>Hi,
>
>> > I've seen this problem on two 133 MB/sec IDE cards - one gave the same
>> > problems you're seeing, and the other crashed the system when using both
>> > channels heavily. I decided to use only one drive per IDE card for those
>> > particular cards.
>>
>> FWIW, my similar problems are occuring on a one-drive, one-card system (a
>> PM 7300 with a Sonnet G3/500, and a Sonnet Tempo Trio/133 connected to a
>> Western Digital 40GB ATA/100).
>
>All the other times I've seen lost interrupt messages have been due to
>bad cables. Have you tried another cable? There can be a lot of cable
>voodoo with IDE at higher speeds.
>
>John Klos

I'll check into the power supply issue.  I actually removed the CD-ROM and
floppy drives to compensate somewhat for the "extra" drive, but I believe
that the Beige should handle the load OK.  Newer drives are far thriftier
with power than older SCSI drives, and I know that the Beige could handle
two of the older drives.

I don't have the right kind of equipment to accurately measure ripples
and dropouts in my power supply output of the kind that would be caused
by disk seeks and such, so beyond some crude measurements, and removing the
internal disk for testing, I can't really determine whether this is really
caused by PS "weakness".

Cable issues are also a candidate, but should also be independent of activity
on the "other" drive, so I'm doubting that this is my problem.

There are two cables from different
manufacturers.  I can load either disk up with activity, and it will
run fine.  While the one is pounding away, if I "touch" the other one,
I get the hiccups.  The internal drive can be busy or idle, and it makes
no difference.  "touch" in this context is something like a "du -s dir" or
something with a non-trivial number of disk accesses.  The quick ls or
df does not trigger hiccups.

IMHO, behavior suggests a software issue.
I plan to try putting both drives on a single channel and cable tonight,
and we'll see if this makes a difference.  If I get brave, I'll build
some experimental kernels this weekend and see if I can make it behave
..... differently... ;-o

-dgl-