate each argument, which we expect to be a directory
while [ $# -gt 0 ]; do

# Grab next argument, which should be a directory
NEXT_DIR=$1
shift

echo
echo '-----------------------------------------------------------------'
echo

# Make sure each argument exists and is a directory
if [ ! -d $NEXT_DIR ]; then
echo "$NEXT_DIR does not exist. Skipping."
echo

# Exists, lets test
else
echo "Testing $NEXT_DIR"
echo

# Need to recursively find all items!
CONTENTS=`ls -R1 $NEXT_DIR`
#echo "$CONTENTS"

# Keep track of current directory.
# This will change as recursively iterate
# directories.
CURRENT_DIR="$NEXT_DIR";

for result in $CONTENTS ; do

# Try to match directories, which will end with colons
GREP_DIRECTORY=`echo $result | grep -i "\:\$"`
GREP_RESULT_FOR_DIRECTORY=$?

if [ $GREP_RESULT_FOR_DIRECTORY == 0 ] ; then
# Replace the final colon with a slash so have a path
CURRENT_DIR="${GREP_DIRECTORY/\://}"

# Remove comment to print out when
# changes to new current dir
#echo "---] SET CURRENT DIRECTORY: $CURRENT_DIR"

else

# Keep last item
LAST_FILE=$NEXT_FILE

# Create next file by prepending path to filename
NEXT_FILE="$CURRENT_DIR$result"

# Skip if last item doesn't exist. This happens on first
# run since no last file and every odd run after that.
if [ ! -z $LAST_FILE ]; then

#echo "last=$LAST_FILE; next=$NEXT_FILE"

# Do a diff. Might trigger read errors
diff $NEXT_FILE $LAST_FILE ] /dev/null 2]&1

# Don't need to read same file twice. Set next file to blank
# to skip diff next time around.
# This will effectively run a diff every two loops, comparing everything
# on disk.
NEXT_FILE=""
fi
fi

done
fi

# Give user opportunity to notice current iteration
sleep 3

done # Iterating each argument

echo
echo '-----------------------------------------------------------------'
echo

# Print usage message when no arguments
else
echo
echo 'No arguments provided.'
echo 'Usage: ./try_find_disk_error.sh [/path/to/mounted/disk1] {[/path/to/mounted/disk2], ...}'
echo
echo 'Description: tries to find a bad disk among multiple mounted disk by extensively reading (diff on each file). The arguments are simply directories, and do not necessarily need to be on separate disks, though that\'s the intended purpose.'
echo

exit 1
fi
---------------------------------------------------------------------
---------------------------------------------------------------------

Correction, Not the disk!

Actually, turns out the disk wasn't the problem! 

I put in a new disk, created a partition then formatted it, only to see the same DriveStatusError seek problems.

I then swapped the SATA cable with another disk that I knew was fine, and it threw the same errors!

Turns out that the SATA host adapter (that's the slot on the motherboard where you plug in the SATA cable from the disk) is bad. Just going to shuffle some disks around in another machine to make room!