3 x LTO-1 Ultrium 230 drives (all reporting same error)?

[Rick Kirk]

Hi all,

I've recently acquired from work some decomissioned Ultrium 230 drives. 1 from an autoloader and 2 from a 1/20 library. I'm intending to use them under windows. I have installed an adaptec 2940u2w ultra2 scsi and a foxconn Ultra160 cable (should be compatible with Ultra 2 speeds 80M/s).

All is well and I can see them in LTT. This is what is interesting. I have used the Drive assesment tests on each drive in turn and it fails on all of them. Looking at the LTT logs tells me that Servo performance is around 4-6% and read/write performance is around 50%.

The drives vary in age from 2 years and up, some have had many loads (9000+) others much less.

What I find odd here is that all 3 report similar results (within 2-3%) for the assesment test. Drive health reports no issues with any however I notice the negotiated burst transfer rate is 40 (as apposed to the 80 the card and drive are capable of).

I'm doing all this on Windows XP.

I have also tried other SCSI cabels (though I only have the one Ultra 160), others are SE differential terminators). I have also tried two SCSI cards (adaptec and LSI). Nothing seems to make any difference.

Given the rated life of the heads etc on these drives I find it hard to believe that I could be so unlucky as to get 3 drives with such bad performance. I've been pulling my hair out for 4 weeks testing these now. I would be open to any suggestions anyone can recommend.

Also, I have tried cleaning the drives (with cleaning tape), and even cleaned one manually with cotton bud and isopropyl). I haev tried various data carts from both imation and HP..

Each has been flashed to latest firmware level however I have noticed that the Ultrium 230 firmare from the site is the "wrong identity" and I have had to use E3BR for these drives. I am curious if this is because they are library drives instead of standalone drives... Each has been plugged in as standalone in the PC and is the only device on the chain. Is there any difference in a library 230 as apposed to a standalone? and if so can a library drive be flashed with standalone firware via LTT with a service password?

I'm at my wits end. Could they really all be dead?

Sorry for the big post.

Thanks for any help and suggestions!

Regards,

Rick.

[Vincent Farrugia]

Hello,

That is your problem, you cannot use library drives as standalones. As you said, they have a different identity, and you cannot flash them with standalone firmware.

Once I tried this and had the same exact problem as you.

If you have drives from a 1/20, how did you manage to convert them to standalones, may I ask? If I got you correctly, these autoloader/libraries are from the pre-merge HP series, not the modern MSL series.

Kind Regards,
Vince

[Rick Kirk]

Vince,

Thanks for the response. The 1/20 was a typo. I actually meant a 2/20. It's a surestore library so no, not from the MSL series.

All I've done to convert them is pull them out of the cassette system they are in and direct attach 1 at a time to the internal SCSI chain in the PC. They work and LTT see's the drives with no issues... It's just write and servo performance where I get poor performance.

I can successfully backup and restore with them however I can't put any faith in them given the report of L&TT. I'll try running the same tests with the drives in the library and see what happens.

Can you explain why they would work but in a reduced capacity installed individually as apposed to in a library? I can't make sense of how that could affect it. Looking at the internal smarts of the library it appears the control board on the cassette is only used for changer->drive communication. Servo performace would related to the ability of the drive to key to the servo track on the tape, surely that would not change as a result of being plugged into a library vs a standard SCSI chain?

Thanks for your help!

Cheers,

Rick.

[Rick Kirk]

Spent last night cabling and running L&TT directly on the library. Seems my worst fears have been realised. Both drives report exactly the same problems as they did hooked up as standalone. It seems they are just worn out.

If someone from the L&TT team reads this I have attached the log from last nights activities if you could be so kind as to take a look?

The library is out of warrantee and service so I would be interested to know what could be done to have the drives swapped or repaired and the likely cost for such activities.

Thanks for the help.

Regards,
Rick.

[Richard Bickers]

Some drives will work outside of the library environment and some won't - it depends how specific the drive firmware is. You may well be OK with autoloader units but likely not from the bigger libraries. Looks like you're OK in this case - functionally.

Note that you'll only be able to upgrade with the same firmware variant - though any drive fixes are applied equally to all variants. It's worth updating your FW if you can.

I took a look at the log and the drive has failed the assessment test twice. In this case it's a write/read failure with a loss of capacity between 25 and 50%. This is very high so not good. We consider 20+% to be failures - though you may be able to recover the situation with a cleaning tape. It could just be that the heads are clogged.

A loss of capacity doesn't mean it doesn't work - these drives have lots of margin - but you're on the edge, your performance will be reduced and I doubt you want to trust your data to it.

My suggestion:
1) Update the firmware (ah, just noticed tht you did...)
2) Use a cleaning tape
3) Try the assessment test again
4) Pull a support ticket and take a look through (detail level set to more details). This should tell you something about your drive and its margin. LTO tickets are pretty good now (use LTT 4.0 SR2).
5) Save the ticket to a directory and send the files (1x header, 1x .dat file) to LTT_team@hp.com marked for my attention and I'll take a look too.

You could well be right - the drives are old and worn. Having said that, the HP LTOs go on for a long time so let's take a closer look before condemming them.

Good luck, Richard (LTT program manager)

[Rick Kirk]

Hi Richard,

Thanks for taking the time to look at this for me. Much appreciated.

I'll pull the support tickets and post them up tomorrow..

To be honest I'm amazed that I could have 3 drives in this condition. LTO-1 specs list a MTBF at 250000 hours at 100% duty cycle. One of these drives was only in service for 660 odd days (which is only 15840 hours at 100% utilisation!).

