Hanging when trying to detect Quantum LTO 3 HH?

[apinsley]

We recently purchased a Quantum Ultrium 3 LTO HH drive, as it was on the qualified hardware list. After swapping out the Quantum (Benchmanrk) DLT IV drive we had been using, OS X 10.4.11 detects the new drive properly, but Retrospect hangs on the "scanning" screen when I click on Configure->Devices.

 

Current Setup:

G5 Dual 1.8ghz

1GB RAM

150GB internal + 1TB Firewire Drive

Adaptec APD 29160 MAC*

Quantum LTO 3 HH

 

Retrospect 6.1.230

Device Access Version 1.0.107

Driver Update Version 6.1.15.100

 

*Now, I'm pretty sure this is a problem, but it worked with the DLT IV Drive. If necessary, I'd replace it with an ATTO UL4S.

 

 

Edited July 17, 2008 by Guest

[rhwalker]

OS X 10.4.11 detects the new drive properly, but Retrospect hangs on the "scanning" screen when I click on Configure->Devices.

 

...

 

Adaptec APD 29160 MAC*

 

*Now, I'm pretty sure this is a problem, but it worked with the DLT IV Drive. If necessary, I'd replace it with an ATTO UL4S.

Yep, that's the problem. Adaptec stopped supporting the Mac platform as of MacOS 10.3.9 and did not develop any drivers for 10.4.x (or later).

 

Adaptec used to have a notice on their web site announcing this, but the link is now dead:

Adaptec abandons Mac platform

 

Get the UL4S. We have a UL4D in our Xserve G5, and it works fine.

 

Russ

[apinsley]

Is there any benefit for the UL4D over the S?

 

It'll only be driving the single device, so I'd think the second channel would be superfluous.

[rhwalker]

Is there any benefit for the UL4D over the S?

 

It'll only be driving the single device, so I'd think the second channel would be superfluous.

Well, just to give you my experience:

 

There seems to be some sort of an odd bug with Retrospect that has existed for several years. We reported it to Retrospect support under our support contract back in 2006, and, together with Retrospect support, came up with a workaround, but I believe that the problem still exists and requires multiple SCSI channels to solve.

 

Here is the bug, which is timing dependent, hard to reproduce, but it will bite you.

 

Apparently Retrospect does something during its device scan prior to backup that causes a SCSI bus transaction to be dropped on an ATTO UL4D/UL4S; perhaps it's a bus reset, who knows. We had a SCSI RAID 1 mirror drive on the same SCSI channel as our Exabyte VXA-2 1x10 1u (SCSI) tape drive with autoloader that is our backup device. When Retrospect autolaunched to do the nightly backup, occasionally it would cause the SCSI RAID 1 mirror drive to hang on a pending I/O transfer. That drive, being a mirror of our OS boot drive, caused the system to hang (although you could still poke about if you already had an SSH session open). Apparently the system was waiting on an I/O transfer completion on the SCSI disk drive that never completed.

 

The only workaround that reliably worked was to put the tape drive (Retrospect's backup device) on its own SCSI channel, which we were able to do because we had two channels with the UL4D. Once we isolated the tape drive on its own SCSI channel, the problem has not reoccurred - before, it would happen about once a week.

 

If you are able to put all of your tape backup devices on the same SCSI channel without any disks on that SCSI channel, you won't run into this issue.

 

Russ

[apinsley]

thanks!

 

I just ordered the UL4S.

[jamiemac]

Please could you tell me which slot you have your ATTO plugged into. I have and old G5 and i'm only getting around 25 mb/sec instead of the quoted 68mb/s for the LTO3-HH using the ATTO ExpressPCI UL4S.

 

Many thanks