Fatal SCSI error at script address 148 Illegal instruction

Tried connecting a fully populated D2 storage array to a Netra t105

Results in :

ok probe-scsi-all
/pci@1f,0/pci@1,1/scsi@2
Fatal SCSI error at script address 148 Illegal instruction

Would anyone be kind enough to give me some explanation of this error
and/or possible solution?

Regards,

Richard Webb

On 2009-07-24, rcswebb <> wrote:
> Tried connecting a fully populated D2 storage array to a Netra t105
>
> Results in :
>
> ok probe-scsi-all
> /pci@1f,0/pci@1,1/scsi@2
> Fatal SCSI error at script address 148 Illegal instruction
>
> Would anyone be kind enough to give me some explanation of this error
> and/or possible solution?

O.K. Is the D2 a JBOD like the D1000, or is it a hardware RAID
system?

If hardware RAID have the drives already been integrated into
the array's logic?

If they are not integrated, or if the box is a JBOD, I would
start by pulling out all drives, and see what information you get. If
it is clean, add one drive at a time and repeat the test seeing if any
particular drive brings up the error. Always start with the minimum
configuration for a test.

Also -- what is the interface between the D2 and the Netra t105?
I know that the D1000 (which I have) uses a 68-pin HVD SCSI, and as far as I
can tell, the t105 has by default a 68-pin SE SCSI. I have to use a HVD
SCSI card in my Sun Fire 280R to talk to the D1000.

Also, a HVD SCSI terminator is needed on the other end of the
bus in the box. If you want to access all the drives, and use only a
single SCSI controller, you will need a jumper cable to connect the
terminator output from one side to the input connector on the other
side, and that side's output will need a terminator. And I see that
this one needs the VHDCISCSI connectors, not the 68-pin ones. (Other
than the connectors, it looks quite similar to my D1000, except that the
D1000 has both sides in a single plugin, while the D2 has two plugins.

The switches on the plugins define the range of SCSI IDs
assigned to the drives, and if both sides are set the same, you will
have conflicts with two drives answering to each SCSI ID.

O.K. I've finally found something which tells me that it uses
LVD SCSI (not the SE SCSI which is likely what the Netra has), so do you
have a LVD controller to talk to it with, and a LVD terminator for it?
And the choice of RAID or JBOD seems to be based on which card in the
computer it is connected to, not on which card is plugged into the back
of the D2.

And it looks as though the D2 should be powered on before the
computer -- at least for booting from the D2.

If you want to use it as a JBOD, the interface card in the
computer should be the 375-3057 -- except that I don't see the Netra
t105 in the list of systems into which Sun says you can place it.

Have you looked at the LEDs on the D2? This page documents the
meanings of both the box and the drive LEDs:

<http://www.sunshack.org/data/sh/2.1.8/infoserver.central/data/syshbk/General/LEDs/D2_LEDs.html>

and they may be able to give you a hint.

So -- these are the things which I can find which may apply.

Good Luck,
DoN.

--
Email: <> | Voice (all times): (703) 938-4564
(too) near Washington D.C. | http://www.d-and-d.com/dnichols/DoN.html
--- Black Holes are where God is dividing by zero ---

