I have a AMD system with 3 200G and 1 250G hard drive arranged in a 4 drive
RAID 5 array. The 250G drive is connected to the /dev/hda, a DVD-RW on
/dev/hdc, the other 200G drives in /dev/hde, /dev/hdg and /dev/hdi. Two of
the controllers are ATA100/133 and the last is ATA66. /dev/hdi is connected
to the ATA66 controller. I have reiserfs on /dev/md0 using the entire array.
With this setup I have a read problem accessing files on the array. If I do
an md5sum on a file less then 200M it is fine. If I md5sum on a file greater
then 200M the results change every time I run it. Checking dmesg hda, hde
and hdg are UDMA(100) and hdi is UDMA(66). /proc/mdstat shows the array as
clean with all drives mounted. Any idea of what can be causing this problem?

Some history the array was built in degraded mode with hda, hde and hdg
only. The data was copied over and a cmp was done on the original and copied
data to insure it copied correctly. After the data move was verified. I took
down the 2 disk lvm array and put one of the drives in the RAID 5 array as
hdi. I then did a hot add of the new drive to complete the array. 

Joseph Key