Whenever two drives plugged in, second drive fails?

Whenever two drives plugged in, second drive fails?

[Jeff Hill]

A bit of a problem that appears to be more complex than I thought:

When I try to add a recently repaired drive to my SCSI chain, the Adaptec 
2940U2W controller  times out trying to find it, even reports occasionally 
a "cable or termination error, please fix."

Sounds simple, but I don't think it's termination on the drives, since 
they're an identical pair of Seagate Cheetah LVD drives.

  The LVD cable is an expensive custom, teflon job with active termination, 
but I decided to try using my spare LVD cable, the original from Adaptec. 
Same result. I've also tried changing the order of the drives, but there is 
little difference: when the repaired drive is the first one in the chain, 
the controller just seems to freeze up faster.

My old SCSI drive works fine alone on the cable as sda: the  controller 
recognizes it without any problem and sda boots up as the first drive in a 
software RAID-1 where sda is the only drive in the array (I was waiting for 
the repaired drive [which came out of another system] to return).

I started to suspect it was some fault with the repaired drive. After 
trying a number of tests, I finally put it on the scsi cable by itself, 
booting the system off an IDE drive. In this configuration, the drive is 
recognized with no problem. I partitioned it, did a bad block check, put 
some ext2 file systems on it, did some further testing, couldn't find any 
problem with the repaired drive.

Then I thought maybe it was a power issue, so I unplugged the IDE drives 
and CD-ROM to make certain that wasn't it (even though I have a pretty good 
300Watt power supply). No change.

Any suggestions on what else to try would be appreciated.

Regards,

Jeff Hill



----
Jeff Hill
www.HRpost.com
Ph: 416-047251
Fx: 647-439-1414
----

Re: Whenever two drives plugged in, second drive fails?

[Doug Ledford]

On Thu, Jun 20, 2002 at 02:15:36PM -0400, Jeff Hill wrote:
> When I try to add a recently repaired drive to my SCSI chain, the Adaptec 
> 2940U2W controller  times out trying to find it, even reports occasionally 
> a "cable or termination error, please fix."
> 
> Sounds simple, but I don't think it's termination on the drives, since 
> they're an identical pair of Seagate Cheetah LVD drives.

[ snipped description ]

Hmmm...from what you've given I almost wonder if there isn't a termination 
power issue with too many devices providing term power.  I would check the 
drives to make sure they don't have term power enabled since it shouldn't 
be needed.  The other thing that it could be is actually an underpowering 
of term power with the two drives pulling more power from the reset line 
than the term power is capable of providing, resulting in the bus seeing 
infinite resets when both drives are plugged in.

(Note: it's been a long time since I read the electrical specs of the LVD
SPI bus.  Don't flame me too hard on innaccuracies in this bit.  It may
not be technically correct in all places, but the basic information
provided is correct. The reset line on the SCSI cable works by having
something on the bus provide +2v (I think, it may be +1v on LVD, been a
while since I read the electrical specs on LVD) to the terminator which
then connects that power source to the reset line on the bus, pulling the
voltage on the reset line up high.  When the controller or any device on
the bus wants to cause a bus reset, it temporarily connects the reset line
to ground, pulling the voltage low and triggering all the devices to
reset.  However, reading the voltage on the line requires consuming a
small amount of the power provided by the terminator.  So, the more
devices you have on the bus, the more trickle power the terminator has to
supply in order to keep the voltage on the reset line up.  The trickle
power is limited though, or else devices wouldn't be able to pull it low
at will.  So, if the terminator is providing too small of an amount of
power on the line, then two drives can force it low by just reading the
line, resulting in what you are seeing.  On the other hand, it could be
that the terminator is in fact providing all the power it is suppossed to,
but each of the drives you have are pulling way too much power when just
sitting there and as it may work out, each device will work on a bus by
itself but when they are put on a bus together the combined power pull
becomes too much for the terminator and the bus goes into a permanent
reset state.  This would be my actual guess for what's wrong on your bus.)

-- 
  Doug Ledford <dledford@redhat.com>     919-754-3700 x44233
         Red Hat, Inc. 
         1801 Varsity Dr.
         Raleigh, NC 27606