DoN. Nichols <dnichols@d-and-d.com> wrote:

> Interesting -- since the label says (in the finer print to the
> right of the bold-print ST3145807FCV:

> VDC Amps
> +5 -
> +12 1.4


Well, if that is what is on the label, that is what it must be and I'm
wrong.

Only suggestion at this point is download and examine this...

http://www.seagate.com/support/disc/...100195490b.pdf

I don't know how you get to those anymore from seagates site but I found the
old "product manual" when we ran into this, adjusted things accordingly and
googled enough to track down the specific one.

It has everything you want to more about the drive than you probably care to
know, including the connector pinout, misc jumpers and general theory of
operation.

Again, it makes no mention of a 12V only option and the half dozen pages
explaining the power requirements are always showing a 12/5 source. Even has
graphs showing current measurements of both feed lines as the drive is going
through various stages of operations.

Best of luck finding a solution but I'm still betting on it being defective
somehow.

-bruce
bje@ripco.com

DoN. Nichols wrote:
> On 2008-04-20, Bruce Esquibel <bje@e4500.ripco.com> wrote:
>
>>DoN. Nichols <dnichols@d-and-d.com> wrote:
>>
>>
>>> When withdrawn, the drives do not give the typical inertial
>>>behavior of a spinning drive.
>>

Have had a few issues with fc drives, using a couple of the older emc
10 drive boxes with s/h none sun seagate 10k drives. Each box of 10 is
configured under solaris 8 / disksuite as a single raid array, with the
second box powered up only for rsync between the first and second.
Controller is a dual channel sun (FC100 iirc) sbus card, host is Ultra
II. The good thing about the emc boxes is the lower power consumption
compared with stuff like a5100, which draws 600 or so watts with 12
drives fitted.

Part of the problem may be due to the fact that the whole setup is
unsupported and a real dog's dinner, but one thing for sure is that none
of the drives were recognised without a boot -r, when the boot process
probes for all peripherals it can find. After that, it's just a case of
using format to partition, format and bad block scan the drives.
Overall, nice cheap storage, but f/c drives don't seem as
straightforward as scsi...

Chris

On 2008-04-21, Bruce Esquibel <bje@e4500.ripco.com> wrote:
> DoN. Nichols <dnichols@d-and-d.com> wrote:

[ ... ]

> Only suggestion at this point is download and examine this...
>
> http://www.seagate.com/support/disc/...100195490b.pdf
>
> I don't know how you get to those anymore from seagates site but I found the
> old "product manual" when we ran into this, adjusted things accordingly and
> googled enough to track down the specific one.

Thanks! It is printing as I type.

> It has everything you want to more about the drive than you probably care to
> know, including the connector pinout, misc jumpers and general theory of
> operation.
>
> Again, it makes no mention of a 12V only option and the half dozen pages
> explaining the power requirements are always showing a 12/5 source. Even has
> graphs showing current measurements of both feed lines as the drive is going
> through various stages of operations.
>
> Best of luck finding a solution but I'm still betting on it being defective
> somehow.

Well ... I did kick my wife off the server for a little while
today and replaced both 146 GB drives in the hot-swap housing with
these, and probe-scsi did show them.

So -- back to the normal pair of drives and reboot, then copy
everything off the second drive to a zfs pool (four filesystems) and
re-export them from there -- and change the client mount information so
I could do without that second drive for a while.

Now it sees it (after the usual "devfsadm -C -c disk" (we seem
to need the "-C" to clean up the old WWN linkages in the /devices tree),
and I am currently formatting it. 30000 tracks into the first verify
pass.

Once that is done, I will try again to get it recognized in the
EMC housing. It would be nicer if I could trust it to mount there as
well as in the internal cage.

Thanks much,
DoN.

--
Email: <dnichols@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 ---

On 2008-04-21, ChrisQuayle <nospam@devnul.co.uk> wrote:
> DoN. Nichols wrote:
>> On 2008-04-20, Bruce Esquibel <bje@e4500.ripco.com> wrote:
>>
>>>DoN. Nichols <dnichols@d-and-d.com> wrote:
>>>
>>>
>>>> When withdrawn, the drives do not give the typical inertial
>>>>behavior of a spinning drive.
>>>
>
> Have had a few issues with fc drives, using a couple of the older emc
> 10 drive boxes with s/h none sun seagate 10k drives. Each box of 10 is
> configured under solaris 8 / disksuite as a single raid array, with the
> second box powered up only for rsync between the first and second.
> Controller is a dual channel sun (FC100 iirc) sbus card, host is Ultra
> II. The good thing about the emc boxes is the lower power consumption
> compared with stuff like a5100, which draws 600 or so watts with 12
> drives fitted.