On 25 July, 01:01, "DoN. Nichols" <dnich...@d-and-d.com> wrote:
> On 2009-07-24, rcswebb <rcsw...@gmail.com> wrote:
>
> > Tried connecting a fully populated D2 storage array to a Netra t105
>
> > Results in :
>
> > ok probe-scsi-all
> > /pci@1f,0/pci@1,1/scsi@2
> > Fatal SCSI error *at script address 148 Illegal instruction
>
> > Would anyone be kind enough to give me some explanation of this error
> > and/or possible solution?
>
> * * * * O.K. *Is the D2 a JBOD like the D1000, or is it a hardware RAID
> system?
>
> * * * * If hardware RAID have the drives already been integrated into
> the array's logic?
>
> * * * * If they are not integrated, or if the box is a JBOD, I would
> start by pulling out all drives, and see what information you get. *If
> it is clean, add one drive at a time and repeat the test seeing if any
> particular drive brings up the error. *Always start with the minimum
> configuration for a test.
>
> * * * * Also -- what is the interface between the D2 and the Netra t105?
> I know that the D1000 (which I have) uses a 68-pin HVD SCSI, and as far as I
> can tell, the t105 has by default a 68-pin SE SCSI. *I have to use a HVD
> SCSI card in my Sun Fire 280R to talk to the D1000.
>
> * * * * Also, a HVD SCSI terminator is needed on the other end ofthe
> bus in the box. *If you want to access all the drives, and use only a
> single SCSI controller, you will need a jumper cable to connect the
> terminator output from one side to the input connector on the other
> side, and that side's output will need a terminator. *And I see that
> this one needs the VHDCISCSI connectors, not the 68-pin ones. *(Other
> than the connectors, it looks quite similar to my D1000, except that the
> D1000 has both sides in a single plugin, while the D2 has two plugins.
>
> * * * * The switches on the plugins define the range of SCSI IDs
> assigned to the drives, and if both sides are set the same, you will
> have conflicts with two drives answering to each SCSI ID.
>
> * * * * O.K. *I've finally found something which tells me that it uses
> LVD SCSI (not the SE SCSI which is likely what the Netra has), so do you
> have a LVD controller to talk to it with, and a LVD terminator for it?
> And the choice of RAID or JBOD seems to be based on which card in the
> computer it is connected to, not on which card is plugged into the back
> of the D2.
>
> * * * * And it looks as though the D2 should be powered on beforethe
> computer -- at least for booting from the D2.
>
> * * * * If you want to use it as a JBOD, the interface card in the
> computer should be the 375-3057 -- except that I don't see the Netra
> t105 in the list of systems into which Sun says you can place it.
>
> * * * * Have you looked at the LEDs on the D2? *This page documents the
> meanings of both the box and the drive LEDs:
>
> <http://www.sunshack.org/data/sh/2.1.8/infoserver.central/data/syshbk/...>
>
> and they may be able to give you *a hint.
>
> * * * * So -- these are the things which I can find which may apply.
>
> * * * * Good Luck,
> * * * * * * * * DoN.
>
> --
> *Email: * <dnich...@d-and-d.com> * | Voice (all times): (703) 938-4564
> * * * * (too) near Washington D.C. |http://www.d-and-d.com/dnichols/DoN.html
> * * * * * *--- Black Holes are where God is dividing by zero ---

Hi DoN,

Thanks for your response. I took your advice and booted the array
without just a single disk in position 1. This seemed to recognise the
single disk and controller:

Netra t1 (UltraSPARC-IIi 440MHz), No Keyboard
OpenBoot 3.10.24 ME, 256 MB memory installed, Serial #11700839.
Ethernet address 8:0:20:b2:8a:67, Host ID: 80b28a67.

ok probe-scsi-all
/pci@1f,0/pci@1,1/scsi@2
Target 0
Unit 0 Disk SEAGATE ST318404LSUN18G 4207
Target 1
Unit 0 Disk FUJITSU MAP3367N SUN36G 0401
Target f
Unit 0 Processor SUN D2 3034

My knowledge of SCSI is limited to say the least, but I would have
expected to see the disk appearing as a "child" of the d2, not a
target in its own right. I get the error when I place a disk in
position 0 within the array, which obviously conflicts with the
internal disk. Changing the value of the unit id rotary switch on the
rear of the d2 unit appears to have no affect on the values resulting
from the probe-scsi-all

Any ideas what's going on here? If this is normal behaviour, how would
I alter my setup so that I could use both the internal disk (the
segate) and all 12 slots within the d2.

Thanks,

Richard.

On 2009-07-25, rcswebb <> wrote:
> On 25 July, 01:01, "DoN. Nichols" <dnich...@d-and-d.com> wrote:
>> On 2009-07-24, rcswebb <rcsw...@gmail.com> wrote:
>>
>> > Tried connecting a fully populated D2 storage array to a Netra t105
>>
>> > Results in :
>>
>> > ok probe-scsi-all
>> > /pci@1f,0/pci@1,1/scsi@2
>> > Fatal SCSI error *at script address 148 Illegal instruction

[ ... ]

>> * * * * If they are not integrated, or if the box is a JBOD, I would
>> start by pulling out all drives, and see what information you get. *If
>> it is clean, add one drive at a time and repeat the test seeing if any
>> particular drive brings up the error. *Always start with the minimum
>> configuration for a test.

