I fixed Windows Delayed Write error....how?

I have a new X2 build. 2-15K Fujitsu drives. The 2nd drive (73G) never 
installed correctly. I got lots of Windows Delayed Write and Device Not 
Found errors. The drive seemed to be failing.

Tonight I replaced it with another 73G drive; which immediately gave me 
problems. The machine wouldn't boot with it attached. Fiddled around a bit 
to no avail; then I tried reversing the SCSI chain, putting the terminator 
right after the new drive rather than after the LSI adapter. That fixed it, 
so I swapped the original "failing" Fujitsu back in. No more errors. Fast as 
could be.

I've set up half a dozen scsi installations before and never came across 
this problem. Sometimes the machine won't boot until I rearrange things, but 
I've never had a machine that ran, but with constant errors.

Why did changing ther termination position fix it? I'd like to find a little 
practical knowledge here, as opposed to my usual fumbling upon the solution 
way of doing things. Thanks...


0
Dennis
8/16/2006 4:44:02 AM
comp.periphs.scsi 2472 articles. 0 followers. l_x2828 (153) is leader. Post Follow

8 Replies
244 Views

Similar Articles

[PageSpeed] 5

Dennis Gordon wrote:
> 
> [...]
> The 2nd drive (73G) never
> installed correctly. I got lots of Windows Delayed Write and Device Not
> Found errors. The drive seemed to be failing.
> [...]
> then I tried reversing the SCSI chain, putting the terminator
> right after the new drive rather than after the LSI adapter. That fixed it,
> so I swapped the original "failing" Fujitsu back in. No more errors. Fast as
> could be.

There must always be two terminators, where is the second one?

[...]
> Why did changing ther termination position fix it?

Most likely because it was in the wrong position before.

The SCSI bus must be a linear sequence of one or multiple cables with a
terminator on each end. The terminators "terminate" any signal that runs
into them - by converting its energy to heat - preventing it from being
reflected at the end of the cable (reflected signals can lead to the
situation that a device detects the same signal multiple times).

This means the terminators must be placed on the two _physical_ ends of
the bus (regardless where the devices are placed between them). Note
that a hostadapter counts as a device like e.g. a disk.

The devices should be uniformly distributed across the bus for best
electrical characteristics and should have a distance of 30cm from
each-other if possible. A minimum distance between the terminators and
devices is not required (therefore it is possible to integrate
terminators into devices).

For your configuration this might look like this:

     +---------+   +-------+   +------T
     |          \ /         \ /
  +--O--+     +--O--+     +--O--+
  |  T  |     |     |     |     |
  |     |     |     |     |     |
  +-----+     +-----+     +-----+
Hostadapter    Disk1       Disk2

(T: Terminator)


Micha
0
Michael
8/16/2006 10:23:13 AM
"Michael Baeuerle" <michael.baeuerle@stz-e.de> wrote in message news:k98br3-j6t.ln1@micha.freeshell.org
> Dennis Gordon wrote:
> > 
> > [...]
> > The 2nd drive (73G) never
> > installed correctly. I got lots of Windows Delayed Write and Device Not
> > Found errors. The drive seemed to be failing.
> > [...]
> > then I tried reversing the SCSI chain, putting the terminator
> > right after the new drive rather than after the LSI adapter. That fixed it,
> > so I swapped the original "failing" Fujitsu back in. No more errors. Fast as
> > could be.
> 
> There must always be two terminators, where is the second one?

On the scsi controller card, maybe? Just a lucky guess.

> 
> [...]
> > Why did changing ther termination position fix it?
> 
> Most likely because it was in the wrong position before.
> 
> The SCSI bus must be a linear sequence of one or multiple cables with a
> terminator on each end. 

> The terminators "terminate" any signal that runs into them - 

> by converting its energy to heat - 

Never heard it explained like that. 
Pity reflections are frequency/impedance dependent and not energy related.

> preventing it from being reflected at the end of the cable 

The popular explanation is that a terminator mimics the impedance of the 
cable, making it appear as if the cable is endless and so no reflections
do appear. It also functions to supply the signalling voltage to the bus.

> (reflected signals can lead to the situation that 
> a device detects the same signal multiple times).
> 
> This means the terminators must be placed on the two _physical_ ends 
> of the bus (regardless where the devices are placed between them). 
> Note that a hostadapter counts as a device like e.g. a disk.

> The devices should be uniformly distributed across the bus 

Nope. 
Pity that is usually completely impossible with both internal and external 
devices on the bus.

> for best electrical characteristics 

> and should have a distance of 30cm from each-other if possible. 

Actually the distance is dependent on the stub length of the devices
(conductor length from cable connector to device controller chip).
The 30cm is based on the max allowed stublength.
The (minimum) requirements are actually different for SE and LVD.

> A minimum distance between the terminators and devices is not required 

Yes it is. 
It's either within a short distance (stub) or the minimum spacing distance.

