Mysterious Problems

It's déjà vu all over again!

Last month, it appeared that I'd successfully wrestled Windows NT into shape on its test-bed PC, making it work despite balky hardware and numerous software problems. As usual, the reality was a bit more complex.

After a week or so, the NT system stopped booting. No particular symptoms; it just stopped in mid-boot, leaving the NT startup screen and not bringing up the usual "Ctrl-Alt-Del to sign on" message. (About this time, I stopped worrying about using the three-fingered salute to start a PC. Am I the only inveterate DOS user alarmed by NT's choice of login keystrokes?) I'd been messing with the hardware configuration, so I used the "boot to previous configuration" choice in NT. Same problem; the system might be coming up, but it wasn't accepting any input.

I booted from the emergency boot floppy (if you don't have one, make one now; I wish I had earlier). Files seemed to be missing from NT's Setup. I booted from the Setup disks, told it to fix a damaged installation, and put in my other hardware selections. That didn't help. I reinstalled all of NT from ground zero. Still nothing! Had I messed up NT when I'd booted back to regular Windows? If so, how? Nothing I did in NT told me where the problem lay. I had to boot DOS and run Norton's Disk Doctor to discover that "bit rot" was setting in; changes and new files were not being saved correctly--under DOS or NT. Directories were corrupted in NT's hierarchy; it's a wonder the system could boot as far as it did.

Yanking the covers off the machine and restarting it exposed the problem right away. The five-year-old 51Ž4" full-high Siemens SCSI hard drive was beginning to fail. The physically larger, older drives make distinctive sounds; when a drive is beginning to have motor problems, the sound becomes irregular, because it can't hold synch. Smaller drives make these sounds, too, but it takes a stethoscope to hear them. (Years of fixing other people's hard drives do come in handy.)

Disk Sentinels and Spares
What does all this have to do with NT? It checks the drives at startup, running a chkdsk before booting, but it didn't know that its host volume was failing. It also didn't know that its own files were damaged beyond recovering.

Other operating systems have some checking of this sort: NetWare "hot fixes" a bad spot to a spare cluster; UNIX keeps spare tracks at the end of the disk. To be fair, though, NetWare and PC-based UNIX versions aren't intelligent enough to recognize the symptoms of a failing drive. They just suddenly have a growing number of bad spots. None of them truly watchdogs its storage the way mainframe operating systems do. There's definitely a place for a memory-resident product that watches for the classic symptoms of drive failure and alerts the user.

I assume that an NTFS volume, one formatted for NT only, would be more protection against these kinds of problems. (I have a question in to Microsoft about this.) I'm only running FAT volumes, though, so I can use the same disk space for either DOS or NT.

About 18 months ago, hard-drive manufacturers considered putting more intelligence on the drives themselves, so they could inform the operating system of problems before they became serious. Since then, hard drives have become virtual commodities, with the emphasis almost entirely on higher capacity, smaller size, and lower price. I haven't heard anything about drive-based sentinels since, however, so there's definitely room for a software product.

NT does have built-in support for RAID. At its simplest--Level 1--RAID consists of writing everything to two drives at once, transparently. Anyone familiar with NetWare knows this as disk duplexing; if you use two controllers and an external power supply for the second drive, it's called mirroring.

New Hardware, New Problems
The easiest fix to my problem appeared to be disk-to-disk copy via the network, install a new drive with DOS, and copy everything back. This "easy" fix chewed up two days and uncovered a new and more mysterious problem.

The new drive was a DEC 3GB hard drive, probably the last new full-high 51Ž4" disk drive I'll ever use. Almost everyone has migrated to 31Ž2" drives, which currently max out at 4GB--this migration has made the remaining 51Ž4" drives very inexpensive. I copied DOS and Laplink to the newly formatted drive and copied all the data back--minus NT which I hadn't backed up.

Then I ran into another anomaly, one I still can't explain. DOS booted, displayed the "Starting MS-DOS..." message, and then said "Unrecognizable disk/Press any key to continue." When I pressed the spacebar, it booted normally! Reinstalling DOS didn't fix this; re-SYSing the drive didn't; nor did reinstalling QEMM or BOOTCON. The only fix was to install NT version 3.51, which I'd gotten in the interim. Somehow, NT had became intertwined with DOS, and DOS wouldn't work correctly without it.

After arguing with computers for years, I've decided there are two kinds of PC problems: those you fix and those that disappear mysteriously after the proper rituals are observed. This one definitely falls into the second camp.

Installation--Again!
Whatever the cause, NT (and DOS) worked and presumably had all the fixes that shipped in version 3.51, too. NT now autorecognized my SoundBlaster-connected IDE CD-ROM drive but not the Qlogic VL-bus SCSI disk controller. However, once I used the Qlogic install disk, NT recognized my hard drive.

Incidentally, the AdvanSys SCSI controller, the one that wouldn't work under NT last month, is happily running a NetWare 3.11 server elsewhere at Chaos Manor. I suspect drivers caused NT's allergy to it. The power supply on that particular NetWare server cooked its old motherboard; I'm amazed the hard drive survived the experience. This is exactly why mirrored drives--whatever the operating system--should have separate power supplies. 110V will cook two drives as quickly as one.

NT did recognize my Stealth 64 VL-bus video card as an S3-chip-based product. After I finished installing with the default settings and rebooted, I used Diamond's own NT drivers and set my video to 1024 * 768 pixels again. But NT 3.51 had no better luck recognizing my sound card. I'm puzzled as to why a SoundBlaster-16, the most popular sound card on the market, isn't recognized, although its onboard CD-ROM is. Once again, installing the drivers from within NT fixed it.

Printing
NT 3.51 recognized my Intel EtherExpress Ethernet card and the network. It was time to hook NT up to our print queue via the network.

Another subtle difference between "classic" Windows 3.1 and NT is in printing. Instead of having a separate printer control panel and print manager, the two are one, which makes management easier. Printing isn't quite object-oriented á la Windows 95 yet, but it's definitely smoother.

The main printer we use here at Chaos Manor is a LaserJet III hooked to Valiant, an OS/2 Warp Connect computer. Printing is much faster on OS/2 than on any version of Windows. Windows 3.1's printer connect function sees Valiant; another OS/2 machine can see it; but Windows 95 and NT can't.

There is a fix. If you type the exact Universal Naming Convention code (UNC) for the printer, \Valianthplaserj, NT can connect and print to it and even tell you its status. Somewhere between Windows 3.1 and today, Microsoft stopped listening to OS/2's service broadcasts. Anyone using a mixed network must record the exact names of OS/2, LANtastic, etc., file and print servers for use on Windows 95 and NT.

NT adds one nice feature to printing: When you select the target printer, NT uses its description object if possible. On a network, this makes lots of sense: The workstation gets the description from the printer itself, so only one object needs to be updated if the printer changes.

Not surprisingly, the OS/2 computer couldn't supply a satisfactory description to NT, so I had to load the LaserJet III definition from CD-ROM. It worked! I could finally print from Word for Windows 6.0c. And using the exact UNC let me share Windows 95 and Windows 3.1 disks, too.

Off for Now
Next month, I'll talk about applications, from the Internet to graphics to Microsoft Office, discuss more about disk drives under NT, comment on NetWare vs. NT, and doubtless have one or two more tales about hardware.

Until then, I'd like to hear from readers who are using NT, either in real-world business situations or as a Windows 3.1 workstation replacement for large tasks. Now, I'm off to SIGGRAPH, the yearly computer graphics showcase, to see if the computer-graphics houses are using NT instead of Silicon Graphics and Macs.