curt@mischief.ecn.purdue.edu (Curt Freeland) (04/14/90)
Several weeks ago, I posted to this group needing information on how to get an Exabyte 8mm tape drive to run on our DN3500 systems. I would like to thank everyone who replied to my post. With the help of you all, and a few people at Apollo, we are up and running. A brief synopsis: The problem: We had an Exabyte drive we removed from a SUN system that we wanted to connect to the WD controller in our DN3500. We also had purchased Omniback 1.1 from Apollo to use as the dump software for the Exabyte. We tried, but could not get things running. The diskless_node_mkdev command kept exploding trying to link the rmt_scsi binary and the tape drive special device files. The responses: The first response I got was from Susan Liebeskind (sl11@prism.gatech.edu). She had all of the answers, but I didnt put them all together until many weeks later. She stated that the SCSI stuff didnt work right for them under SR10.1. She told me of a "patch tape" that was required to run the SCSI devices on the DN3500. She later found the patch tape number, and emailed that to me also. Had I listened to her good advice, I would have been up and running several weeks ago! The next response came from Alan Holbrook (holbrook@apollo.com). He was surprised that nobody at Apollo had been able to help me (we had spoken to a growing list of Apollo folks, and none had been able to get answers for us). I mailed Alan some questions, and stumped him too. He finally put me in touch with Abe Shenker (abe@apollo.com) of Apollo. Abe determined that one of the binaries that was supplied to us was for the wrong architecture. Aparently, the SCSI stuff did not work correctly on the DN3500 (m68k architecture), but did on some other architectures (my interpretation, may not agree with Apollo's :-)). The SR10.1 release we were running had an 88000 (a88k) binary instead of the 68000 (m68k) binary shipped for the /sys/mgrs/rmt_scsi" file. Abe told me the fix was to get patch SFW-PSK4-MED10 (016955-A00). He also suggested we purchase the newest version of Omniback (1.2?), as it has many new features, and bugfixes. Wewanted to see the tape do something before we committed to spending any more money on it. The fix: We ordered the patch tape Abe and Susan mentioned, and received it a week later. Once we installed the patch, we had an m68k binary for the /sys/mgrs/rmt_scsi file. I ran the diskless_node_mkdev command, and all worked. Using the /etc/omniback/bin/new_mt command I could rewind, fsf, and generally diddle the tape drive. I then used tar to write some stuff to the drive, and read it back. The lab manager was very excited that all of this worked. So was I. Our multi-month quest was finally over! Thanks again to all who responded. Curt Freeland Manager, Systems Engineering Purdue University Engineering Computer Network (317) 494-3715 curt@mischief.ecn.purdue.edu