> (therefore it is possible to integrate terminators into devices).
> 
> For your configuration this might look like this:
> 
>      +---------+   +-------+   +------T
>      |          \ /         \ /
>   +--O--+     +--O--+     +--O--+
>   |  T  |     |     |     |     |
>   |     |     |     |     |     |
>   +-----+     +-----+     +-----+
> Hostadapter    Disk1       Disk2
> 
> (T: Terminator)
> 
> 
> Micha

Are you trying to become the resident babblemouth, Micha.
0
Folkert
8/16/2006 5:21:44 PM
On Wed, 2006-08-16 at 19:21 +0200, Folkert Rienstra wrote:
> "Michael Baeuerle" <michael.baeuerle@stz-e.de> wrote in message news:k98br3-j6t.ln1@micha.freeshell.org

[snip]

> Are you trying to become the resident babblemouth, Micha.

Are you afraid he's after your job, Folkert?

-- 
Jeremy Boden


0
Jeremy
8/16/2006 9:13:35 PM
Dennis Gordon wrote:
> I have a new X2 build. 2-15K Fujitsu drives. The 2nd drive (73G) never 
> installed correctly. I got lots of Windows Delayed Write and Device Not 
> Found errors. The drive seemed to be failing.
> 
> Tonight I replaced it with another 73G drive; which immediately gave me 
> problems. The machine wouldn't boot with it attached. Fiddled around a bit 
> to no avail; then I tried reversing the SCSI chain, putting the terminator 
> right after the new drive rather than after the LSI adapter. That fixed it, 
> so I swapped the original "failing" Fujitsu back in. No more errors. Fast as 
> could be.

Hi

You seem to be suggesting you had the terminator in between the adapter
(LSI) and the first or second disk.  You should have the terminator
right at the end of the cable (last connector of the cable).

If it helps a little

Regards
Charles

> 
> I've set up half a dozen scsi installations before and never came across 
> this problem. Sometimes the machine won't boot until I rearrange things, but 
> I've never had a machine that ran, but with constant errors.
> 
> Why did changing ther termination position fix it? I'd like to find a little 
> practical knowledge here, as opposed to my usual fumbling upon the solution 
> way of doing things. Thanks...
> 
> 


-- 
Please remove _removeme_ to reply.
0
Charles
8/17/2006 12:24:03 AM
Folkert Rienstra wrote:
> 
> Michael Baeuerle wrote:
> >
> > [...]
> > The terminators "terminate" any signal that runs into them -
> > by converting its energy to heat -
> 
> Never heard it explained like that.
> Pity reflections are frequency/impedance dependent and not energy related.

The physical principle of energy conservation say that energy cannot
disappear. If the signal hits the unterminated end of the cable it must
be reflected otherwise this principle would be violated. This is the
physical cause for reflections.

How much of the energy is reflected depends on the impedance mismatch.
This means if your terminator has not the same impedance as the cable it
cannot convert all of the signals energy (the rest is reflected). This
is also the case if you place a correctly matching terminator in the
middle of a long cable: A signal reaching the terminator will see the
terminators impedance and the cable impedance (from the rest of the
cable) in parallel. This is an impedance mismatch creating a reflection
(of nominal 1/3 of the energy).

> > preventing it from being reflected at the end of the cable
> 
> The popular explanation is that a terminator mimics the impedance of the
> cable, making it appear as if the cable is endless and so no reflections
> do appear.

I know. But IMHO this do not explain what terminators _physically_ do -
they have to convert the signal energy, otherwise they have no effect.
The "endless cable" is nothing else than a virtual energy storage
facility with infinite size.

> It also functions to supply the signalling voltage to the bus.

Yes, but this job is always done if at least one terminator is present
(even if it is at the wrong position).

> > [...]
> > The devices should be uniformly distributed across the bus
> 
> Nope. 
> Pity that is usually completely impossible with both internal and external 
> devices on the bus.

I have written "should". It is the ideal case because every device
creates some impedance distortion that should not be lumped (instead it
should be distributed as best as possible).

> > and should have a distance of 30cm from each-other if possible.
> 
> Actually the distance is dependent on the stub length of the devices
> (conductor length from cable connector to device controller chip).

... and the input capacitance of the chip.

> The 30cm is based on the max allowed stublength.
> The (minimum) requirements are actually different for SE and LVD.

Yes, but stub length and capacitance are unknown in nearly all cases. So
one should try to use 30cm if possible to be on the save side.

> [...]
> Are you trying to become the resident babblemouth, Micha.

Maybe :-) But I can be more quiet if the people here have a problem with
it and think my postings contain no useful information ...


Micha
0
Michael
8/17/2006 9:23:20 AM
"Michael Baeuerle" <michael.baeuerle@stz-e.de> wrote in message news:b5pdr3-v3v.ln1@micha.freeshell.org
> Folkert Rienstra wrote:
> > Michael Baeuerle wrote:
> > > 
> > > [...]
> > > The terminators "terminate" any signal that runs into them -
> > > by converting its energy to heat -
> > 
> > Never heard it explained like that.
> > Pity reflections are frequency/impedance dependent and not energy related.