[ ... ]

> Hi DoN,
>
> Thanks for your response. I took your advice and booted the array
> without just a single disk in position 1. This seemed to recognise the
> single disk and controller:
>
> Netra t1 (UltraSPARC-IIi 440MHz), No Keyboard
> OpenBoot 3.10.24 ME, 256 MB memory installed, Serial #11700839.
> Ethernet address 8:0:20:b2:8a:67, Host ID: 80b28a67.
>
> ok probe-scsi-all
> /pci@1f,0/pci@1,1/scsi@2
> Target 0
> Unit 0 Disk SEAGATE ST318404LSUN18G 4207
> Target 1
> Unit 0 Disk FUJITSU MAP3367N SUN36G 0401
> Target f
> Unit 0 Processor SUN D2 3034

Hmm ... so the D2 shows up as a device on its own -- perhaps for
asking about the status of disks installed. IIRC, this was a 12-slot
unit, so target f is unlikely to have a disk.

But it would appear that your Netra does not have a separate
SCSI bus for the exterior. The Sun Blade 1000 and Sun Blade 2000 do
have separate busses -- FC-AL for internal plus a lot of external as one
bus, an internal SCSI bus for the DVD-ROM drive and a possible optional
tape drive, and a totally separate external fast wide SE SCSI for
external drives.

The 12-slot Multipack boxes have an interesting pattern of SCSI
IDs.

SCSI Slot or behavior
ID
0 Skipped -- assumed internal

1 Skipped -- assumed internal

2 drive slot 0

3 drive slot 1

4 drive slot 2

5 drive slot 3

6 Skipped -- assumed internal CD or DVD drive

7 SCSI Host Adaptor id

8 drive slot 4

9 drive slot 5

10 drive slot 6

11 drive slot 7

12 drive slot 8

13 drive slot 9

14 drive slot 10

15 drive slot 11

IIRC -- the Ultra-2 shared internal and external SCSI buses, so
they needed this pattern. The Ultra-60, however, was the first which I
had to have external SCSI bus separate from the internal.

> My knowledge of SCSI is limited to say the least, but I would have
> expected to see the disk appearing as a "child" of the d2, not a
> target in its own right.

In JBOD (Just a Box Of Disks) mode, which is what you apparently
get without a RAID controller in the computer, it simply causes the
disks to appear on the SCSI bus without interference. Only when it is a
hardware RAID box, such as the A1000 (the same as my D1000, except that
the logic card plugged into it does hardware raid) will it process the
disks into fewer logical disks which are presented to the normal SCSI
controller.

Now -- as I remember from looking at the FEH entries for the D2,
there are DIP switches in the back which can control the address
patterns of each set of 6 disks.

O.K. Looking at the man page for the plug-in modules on the
back (375-3059 540-5016) I see that when DIP switch position 1
(right-most switch) is turned on (up) it runs in single bus mode, while
when it is down it runs in split bus mode. In split bus mode, both
halves run SCSI IDs

There is a manual for it -- but I apparently don't have the permissions
to download it.

https://spe.sun.com/info/control/Doc...6/816-2578/pdf

> I get the error when I place a disk in
> position 0 within the array, which obviously conflicts with the
> internal disk. Changing the value of the unit id rotary switch on the
> rear of the d2 unit appears to have no affect on the values resulting
> from the probe-scsi-all

Hmm ... unit IDs don't seem to make sense for SCSI, as you are
limited to 16 SCSI IDs for wide SCSI (which is what you have here). If
it were a Fibre Channel (FC-AL) box, it would make sense, as it has
the ability to talk to up to 125 or 126 total drives (the 127th being
the host adaptor again.) I'm using some Eurologic JBOD boxes for FC-AL
drives with both my Sun Fire 280R (server) and my Sun Blade 2000
(desktop tower).

But have you power cycled the D2 after changing the switch
settings? I believe that they are read only at power-up, so a change
while it is still powered up would not show up in the behavior until the
next power cycle.

But I think that some of the switches change the IPs of the
drive slots. It never mattered to me, because I had to connect the
D1000 to a separate controller card anyway because it needed HVD, not
LVD (which usually will work as SE SCSI, except for the Exabyte 430 with
Mammoth II tape drives, which will *not* work in SE mode.

> Any ideas what's going on here? If this is normal behaviour, how would
> I alter my setup so that I could use both the internal disk (the
> segate) and all 12 slots within the d2.

Get a PCI LVD SCSI card for the Netra and you can use all 12
slots with no worry about conflicts. You can probably even work with a
SE SCSI card. There are quite a few dual bus cards, so you could run
the D2 in split mode (SCSI IDs 10 through 15 IIRC) with each bus talking
to its own group of six drives -- for a total faster behavior than if
all twelve drives were sharing a single bus.

From my last dead-trees edition of the FEH, the following PCI
bus (I presume that your Netra has PCI bus) cards all offer only SE.

501-2741 68-pin HD SCSI connector, SE

375-0097 68-pin HD SCSI connector, SE -- Symbios SYM8751SPE

375-0005 68-pin VHDCI SCSI dual SE -- Symbios SYM22801 w FCode
375-0013 68-pin VHDCI SCSI dual SE -- Symbios SYM22801 w/o FCode

The only other is a HVD SCSI,

There are PCI-bus LVD SCSI cards for the Suns, but I can't
identify mine at the moment because one is in the Sun Fire 280R server,
and the other in the SB-2000 on which I am typing this. Sun's prtfru
command only prints the system board, the CPU modules, and the DIMMs,
not anything in the PCI slots.

Good Luck,
DoN.

--
Email: <> | Voice (all times): (703) 938-4564
(too) near Washington D.C. | http://www.d-and-d.com/dnichols/DoN.html
--- Black Holes are where God is dividing by zero ---

On 25 July, 22:58, "DoN. Nichols" <dnich...@d-and-d.com> wrote:
> On 2009-07-25, rcswebb <rcsw...@gmail.com> wrote:
>
> > On 25 July, 01:01, "DoN. Nichols" <dnich...@d-and-d.com> wrote:
> >> On 2009-07-24, rcswebb <rcsw...@gmail.com> wrote:
>
> >> > Tried connecting a fully populated D2 storage array to a Netra t105
>
> >> > Results in :
>
> >> > ok probe-scsi-all
> >> > /pci@1f,0/pci@1,1/scsi@2
> >> > Fatal SCSI error *at script address 148 Illegal instruction
>
> * * * * [ ... ]
>
> >> * * * * If they are not integrated, or if the box is a JBOD, Iwould
> >> start by pulling out all drives, and see what information you get. *If
> >> it is clean, add one drive at a time and repeat the test seeing if any
> >> particular drive brings up the error. *Always start with the minimum
> >> configuration for a test.
>
> * * * * [ ... ]
>
>
>
>
>
> > Hi DoN,
>
> > Thanks for your response. I took your advice and booted the array
> > without just a single disk in position 1. This seemed to recognise the
> > single disk and controller:
>
> > Netra t1 (UltraSPARC-IIi 440MHz), No Keyboard
> > OpenBoot 3.10.24 ME, 256 MB memory installed, Serial #11700839.
> > Ethernet address 8:0:20:b2:8a:67, Host ID: 80b28a67.
>
> > ok probe-scsi-all
> > /pci@1f,0/pci@1,1/scsi@2
> > Target 0
> > * Unit 0 * Disk * * SEAGATE ST318404LSUN18G 4207
> > Target 1
> > * Unit 0 * Disk * * FUJITSU MAP3367N SUN36G 0401
> > Target f
> > * Unit 0 * Processor * * SUN * * D2 * * * * * **3034
>
> * * * * Hmm ... so the D2 shows up as a device on its own -- perhaps for
> asking about the status of disks installed. *IIRC, this was a 12-slot
> unit, so target f is unlikely to have a disk.
>
> * * * * But it would appear that your Netra does not have a separate
> SCSI bus for the exterior. *The Sun Blade 1000 and Sun Blade 2000 do
> have separate busses -- FC-AL for internal plus a lot of external as one
> bus, an internal SCSI bus for the DVD-ROM drive and a possible optional
> tape drive, and a totally separate external fast wide SE SCSI for
> external drives.
>
> * * * * The 12-slot Multipack boxes have an interesting pattern of SCSI
> IDs.
>
> SCSI * *Slot or behavior
> ID
> 0 * * * Skipped -- assumed internal
>
> 1 * * * Skipped -- assumed internal
>
> 2 * * * drive slot 0
>
> 3 * * * drive slot 1
>
> 4 * * * drive slot 2
>
> 5 * * * drive slot 3
>
> 6 * * * Skipped -- assumed internal CD or DVD drive
>
> 7 * * * SCSI Host Adaptor id
>
> 8 * * * drive slot 4
>
> 9 * * * drive slot 5
>
> 10 * * *drive slot 6
>
> 11 * * *drive slot 7
>
> 12 * * *drive slot 8
>
> 13 * * *drive slot 9
>
> 14 * * *drive slot 10
>
> 15 * * *drive slot 11
>
> * * * * IIRC -- the Ultra-2 shared internal and external SCSI buses, so
> they needed this pattern. *The Ultra-60, however, was the first which I
> had to have external SCSI bus separate from the internal.
>
> > My knowledge of SCSI is limited to say the least, but I would have
> > expected to see the disk appearing as a "child" of the d2, not a
> > target in its own right.
>
> * * * * In JBOD (Just a Box Of Disks) mode, which is what you apparently
> get without a RAID controller in the computer, it simply causes the
> disks to appear on the SCSI bus without interference. *Only when it is a
> hardware RAID box, such as the A1000 (the same as my D1000, except that
> the logic card plugged into it does hardware raid) will it process the
> disks into fewer logical disks which are presented to the normal SCSI
> controller.
>
> * * * * Now -- as I remember from looking at the FEH entries for the D2,
> there are DIP switches in the back which can control the address
> patterns of each set of 6 disks.
>
> * * * * O.K. *Looking at the man page for the plug-in modules on the
> back (375-3059 540-5016) I see that when DIP switch position 1
> (right-most switch) is turned on (up) it runs in single bus mode, while
> when it is down it runs in split bus mode. *In split bus mode, both
> halves run SCSI IDs
>
> There is a manual for it -- but I apparently don't have the permissions
> to download it.
>
> https://spe.sun.com/info/control/Doc...oc=566/816-257....
>
> > * * * * * * * * * * * * *I get the error whenI place a disk in
> > position 0 within the array, which obviously conflicts with the
> > internal disk. Changing the value of the unit id rotary switch on the
> > rear of the d2 unit appears to have no affect on the values resulting
> > from the probe-scsi-all
>
> * * * * Hmm ... unit IDs don't seem to make sense for SCSI, as you are
> limited to 16 SCSI IDs for wide SCSI (which is what you have here). *If
> it were a Fibre Channel (FC-AL) box, it would make sense, as it has
> the ability to talk to up to 125 or 126 total drives (the 127th being
> the host adaptor again.) *I'm using some Eurologic JBOD boxes for FC-AL
> drives with both my Sun Fire 280R (server) and my Sun Blade 2000
> (desktop tower).
>
> * * * * But have you power cycled the D2 after changing the switch
> settings? *I believe that they are read only at power-up, so a change
> while it is still powered up would not show up in the behavior until the
> next power cycle.
>
> * * * * But I think that some of the switches change the IPs of the
> drive slots. *It never mattered to me, because I had to connect the
> D1000 to a separate controller card anyway because it needed HVD, not
> LVD (which usually will work as SE SCSI, except for the Exabyte 430 with
> Mammoth II tape drives, which will *not* work in SE mode.
>
> > Any ideas what's going on here? If this is normal behaviour, how would
> > I alter my setup so that I could use both the internal disk (the
> > segate) and all 12 slots within the d2.
>
> * * * * Get a PCI LVD SCSI card for the Netra and you can use all12
> slots with no worry about conflicts. *You can probably even work with a
> SE SCSI card. *There are quite a few dual bus cards, so you could run
> the D2 in split mode (SCSI IDs 10 through 15 IIRC) with each bus talking
> to its own group of six drives -- for a total faster behavior than if
> all twelve drives were sharing a single bus.
>
> * * * * From my last dead-trees edition of the FEH, the followingPCI
> bus (I presume that your Netra has PCI bus) cards all offer only SE.
>
> 501-2741 * * * *68-pin HD SCSI connector, SE
>
> 375-0097 * * * *68-pin HD SCSI connector, SE -- Symbios SYM8751SPE
>
> 375-0005 * * * *68-pin VHDCI SCSI dual SE -- Symbios SYM22801 w FCode
> 375-0013 * * * *68-pin VHDCI SCSI dual SE -- Symbios SYM22801 w/oFCode
>
> * * * * The only other is a HVD SCSI,
>
> * * * * There are PCI-bus LVD SCSI cards for the Suns, but I can't
> identify mine at the moment because one is in the Sun Fire 280R server,
> and the other in the SB-2000 on which I am typing this. *Sun's prtfru
> command only prints the system board, the CPU modules, and the DIMMs,
> not anything in the PCI slots.
>
> * * * * Good Luck,
> * * * * * * * * DoN.
>
> --
> *Email: * <dnich...@d-and-d.com> * | Voice (all times): (703) 938-4564
> * * * * (too) near Washington D.C. |http://www.d-and-d.com/dnichols/DoN.html
> * * * * * *--- Black Holes are where God is dividing by zero ---- Hide quoted text -
>
> - Show quoted text -

Thanks Don, for your comprehensive replies.. time to trawl ebay :-)

Rich.