O.K. I've managed to get the drives in question working in the
internal drive cage in the SF 280R, and after I finish the full format
of the first of them, I'll see whether it moves to the EMC housing this
time -- though it did not before. I'll also test the non "12V ONLY"
drive of the same size in there and see what it does.

Do you happen to know whether the EMC box needs the loopback
"terminator" on the "out" connector? I've got one (home-made) there and
it is working fine with the original drives at least. I'm reluctant to
pull the terminator because that also potentially shuts down my internal
drives thus shutting the whole system down. :-)

> Part of the problem may be due to the fact that the whole setup is
> unsupported and a real dog's dinner, but one thing for sure is that none
> of the drives were recognised without a boot -r, when the boot process
> probes for all peripherals it can find. After that, it's just a case of
> using format to partition, format and bad block scan the drives.

Hmm ... for me, with the original drives (seven 36 GB FC drives)
all I needed to do was "devfsadm -c disk" to get it to see the "new"
drives in the array. When I went to test the ten 9 GB FC drives out of
the second box, three at a time, I had to use "devfsadm -C -c disk" to
clear the old drives out of the /devices/ tree.

> Overall, nice cheap storage, but f/c drives don't seem as
> straightforward as scsi...

Perhaps in part because of the unique "WWN" (World Wide Number)
in each drive -- which gets encoded into the /devices tree. Here is an
example from the first drive in my system:

/devices/pci@8,600000/SUNW,qlc@4/fp@0,0/ssd@w2100000c50a8710d,0:a

The part between the '@w' and the ",0:a" at the end is the WWN, and I
think that having multiple ones tied to the same drive path could
confuse the system. The "-C" option to devfsadm tells it to clear any
unused ones. I use the "-d disk" option simply to avoid the risk that
it could confuse the sgen entry for the tape jukebox, and the entries
for the tape drives.

Thanks,
DoN.

--
Email: <dnichols@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 ---

On 2008-04-22, DoN. Nichols <dnichols@d-and-d.com> wrote:
> On 2008-04-21, ChrisQuayle <nospam@devnul.co.uk> wrote:
>> DoN. Nichols wrote:
>>> On 2008-04-20, Bruce Esquibel <bje@e4500.ripco.com> wrote:
>>>
>>>>DoN. Nichols <dnichols@d-and-d.com> wrote:

[ ... ]

> O.K. I've managed to get the drives in question working in the
> internal drive cage in the SF 280R, and after I finish the full format
> of the first of them, I'll see whether it moves to the EMC housing this
> time -- though it did not before. I'll also test the non "12V ONLY"
> drive of the same size in there and see what it does.

The plain vanilla drive does not work in the EMC housing either.
I suspect that the EMC is 1 GHz only in the FC, and is not setting the
pins to tell the drive to back down to 1 GHz from 2 GHz. I've only used
the EMC housing with the 36 GB and 9 GB drives which came in the
housings.

Thanks all,
DoN.

--
Email: <dnichols@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 ---

DoN. Nichols wrote:

>
>> O.K. I've managed to get the drives in question working in the
>>internal drive cage in the SF 280R, and after I finish the full format
>>of the first of them, I'll see whether it moves to the EMC housing this
>>time -- though it did not before. I'll also test the non "12V ONLY"
>>drive of the same size in there and see what it does.
>
>
> The plain vanilla drive does not work in the EMC housing either.
> I suspect that the EMC is 1 GHz only in the FC, and is not setting the
> pins to tell the drive to back down to 1 GHz from 2 GHz. I've only used
> the EMC housing with the 36 GB and 9 GB drives which came in the
> housings.
>
> Thanks all,
> DoN.
>

Don,

Thanks for the tip about replacing the drives hot. It probably is the
WWN number change that's confusing the system - just never figured out
which utils to use to fix it. More frequent rtfm I guess :-). One thing
to watch with the emc stuff is that some of the drives are formatted for
a non standard sector size - iirc 578 bytes rather than 512. If you
download the fc utils from the seagate web site, and use an Emulex
lp8000 or similar pci card in a pc running the seagate utils, there is
an option to change the sector size. Have never tried this, as all my
drives were already at 512 bytes, but the utility will tell you what the
sector sizes are and iirc, bad block the drives at low level, prior to
sun format.

As for the terminator, don't konw, but not using one here. The sun sbus
controller has removable transceivers. One array has a 9 pin D
transceiver, while the other uses fibre and a fibre to 9 pin mia adapter
at the array end. Both seem to work flawlessly. The drives are a mix of
early f/h st173404fc and the same type but later h/h drives, iirc
....405fc...