[snip]
 
> > > preventing it from being reflected at the end of the cable
> > 
> > The popular explanation is that a terminator mimics the impedance of the
> > cable, making it appear as if the cable is endless and so no reflections
> > do appear.
 
> I know. 

Then why not keep it at that. 
Why explain it differently and stirr up controversy.

> But IMHO this do not explain what terminators _physically_ do -

Who cares. They avoid reflections occurring.

> they have to convert the signal energy, otherwise they have no effect.

> The "endless cable" is nothing else than a virtual energy storage
> facility with infinite size.

Which is good enough for me and far more appealing.

> 
> > It also functions to supply the signalling voltage to the bus.
> 
> Yes, but this job is always done if at least one terminator is present
> (even if it is at the wrong position).
> 
> > > [...]
> > > The devices should be uniformly distributed across the bus
> > 
> > Nope.
> > Pity that is usually completely impossible with both internal and external
> > devices on the bus.
> 
> I have written "should". It is the ideal case because every device
> creates some impedance distortion that should not be lumped (instead it
> should be distributed as best as possible).
> 
> > > and should have a distance of 30cm from each-other if possible.

> > Actually the distance is dependent on the stub length of the devices
> > (conductor length from cable connector to device controller chip).

> .. and the input capacitance of the chip.

Nope, not in the SE specs. Capacitance plays a role in the LVD specs.
With SE it is just spacing distance to stublength is 5:1
With a 6cm maximum for stub length that makes 30cm a safe distance.

> 
> > The 30cm is based on the max allowed stublength.
> > The (minimum) requirements are actually different for SE and LVD.
> 
> Yes, but stub length and capacitance are unknown in nearly all cases. 
> So one should try to use 30cm if possible to be on the save side.
> 
> > [...]
> > Are you trying to become the resident babblemouth, Micha.
> 
> Maybe :-) 
> But I can be more quiet if the people here have a problem with
> it and think my postings contain 

> no useful information ...

Depends on the usefulness and correctness.

> 
> 
> Micha
0
Folkert
8/17/2006 11:17:43 PM
Folkert Rienstra wrote:
> "Michael Baeuerle" <michael.baeuerle@stz-e.de> wrote in message news:b5pdr3-v3v.ln1@micha.freeshell.org

<heavy snip>

> 
>> But IMHO this do not explain what terminators _physically_ do -
> 
> Who cares. They avoid reflections occurring.

...^^^^^^^^^

<snip>

>>> [...]
>>> Are you trying to become the resident babblemouth, Micha.
>> Maybe :-) 
>> But I can be more quiet if the people here have a problem with
>> it and think my postings contain 
> 
>> no useful information ...
> 
> Depends on the usefulness and correctness.

as you can see you too may choose to avoid correctness.  Usefulness is
important even if it is not perfectly correctly explained.  You could
make yourself more useful by giving useful answers as opposed to picking
on those that are not always correct but may still be useful.

Regards
Charles

-- 
Please remove _removeme_ to reply.
0
Charles
8/18/2006 12:55:25 AM
Thanks for the explanations and...ummm entertiaing banter. As I checked the 
other machines I've set up I see that I'd done it correctly on the other 
three, but it was a long time ago and I just didn't think things through 
this time. What confused me was, in the past, when I had impropery cabled 
the drives, one of them simply wouldn't show up. I'd not seen the delayed 
write error before, so it took a day for my small brain to catch on.


"Charles C." <c.k.christacopoulos.removeme.@dundee.ac.uk> wrote in message 
news:44e3b700$0$1390$db0fefd9@news.zen.co.uk...
> Dennis Gordon wrote:
>> I have a new X2 build. 2-15K Fujitsu drives. The 2nd drive (73G) never
>> installed correctly. I got lots of Windows Delayed Write and Device Not
>> Found errors. The drive seemed to be failing.
>>
>> Tonight I replaced it with another 73G drive; which immediately gave me
>> problems. The machine wouldn't boot with it attached. Fiddled around a 
>> bit
>> to no avail; then I tried reversing the SCSI chain, putting the 
>> terminator
>> right after the new drive rather than after the LSI adapter. That fixed 
>> it,
>> so I swapped the original "failing" Fujitsu back in. No more errors. Fast 
>> as
>> could be.
>
> Hi
>
> You seem to be suggesting you had the terminator in between the adapter
> (LSI) and the first or second disk.  You should have the terminator
> right at the end of the cable (last connector of the cable).
>
> If it helps a little
>
> Regards
> Charles
>
>>
>> I've set up half a dozen scsi installations before and never came across
>> this problem. Sometimes the machine won't boot until I rearrange things, 
>> but
>> I've never had a machine that ran, but with constant errors.
>>
>> Why did changing ther termination position fix it? I'd like to find a 
>> little
>> practical knowledge here, as opposed to my usual fumbling upon the 
>> solution
>> way of doing things. Thanks...
>>
>>
>
>
> -- 
> Please remove _removeme_ to reply. 


0
Dennis
8/21/2006 3:09:20 AM
Reply: