by Wayne M. Krakau - Chicago Computer Guide, September 1995

It’s like the old saying "With friends like these, who needs enemies?" With technical support like this, who needs bugs.

It started simply enough. We planned a network upgrade using a new, more powerful file server with a 32-bit bus. We would reuse the existing SCSI-2 (Small Computer Systems Interface, Version 2) duplexed Toshiba hard disks and Data Technology Corporation (DTC) controllers. (Duplexing is a variation of mirroring, employing both dual disk drives and dual controllers to provide redundancy while improving performance over the equivalent mirrored configuration, which uses a single controller.)

The client already had a copy of NetWare 4.1. He also had a Pacific Micro Data DAT (Digital Audio Tape) tape drive (really an older Hewlett-Packard model in disguise), an Adaptec AHA-1522 SCSI controller, and Palindrome’s DOS-based backup software to implement a server-based backup system, replacing the workstation-based backup system they had previously used. As per to common practice, we were giving the tape drive its own controller to maximize performance.

We were supplying the file server, a DTK 486/DX4-100 with three bus-mastering VL-Bus slots. This machine was a little light for their needs, but it is a not-for-profit organization that has to make due on a shoestring budget. The same was true of the salvaging of old parts, especially the three 16-bit SCSI cards that were being installed in a machine with 32-bit slots.

The first bad news was from Data Technology. After repeated attempts to install the new tape drive on the old file server, the client was told by DTC that Adaptec’s ASPI (Advanced SCSI Programming Interface) software would not coexist with their ASPI-emulator software in a NetWare server. Also, their ASPI-emulator wasn’t close enough to real ASPI (designed by Adaptec) to run an Adaptec controller while their Adaptec-emulator chipset on their boards wasn’t close enough to the real Adaptec chipset to be run via real ASPI software. Stalemate! We decided to switch to a single 32-bit Adaptec 2840 VL-Bus card (to stay at least close to the budget) to mirror the disks.

When we delivered the new file server, with NetWare already installed on the redundant hard disk taken from the original server, we couldn’t get the tape drive to consistently respond to system level commands. Finally, after two extended sessions with Adaptec’s very patient technical support team, the server itself failed. Using a hardware diagnostic board, I determined that the motherboard-based processor interface had failed. Oh well, back to the office.

DTK gave us a warranty replacement motherboard immediately. (I drove over to their Elk Grove Village, IL plant to pick it up right after I requested it. The location of their Midwest plant - one of six nationwide - is one reason that I deal with them. Immediate parts availability and in-person tech support can be very handy.) Even with the new motherboard, the system wouldn’t work. I started swapping parts with some of our in-house computers and found that the motherboard wasn’t the problem! (Oops. New personal rule - don’t try sophisticated troubleshooting when spaced out from too many hours of continuous work.) The Adaptec AHA-1522 was bad, and when it finally completely failed, it took the IDE/I/O board with it. Replacing both boards was the solution. (Note that I am switching between "we" and "I" because most of this project was a client/KBCS staff team effort, not because I have a big enough ego to use the royal "we".)

Since there were two open 32-bit slots in the server, the client wanted to replace the 1522 with another 2840 VL-Bus card. He checked with Palindrome and was assured that was a valid combination, but when we tried to install the Palindrome software, we had trouble getting it to work. Sometimes it worked. Sometimes it locked up the workstation that ran the control module (the actual working modules run on the file server). Sometimes it locked up the server. After an extended session with Palindrome tech support during which I could tell that the person I was talking to was constantly consulting his colleagues, we brought the system back to our office again.

On the next day, the client called Palindrome before I had a chance to. He was informed in no uncertain terms that Palindrome had never approved the use of the 2840 card(odd, since it uses the same chipset as the 2740 EISA series which is approved by Palindrome), and that there was no way that anyone on their staff would ever even imply that was a valid controller to use with Palindrome’s software! This was after two phone calls approving the use of the 2840 card, one of which included over an hour and a half of experimenting with different configurations of the board. The tech support person (as I stated, backed by his colleagues) even conversed in depth with me about the ins and outs of VL-Bus busmastering technology. There could be no mistake in either my conversation or my client’s that the 2840 VL-Bus card was under discussion.

At this point I gave the client one of our in-house Adaptec AHA-1542 16-bit SCSI controllers. I was able to back up successfully with the 1542 controlling the tape drive while a 2840 controlled the disk drive, so I redelivered the system.

Since the client now had an extra 2840 32-bit controller, he decided to implement disk duplexing, but he kept having trouble with the tape software. Thinking that he would save me some trouble, he called Adaptec to find out how to get the three SCSI cards to coexist. This turned out to be a well-meaning mistake since I had talked to Adaptec and had instructions already prepared for the three-card configuration. He managed to get an inexperienced technician who didn’t seem to know his own limits. The configuration advice that he got was so bad (though admittedly Adaptec’s only error in a long stream of calls on this project) that it the system started generating random hard disk errors, one of which, by chance, trashed sector zero, track zero - the one vital spot on the drive. After repeated attempts to recover the drive using various disk utilities, we finally had to give up. We reformatted the drive, losing all of the data, and hoped that the one backup we had was good. Note that the old server was still in production and could be used if the tape was bad, but only with much redundant work.

After reinstalling NetWare, the client could not get the tape software to work. This time, when he called Palindrome, they told him that their software wouldn’t work in a server that had even one, much less two, 2840 cards in it, even though those cards didn’t have anything to do with the tape drive! They further stated that they had called both Adaptec and Novell and found that the 2840 was "NOT APPROVED FOR USE IN A NOVELL SERVER"!!!!

Right after this, the client pulled out the 2840 box and documentation. The box was plastered with "Novell Approved" and "Works With NetWare" stickers. The documentation had multiple chapters on how to install the board in a NetWare server. He called me regarding this shocking conversation, so I immediately called Adaptec to get the facts. They told me that not only was Palindrome wrong, but that Palindrome might be expecting a letter from Adaptec’s legal department with the magic phrase, "Cease and Desist" on it, along with a not-to-subtle threat of legal action.

At this point, the client demanded that I take action to solve this dilemma, so I gave him a new AHA-1540 16-bit card to use with the disk drives in place of the now-forbidden 2840s. This combination at least partially worked, though using 16-bit controllers slowed the server appreciably. I use the term "partially" because the Palindrome software was unable to successfully restore NetWare 4.1 rights and NDS (NetWare Directory Services) information. Only the raw data was retrieved. The client had to reenter the NetWare-specific information, even though the software is, in theory, approved for use with NetWare 4.1.

I knew I should have been suspicious when the Palindrome installation routine demanded to be run in bindery mode using the old NETX Workstation Shell instead of the current VLM DOS Redirector. This is of course, not to mention a user interface that would have embarrassed a 1970's era mainframe programmer (keeping in mind that I was one).

The final insult happened when the client called Palindrome to demand a refund, since, after even more tech support calls, the backup system still wasn’t trustworthy. Palindrome management told him, in essence, "tough luck". They cast aspersions on my company’s competence and ethics, and hinted, subtly, that the client had been stretching the truth for effect.

Oh, I thought that it might be interesting to point out that the client system administrator in question is an ordained cleric who uses his title in all introductions, and that organization that he works for is, naturally enough, a religious one. If you believe in that sort of thing, is there a special place in Hell for those who screw members of the clergy in business transactions?

1995, Wayne M. Krakau