SGI: Hardware

I'm trying to get a SCSI enclosure taken from an old editing machine to work with my Octane. The enclosure has four Atlas 10K V 72.8 GB hard disks with 68-pin connectors inside, which as far as I can tell are functioning. There is no model brand or number on the box, but it's has two 68-pin connectors on the back. The upper one has a terminator with LVD written on it and a green LED. Also on the back are four switches for setting the SCSI ID number (they are set to 1 2 3 4 at the moment). The enclosure also came with a SCSI cable with two male 68-pin connectors on both sides, which I used to connect the box to the Octane. The problem is when I turn on the Octane I get the following messages on the System Diagnostics (the front light stays red):

timeout on adapter 1 target 8
tm0=0xec31b5ceb, tm1=0xec31b5d01, timeout=0xb
timeout on adapter 1 target 9
tm0=0xec31b5d01, tm1=0xec_something, timeout=0xb
...
timeout on adapter 1 target f
tm0=0xec_something, tm1=0xec_something, timeout=0xb

When it gets to target f the front light turns white, but the messages continue, starting now from target 2 with an additional error message included:

timeout on adapter 1 target 2
tm0=0xec_something, tm1=0xec_something, timeout=0xb
ql_isr - Invalid driver command issued entry_cnt=0x0 entry_type=0x0,
flags=0x4, sys_def_1=0x0
...
and it goes on to f again.

After this I can get in the PROM where hinv doesn't find anything on the SCSI 1 bus. If I don't enter the PROM or I try to start the system from the PROM the machine freezes on

IRIX Release 6.5
...
NOTICE: Initializing Guaranteed Rate I/O v2 (Jul 20 2006, 19:01:52)

The activity LEDs on the front of the box work so it seems that the problem is not in the disks. I tried changing the SCSI IDs but with no luck. I also tried leaving only one disk connected and while I don't get the error messages on startup the system still freezes with the above message and hinv still doesn't recognize the disk. If I remove the terminator the system freezes on System Diagnostics with

timeout on adapter 1 target 3
tm0=0xec_something, tm1=0xec_something, timeout=0xb

I think that the problem may be connected to the termination of the chain, but I don't know what to try next. Any ideas what may be causing the problem or suggestions how to solve it?

Thanks in advance,
Goran

How long is the SCSI cable? The UW SCSI used in the Octane has some fairly severe cable length limitations, as little as 1.5 meters, depending on the number of connected devices. To calculate the overall cable length you'll have to include the distance(s) from the external connector to the hard drives in the drive box.

LVD SCSI allows for much longer cable runs. If your drive box and cabling was intended for use with an LVD controller you might be running into cable length issues.

If the cable length is a problem, the LVD QLogic 12160 is supported <but not bootable> on the Octane - and would have the fringe benefit of offering an increase in SCSI I/O.

I just checked the SCSI cable length and it's 90 cm by itself, but there is around 115 cm inside the enclosure (5 * 23 cm - from the back connector, through the four disks, to the other back connector with the terminator attached) bringing the total to 205 cm. I'll try to find another SCSI cable, hopefully one that's 50 cm long, and check if it makes any difference.

Also, I'm a little reluctant to buy a SCSI card for the Octane without being 100% sure that everything is ok with the enclosure and the disks.

Do you have any suggestions how I can connect only one of the disks and terminate the chain without going through the full length of the enclosure cable (both the terminator and the SCSI connectors on the enclosure cable are male, so I can't connect them directly)?

In any case thank you very much for the help.

Goran

That's understandable - do you have access to another system you could test the drive box and cabling against?

Working blind is always a challenge - could you post a few photos of the enclosure?

If the drives are SCA then you can just run them off your Octane's internal bus. If they're 68-pin you can violate shielding for a short bit and use an internal cable (the ones for LVD should have an integral terminator) and plug it in your Octane's external SCSI port.

Unfortunately I don't have any other machine that has SCSI support, so I can only test the enclosure with the Octane.

The disks are 68-pin not 80-pin SCA, so I can't connect them to the internal controller without an adapter, but I did manage to find a way to connect a single disk to the Octane. I connected the disk directly to the Octane SCSI port using the external SCSI cable, and then provided the power from the enclosure. This way I managed to test all the disks one by one and they are all recognized correctly. One of them had a problem with the SCSI ID jumpers, but after correcting it, the disk worked just fine (at first I thought that I can set the SCSI IDs for the disks using the enclosure, but then I noticed that the corresponding cables were not connected to the disks).

Unfortunately, even after correcting the SCSI ID problem, the disks are not recognized when used with the enclosure. I tried connecting the internal enclosure cable directly to the Octane SCSI port using one of the disk connectors, but I encountered the same problem as before. I suppose that if I use the enclosure cable directly with the Octane I need a terminator on both SCSI connectors on the back of the enclosure.

For now I'll be using just one of the disks connected in the way I mentioned before. Since the Octane SCSI port is limited to 40 MB/s and even a single Atlas disk should easily manage that, AFAIK it won't make a big performance difference if I use one or all four disks.

I've also attached several pictures of the enclosure (the first pic shows one of the disks connected directly to the Octane with the external cable).

Thanks,
Goran

Just a quick update for anyone interested...

I finally managed to solve the problem. It seems that the LVD terminator was the culprit and once I used another terminator everything worked without problems (in fact I used a self-terminating internal SCSI cable, since I didn't have a real terminator, but whatever works...)

Thanks for the help guys!

Cheers!
Goran

I'm glad you figured it out and got it to work! It can really be that ‘simple’ sometimes, so I've experienced also.

The lovely topic of SCSI IDs and termination. The last time that I've spent a fair amount of time messing around with this was with a DEC AlphaServer 1000 system that I used to own (which, by the way, didn't and don't in general come with enough controllers and internal flat-cables and such to accommodate all the physical SCSI channels out-of-the-box, which I only found out after carefully reading the manual). In the end, I luckily nearly always figured it out with SCSI, even though it sometimes required lots of trying around and different SCSI converter/adapter, ID, etc. configurations and setups.

Spoke too soon...

I got file corruption with the above mentioned setup, which as far as I can tell is caused by the total length of the SCSI cables (external cable plus internal enclosure cable plus internal cable used as a terminator). I exchanged the external cable with the internal cable that I used as a terminator (it has three connectors which I had to attach to the Octane port and to BOTH enclosure ports in order for the disks to be recognized) and it seems that the file corruption is gone and everything works as it should.

Are there any risks or possible problems in running the enclosure off an internal SCSI cable type?

Thanks,
Goran