Soft Write / CRC errors / End of tape reached

[jerryk]

I am aware of all of the other similar posts, but this problem isn't solved and it's driving me crazy.

The problem: zillions of soft write errors (CRC), causing tape advances until the end is reached and the backup crashes. Tape cleaning doesn't help.

I have an IBM DDS3 drive, and Veritas 9. It's in a new Dell 2600, connected to *it's own* adaptec 29160w scsi card.
The drive is connected via a new 50 pin ribbon cable to the card's internal 50pin connector. The drive is scsi terminated via a jumper on board.

This is the 2nd drive IBM has shipped in 7 months. This drive was originally in an older "homemade" server built by my predecessor, and had the same problems. I promised the company that a new server will fix the problem, and now I'm in a pickle.

Could IBM have shipped 2 dds3 drives with damaged read/write heads? Are there some arcane SCSI card settings I've missed? (I've set all the speeds on that scsi channel to it's lowest setting of 10mb/s)

Many many thanks for any advice you can provide.

 

[Rols]

Hi there,

Typically CRC errors are the result of a hardware fault (usually faulty read/write heads) or bad media.

If you have already had this drive swapped out - try using some different media.

Let us know how you go.

Cheers

 

[justin2000]

DDS3 are 4mm tape drives aren't they?

they write to the tapes exactly like your video player does.

they should be cleaned at least once a week, depending on how much use they have.

the only way around this, as rols said, try different tapes, otherwise replace the tape unit again, or get a different type - DLT or LTO are good, as they require little to no cleaning.

The other thing is, when you replaced the tape drive, did you also use new tapes? using old tapes, especially in a 4mm drive just transfers dirt from one unit to another.

 

[jerryk]

Thanks for responding guys. I can't tell you how *alone* I feel with this problem, and I really appreciate your thoughts.

Here's the unbelievable thing:
This is the second drive, and we just bought all new tapes this week. Same problems. It's amazing.

I'll call IBM and demand another drive, I guess. Can't believe they'd send 2 bad drives, but nothing about this problem makes any sense anymore.

What about making arcane tweaks to the SCSI channel settings on the card? Could the drive be so slow that no SCSI card can talk to it? Realistically, isn't there some kind of handshaking function on a SCSI card that should make it aware of a devices throughput?

 

[Rols]

I cant say that ive ever seen a CRC error that has been caused by a SCSI issue (thats not to say that it isnt possible).

Question: Do you format (ie: Quick erase or Label) the tapes before you use them??

Let us know

Cheers

 

[justin2000]

jerryk

if it is SCSI problems, check the System Log for any Event ID's errors, 7,9,11 and 15, these all relate to SCSI errors - which are a sign of hardware problems.

You can lower the transfer rate of the SCSI card for the ID the tape uses, also disable wide negotiation on the ID.

Check first to see whether you are getting the ID's above - it will give you a place to go on.

 

[jerryk]

It seems that each time the backup dies I get
error event 7 Once, followed by event 11 3times.


I'm guessing the drive is trashed. Could IBM send 2 busted drives though? Also, the new drive was in my OLD server (that I don't trust) for about 2months before putting it in my new Dell. Is it concievable that an older power supply could be spiking the drive or something?

 

[jerryk]

As for the question about Quick erases or labels... no.

I just set Veritas to overwrite, no append, and pop the tapes (labeled mon-fri on the cover) in.

Is failing to format or setting up electronic tape labels ahead of time going to cause backup failures? My assumption has been that Veritas have an error message like: "Please format the tape first" if that was the case. ...but at this point I'll try anything.

...and THANKS again for your help!

 

[SQLBill]

It's possible that IBM sent two bad hard drives (or that something happened to them in shipping). Did you send the original one back to them? If so, try calling them and seeing if they found anything wrong with it.

It's also possible that it was installed incorrectly. Call IBM and have them walk you through the installation steps. Tell them you are troubleshooting a problem and you want detailed steps.

Different situation-we had what APPEARED to be a hard drive crash on a cluster. Three hard drives seemed to fail at once. After finally getting a tech that knew their job, it was determined to be a RAID controller that failed. They sent us a new one but it didn't work. A second and third one also didn't work. We talked to the tech and asked him to 'treat us as though we didn't know anything' and walk us through each step. What-do-you-know! They never told us that we had to take a memory card out of the old/bad controller and put it in the new one. Problem solved.

-SQLBill

 

[jerryk]

Thanks SQLBill, I'm dealing with IBM tech support now. As for controllers, the tape drive is not on my RAID controller, I bought a separate adapatec 29160w just for the tape drive. (connected via the 50pin connector and a 50pin ribbon cable)

We'll see how it goes. I'll definitely post any successes I have.

 

[SQLBill]

Right, I didn't think it was on your RAID controller. I guess I didn't make myself clear enough. What I was trying to get across is that you might have missed a step that IBM didn't tell you about because they thought EVERYONE knows to do this.

Our problem happened with DELL equipment and when they finally told us about the memory card their comment was that they thought we knew we were supposed to remove the old one and put it in the new controller.

-SQLBill

 

[jerryk]

Well, IBM told me my termination settings on the scsi card may be insufficent, so I called Adaptec.

They had me "force" termination on the SE Channel, to accomodate the IBM people, although adaptec feels the automatic setting was probably ok. Also slowed the channel to 5mb/s, to further permit the drive to be as slow as it wants.

Now, I'll run some more backups and count soft write errors.

One question. How many soft write errors is ok on a 9gig backup? 10? 100? I'm getting about 9500.

 

[SQLBill]

I have an autoloader, so I checked one of my tapes:

Bytes written: 3391253.500 MB
Total Mounts: 286
Hours in Use: 119.991

And the only errors are:
Soft Read Errors: 42

-SQLBill