Thanks again and I'll speak to you soon.

Regards,

Rick.

[Rick Kirk]

Hi Richard,

Have done some more testing today. Here's what happened.

1. Took adaptec card from main PC to another.
2. Did a complete fresh install on WinXP Pro.
3. Ran cleaning cart through both drives.
4. Got some other test tapes (not new).
5. Ran Drive assesment test on both.

Low and behold drive 2 passed with a particular tape it in. Drive 1 failed with two different tapes. Took the tape that worked for drive 2 and put in in drive 1. It passed!!!

Very odd. Will now try the drive back in the old machine.

I have attached the logs and support ticket as requested. Any light you can shed would be very much appreciated.

Thanks.

Regards,

Rick

[Rick Kirk]

More updates...

I've put the best of the 3 drives back in my server and am now running some tests... Drive passed the assesment test and appears to be running well. I've noticed that it writes to the drive in two ways..

1. Tape spinning from card to spool I get 880-920 MB/min (which is fantastic and as expected).
2. Going the other way (different track), spool to cartridge I get 448 MB/min.

Point 2 is stopping me achieving 54 GB/hour but it's far better none the less. I may try an addition clean or two and see if I can unclock the heads a little more.

I'm amazed after everyhing that I've tried that it's finally starting to pay off. I've put cleaning carts through these several times. Today I opened a brand new cleaning cart which seems to have done the trick.

Cheers,

Rick.

[Richard Bickers]

Hi Rick,

Thanks for the tickets. It seems that the slightly odd identity of the drives is causing our decode algorithms some confusion so it's not as easy to decode as I'd hoped. That's a defect which we'll need to take a look at separately. Thanks for bringing it up!

What I could see was that your drives are on the edge of what HP would consider 'fit for purpose'. I.e. a capacity loss of 17% where our limit is 20%. Your drives should still work OK for backup/restore but there will be a corresponding drop in performance of 17% too. Different tapes will have their own slight variations which is probably why you were seeing some passes and some fails of the assessment test.

The drives have certainly been used quite a bit but not excessively. My guess this is a time impact rather than usage - i.e. several years of fans blowing dust through may well have clogged the heads up a bit. The head data doesn't show any particular head down - they're just all a bit low.

Maybe a few more occasional cleans and using new media will improve things still further. Fingers crossed on that one I guess.

As to your performance figures I'm not certain what you mean by card to spool but I assume you're referring to backup and restore performance. This opens up a whole new area - as it's often the system that can't keep up with the drive. Don't forget that a restore requires the directory to be re-built and that takes a lot of disk movement. It's not unusual for a restore to be half the performance of a backup if it's the system that is the bottleneck.

You can use LTT to measure the performance of your system and drive indpendently. Use Sys Perf to measure the system write and read and Drive Perf to measure the drive. There are instructions in the help but I can help if you get stuck.

[Rick Kirk]

Hi Richard,

Thanks once again for looking at the logs... Interesting that I've found something new for you guys to check out with the identity.

What I was refering to above is the write performance only. Using Novabackup (which reports MB/min figures) I notice that it writes one section of tape at maximum speed (900 MB/min), then stops and the next section is at 480 MB/sec. I assumed from this that internally it winds from the data cart onto the internal spool while writing, then moves the head to a separate track and runs the tape backwards from the internal spool back into the cartridge whilst writing that track.

I beleive these drives have dual read/write heads and assumed that it can write in both directions. What I was thinking is that one side of the heads is still dirty thus causing the drop in performance in one direction and not the other.

Just out of curiosity though. Assuming that I am now happy to commit my data to this drive, do you know if HP sell replacement heads for the ultrium Gen 1's and how would I go about sourcing them if they do?

Thanks again for all the info.

Cheers,

Rick.

P.S. I don't have a system performance issue. My machine is all sata 150 and the files streaming to the tape are 350MB each. Transfer rate seems more than adequate from the system side.

[Richard Bickers]

Hi Rick,

The LTO drives write data to the tape in 'serpentine' format. I.e. forwards to the end then backwards to the start. Each of these (single direction) is referred to as a wrap and there are 48 wraps for a full volume in LTO 1.

There are two sets of heads - one for each direction. Each set has 8 write heads and 8 read heads (for read while write - that's where the capacity loss comes from).

The backup app is unaware of all of this. It treats all tape drives the same - a serial write of data from BOT (beginning of tape) until it reaches EOT (end of tape). Any mention of tracks, sections or whatever is independant of the actual tape format.

Having said all that.... it's possible for one set of heads to be in better shape than the other so the different directions could perform differently. The direction change would be based on the amount of data though. For LTO the wrap length is 2.25GB (after compression). You'd find every other 2.25GB goes at a different speed if this is the case.

I'm afraid you can't fix anything inside the box. They are very finely aligned during manufacture and all of the parameters are measured and constants set internally during the production process. What you've got looks to be aligned OK - just a bit dirty. My feeling is that if you can use new media (which is slightly more abrasive than old) and use the occasional clean then you might be able to improve things. Please don't poke around inside or you could upset the alignment.

Good luck,

Richard.

[Rick Kirk]

Richard,

Thanks for putting what I was trying to say into English :)

What you have describes is what I am seeing. Fast/Slow every 2.25 GB. I'll keep at it in the hope that the second set of heads clean themselves up also.

Thanks for the assistance. Much appreciated.

I'm getting 73GB per tape which is fine for my purposes.

Just running a verify process at present though I expect it will perform fine.

Regards,

Rick.