Tape Drive Issues

[Pete Randall]

I've got a rp7410 running HP-UX 11.11 with DLT800 tape drives attached.  This hardware combination has been running without incident for years.  There have been no recent changes to the system, yet recently I've been having "NO_HW" issues with one of the drives (as reported by ioscan).

I swapped in my spare drive and ioscan reported it as "CLAIMED".  However, despite the fact that the SCSI ID was set to 3, just as the original drive had been, ioscan showed it as a completely different Instance Number and H/W Path (10/1/5/0.1.0) and no device files.  Out of curiosity, I set the SCSI ID to 4 and ioscan came up with yet another Instance/H/W Path combination (10/1/5/0.2.0).  Only when I set it to 5 did it show up at H/W Path 10/1/5/0.3.0, as desired.

Just to add to the confusion, the original drive was just fine when I connected it to a different box, so I swapped it back to it's original server, figuring that I wouldn't have to worry about the SCSI ID setting with this one.  This morning, it's state was back to "NO_HW" again and I went through the entire swap scenario again.  I'm now running on the spare drive with it's SCSI ID set to 6 in order to get a H/W PATH that shows an ID of 3.

So, my questions are.  Why does this drive get un-CLAIMED and what the heck is up with the SCSI ID on the replacement drive?

[Patrick Wallek]

Hey Pete,

Are you seeing anything in the syslog.log when the device is going NO_HW on you?  Any SCSI BUS RESET or something similar?

As far as the strange SCSI addresses -- Does that drive behave that way on both system or just on the one?  It almost sounds like the connection between the switches that set the SCSI id and the connection to the controller are off by one set of pins internally.

[Pete Randall]

Hey Patrick!

Hope all is well in the new job, new home, new city and the World Series!!!

As usual, it crossed my mind to check logs, but I forgot before I submitted my question.  There are some SCSI messages in both syslog and dmesg (SCSI Gross Error on 0/0/1/0/1), but they don't seem to apply to this H/W path.  Of course it's been a couple days since it last happened so I may be missing something.  If this recurs again, I'll do a better job of documenting exactly when so I can match that up with the logs.

Cheers,