Chris

On 2008-04-23, ChrisQuayle <nospam@devnul.co.uk> wrote:
> DoN. Nichols wrote:
>
>>
>>> O.K. I've managed to get the drives in question working in the
>>>internal drive cage in the SF 280R, and after I finish the full format
>>>of the first of them, I'll see whether it moves to the EMC housing this
>>>time -- though it did not before. I'll also test the non "12V ONLY"
>>>drive of the same size in there and see what it does.
>>
>>
>> The plain vanilla drive does not work in the EMC housing either.
>> I suspect that the EMC is 1 GHz only in the FC, and is not setting the
>> pins to tell the drive to back down to 1 GHz from 2 GHz. I've only used
>> the EMC housing with the 36 GB and 9 GB drives which came in the
>> housings.

[ ... ]

> Thanks for the tip about replacing the drives hot. It probably is the
> WWN number change that's confusing the system - just never figured out
> which utils to use to fix it. More frequent rtfm I guess :-).

In Solaris 10, at least, use:

devfsadm -C -c disk

If you don't have weird tape drives and jukebox on the SCSI or FC bus,
you don't need the "-c disk" (limit the work to class "disk"). The "-C"
says to clean up links and /devices entries which are no longer in use.

> One thing
> to watch with the emc stuff is that some of the drives are formatted for
> a non standard sector size - iirc 578 bytes rather than 512.

I think that those were ones on a SGI system. :-)

> If you
> download the fc utils from the seagate web site, and use an Emulex
> lp8000 or similar pci card in a pc running the seagate utils, there is
> an option to change the sector size.

That first requires me to get the lp8000 or similar, and move the
token Windows machine out to where it can access the array. :-)

> Have never tried this, as all my
> drives were already at 512 bytes, but the utility will tell you what the
> sector sizes are and iirc, bad block the drives at low level, prior to
> sun format.

Hmm ... another possibility would be to use "sformat" from
schily -- the same fellow who wrote the "cdrecord" program which is
supplied with Solaris 10 in the "Software Companion" DVD-ROM. But
you'll have to find and download the source for that one yourself. And,
you'll have to download and compile the source for his "smake" first.
He has his own version of make, and writes his other sources to expect
it.

So -- that makes "make", "gmake" and "smake" so far, but at
least gmake is in the /usr/sfw/bin directory, so it is already installed
with Solaris 10 anyway -- even if you skip the Software Companion stuff.

> As for the terminator, don't konw, but not using one here.

That means that it is self terminating. That is good to hear
about. I didn't need to make my own terminator.

> The sun sbus
> controller has removable transceivers. One array has a 9 pin D
> transceiver, while the other uses fibre and a fibre to 9 pin mia adapter
> at the array end.

O.K. I'm using the FC interface which is built into the Sun
Fire 280R and the Sun Blade [12]000 systems. That uses a special flat
connector which is like a wider and longer RJ-45 (but black not clear)
at the back of the computer. To use with the EMC system, I have to have
a cable to go from that to the DA-9 connector.

As for the terminator -- it really is just a loopback, and I
made my own (pin 1 to pin 5, and pin 6 to pin 9) rather than depending
on the EMC to be self terminating, since otherwise I would interfere
with access to the two internal drives in the SF-280R, one of which is
the boot drive and thus a bad thing to lose on a server. :-)

> Both seem to work flawlessly. The drives are a mix of
> early f/h st173404fc and the same type but later h/h drives, iirc
> ...405fc...

O.K The drives which came with the EMC array which are still in
use are 36 GB 1.6" high drives marked as:

Vendor: IBM Product: DRHL36L CLAR36

from the output from "iostat -E -n". Also, the 9GB drives (also 1.6"
high) which are labeled ST19171FC by Seagate (though I don't know what
the IOSTAT will call them).

Also it works with a ST336605FSUN36G (pulled from the SB-1000
when I put in one of the 146 GB FCV 12V only drives).

And 1" high ST39102FC (9GB) drives also work well in the EMC
housing.

The 146 GB drives ST3146807FC or SX3146807FC (Seagate drives
which have passed a certain number of service hours magically change the
second character of their identification from 'T' to 'X', and at least
IBM AIX systems bitch like made about them every time they boot claiming
that the drive is ready to fail any minute. However Solaris 10 simply
reports the "SX" in place of the "ST" and still happily used the drive.
(Yes, both of these "12V ONLY" drives had been run to the "SX" point.
But I now have one in each of the SB-1000 and the SF-280R.

I've got a friend and neighbor who has a spare 171 GB 1.6" FC
drive which I will try in the EMC soon -- just to try to figure out what
is worth getting from eBay and what I should skip over. The 1.6" drives
certainly won't work as internal drives in the SF-280R, because it has a
smaller card cage than the SB-[12]000.

Anyway -- whatever it is, the problem is somewhere in the EMC,
since the 146 GB drives work nicely in the SF and SB drive cages.

Obviously, if I could find an affordable Multipack-FC I could
grab a full set of the 146 GB ones and be quite happy.

Thanks and Good Luck,
DoN.

--
Email: <dnichols@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 ---