In article ,
Thomas Maier-Komor wrote:

> Hi all,
>
> I have attached a sun storedge multipack (ultra-scsi version with 6
> slots) to an U160 controller. But the busspeed is stuck at 40MB/s,
> although only U160 devices are on the bus. So I attached an external LVD
> terminator, but this did not change anything.
>
> What might be the reason that the bus is still in HVD mode?
>
[snip]

> Can anybody make anything out of this? I am clueless...
> I would be grateful for any hint.
>
> Tom

See this page for more info:

http://scsifaq.paralan.com/scsifaqanswers.html#8

AFAIK, the controller hardware built into external multipacks isn't
"multi-level" allowing for switching between "High Voltage Differential"
and "Low Voltage Differential", which refers to the signal voltage
levels in the controller and cabling. Usually if there's a mismatch
with the type of controller, you get no communication (LVD device -> HVD
adapter) or fried circuitry (HVD device -> LVD adapter).

I think what you're seeing is the bus not synchronizing at ultra SCSI
_speed_ levels, which has nothing to do with voltage. Usually this is
because there's something else on the bus that won't go at Ultra-SCSI
speeds like a CDROM or a tape drive or the cable is to long. What else
is on the bus? How long is the cable connecting to the box?

--
DeeDee, don't press that button! DeeDee! NO! Dee...

Michael Vilain wrote:
> In article ,
> Thomas Maier-Komor wrote:
>
>
>>Hi all,
>>
>>I have attached a sun storedge multipack (ultra-scsi version with 6
>>slots) to an U160 controller. But the busspeed is stuck at 40MB/s,
>>although only U160 devices are on the bus. So I attached an external LVD
>>terminator, but this did not change anything.
>>
>>What might be the reason that the bus is still in HVD mode?
>>
>
> [snip]
>
>
>>Can anybody make anything out of this? I am clueless...
>>I would be grateful for any hint.
>>
>>Tom
>
>
> See this page for more info:
>
> http://scsifaq.paralan.com/scsifaqanswers.html#8
>
> AFAIK, the controller hardware built into external multipacks isn't
> "multi-level" allowing for switching between "High Voltage Differential"
> and "Low Voltage Differential", which refers to the signal voltage
> levels in the controller and cabling. Usually if there's a mismatch
> with the type of controller, you get no communication (LVD device -> HVD
> adapter) or fried circuitry (HVD device -> LVD adapter).
>
> I think what you're seeing is the bus not synchronizing at ultra SCSI
> _speed_ levels, which has nothing to do with voltage. Usually this is
> because there's something else on the bus that won't go at Ultra-SCSI
> speeds like a CDROM or a tape drive or the cable is to long. What else
> is on the bus? How long is the cable connecting to the box?
>

well the multipack does not have its own controller, like D1000 or the
like. So this should not be an issue. There are only three disks on the
bus, which are in the multipack. It is attached to the HBA with a 2
meter HD68 to VHCI cable. The autotermination of the multipack is off
and the bus is terminated with a LVD ultra terminator.

Are there any tools that support analysis of the scsi bus? Is it
possible to make the driver write syslog messages that describe why it
is negotiating at a certain speed? The HBA is the onboard controller of
a Blade 2500 (mpt driver).

TIA,
Tom

In article ,
Thomas Maier-Komor wrote:

> Michael Vilain wrote:
> > In article ,
> > Thomas Maier-Komor wrote:
> >
> >
> >>Hi all,
> >>
> >>I have attached a sun storedge multipack (ultra-scsi version with 6
> >>slots) to an U160 controller. But the busspeed is stuck at 40MB/s,
> >>although only U160 devices are on the bus. So I attached an external LVD
> >>terminator, but this did not change anything.
> >>
> >>What might be the reason that the bus is still in HVD mode?
> >>
> >
> > [snip]
> >
> >
> >>Can anybody make anything out of this? I am clueless...
> >>I would be grateful for any hint.
> >>
> >>Tom
> >
> >
> > See this page for more info:
> >
> > http://scsifaq.paralan.com/scsifaqanswers.html#8
> >
> > AFAIK, the controller hardware built into external multipacks isn't
> > "multi-level" allowing for switching between "High Voltage Differential"
> > and "Low Voltage Differential", which refers to the signal voltage
> > levels in the controller and cabling. Usually if there's a mismatch
> > with the type of controller, you get no communication (LVD device -> HVD
> > adapter) or fried circuitry (HVD device -> LVD adapter).
> >
> > I think what you're seeing is the bus not synchronizing at ultra SCSI
> > _speed_ levels, which has nothing to do with voltage. Usually this is
> > because there's something else on the bus that won't go at Ultra-SCSI
> > speeds like a CDROM or a tape drive or the cable is to long. What else
> > is on the bus? How long is the cable connecting to the box?
> >
>
> well the multipack does not have its own controller, like D1000 or the
> like. So this should not be an issue. There are only three disks on the
> bus, which are in the multipack. It is attached to the HBA with a 2
> meter HD68 to VHCI cable. The autotermination of the multipack is off
> and the bus is terminated with a LVD ultra terminator.

Let me clarify my language so we're on the same page. The multipack has
circuitry to connect the disks to a controller or HBA. This is contract
to a D1000 or A1000 which has it's own controller which talks to an HBA
and does some intelligent processing and even caching of data.

>
> Are there any tools that support analysis of the scsi bus? Is it
> possible to make the driver write syslog messages that describe why it
> is negotiating at a certain speed? The HBA is the onboard controller of
> a Blade 2500 (mpt driver).
>
> TIA,
> Tom

According to this:

http://www.sun.com/products-n-soluti...05-3953-10.pdf

the cable length should be 0.8m, not 2m. So that's something you might
try. And AFAIK, there are no tools like you speak of that will
instrument the driver to write syslog unless you write your own driver
to do so. That's certainly an option if you need this sort of
reporting. Look at the OpenSolaris site.

What is "seen" by the system when you do a probe-scsi-all at the OBP?
That will tell you what's on that SCSI bus (aka what's connected to the
HBA).

--
DeeDee, don't press that button! DeeDee! NO! Dee...

Scott J Kolodzieski wrote:
> Sunny wrote:
>
>> That's because UltraSCSI _is_ 40mB/s
>>
>> The UltraSCSI 711 multipack backplane is simply not wired for Ultra 2
>> speeds.
>>
>> http://tinyurl.com/d3ycm
>>
>> Sunny
>>
>
> Thanks Sunny.
>
> Well that answers it clearly. I never would have guessed that all 68
> pins were not wired directly through.
>
> I guess I am on to looking for a chassis that can handle half height
> 3.5" drives and doesn't bankrupt a home user.
>
> Also I would think one could get a second 711 chassis, and split the
> drives between the two, running each at 40mB/s. But that doesn't scale.
>
> Scott.

yes, it's too sad. Mine is also for home use. But the speed limitation
does not seem to be too hard for me. Noise is more likely to be an
issue. But the the 711 is by far more quite than a D1000 or the like.
And the D1000 is limited to 40MB/s, too. OTOH it offers redundant PSUs.

At least the 711 takes hot-swapable SCA disks and costs only a few bugs
on e-bay. All alternatives that I took into considerations had some
drawbacks. But as I only use it for playing with ZFS and as a SVM home
mirror, the performance is not such a big issue.

Thanks to all hints and infos.

Cheers,
Tom

P.S.: the non-exisiting phantom-device disappeared after a reboot and
doing a probe-scsi-all on the OBP.