Juggling Windows NT Servers: A Special Report
Robert Bruce Thompson
|It now appears that Robert Bruce Thompson and I
will be doing a book on "good enough hardware." Stay tuned. Thompson is the
author of several O'Reilly books. I have never seen a bad O'Reilly book, which is
Other reports by Thompson:
Juggling Windows NT Servers
As regular readers of my Day Notes journal will recall, I got the new Pentium II box built on the carcass of old kerby, (detailed in An Upgrade for Old Kerby and Finishing the New Pentium II Box) and got it configured as a Backup Domain Controller (BDC) on the TTGNET domain of which thoth is the Primary Domain Controller (PDC). At this point, the new box, temporarily named sophocles, was ready to replace Barbara's main machine, thoth, but was still pretty much a blank slate.
Old thoth is the primary data store for my network, so replacing it will require moving all of the data from thoth to the new box. Because thoth is also Barbara's main workstation, I'll also need to migrate or reinstall all of her applications. thoth with the new box will involve migrating her applications. Finally, because a lot of things point to thoth by name, I decided it would make sense to rename the new box thoth to avoid having to reconfigure everything else.
Yesterday evening (Saturday, October 10th) I finally got around to moving sophocles back to Barbara's office to replace old thoth. I'd done a full backup on Friday afternoon, so that part was taken care of. To start, I copied the /usr folder from old thoth to sophocles. That contained about a gigabyte of data, which took 20 minutes or so to copy. I then promoted sophocles to PDC in the TTGNET domain, which automatically demoted thoth to a BDC. That done, I renamed old thoth to gladly (the cross-eyed bear), powered down all the machines on the network, disconnected gladly and put it aside, and moved sophocles back to Barbara's office.
After reconnecting sophocles in Barbara's office, I powered it and all the other machines except old thoth/gladly back up. I renamed sophocles to thoth and restarted new thoth. Everything appeared to function normally, but I couldn't access any of the other machines from new thoth. When run on new thoth, Network Neighborhood showed only thoth and the TTGNET domain, but no TTG domain. When run on the other machines, Network Neighborhood showed both the TTGNET and TTG domains, and all machines other than thoth. Obviously, thoth wasn't talking to the network.
My first thought was that since new thoth was running only TCP/IP, an IP address problem might exist. I reconfigured new thoth to use 192.168.111.165, the IP address that had been assigned to old thoth. That didn't seem to help. My next thought was that I might have browser/NetBT problem, so I installed the NetBEUI protocol on thoth. With that done, it should have been able to browse other machines on the network with broadcasts, if nothing else. No joy.
I was getting ready to install the Network Analyzer to examine packets when a thought hit me. I walked down to my office. Sure enough, the 100BaseT Ethernet cable that should have been running from the wall jack to the hub was disconnected. That cable had linked sophocles to the hub, and when I moved that machine I unplugged the cable from the Ethernet card and left it lying on the floor instead of plugging it into the wall jack to link Barbara's office to the hub.
The first rule of networking says that the easiest way to increase network throughput is to connect the cable. Once I connected that cable, thoth was visible to the network, and the network was visible to thoth. The lesson here is that you should always check the simple stuff before you assume that a problem is due to something complicated. If I'd done that, I'd have saved the half hour or so I spent puzzling out possible transport protocol problems and browser configuration glitches.
With new thoth up and running, the next step was to get some of Barbara's stuff migrated off the old box and onto the new. I carried gladly back down the hall to my office, disconnected the old 10Base2 cable that used to feed Barbara's office from the hub, and connected gladly directly to the coax port on the hub. I fired up gladly and as it was booting my heart sank. I'd forgotten to change the IP address on gladly, so it was still set for 192.168.111.165--the same IP address currently being used by thoth.
I'd have killed the power, but gladly had already almost finished booting by the time I thought about it. The first rule of TCP/IP networking is that IP addresses must be unique within a network, so I wasn't sure how NT would react. As it happens, NT reacted very gracefully. It informed me that it had found another machine on the network with the same IP address, and started the computer with networking disabled. It took only a minute or two to reset the IP address to an unused value and restart the box. Once I did so, gladly booted normally. When I went back to Barbara's office, I found that thoth was displaying an information message about the attempt to boot another system with its IP address. This message appeared only on thoth, so apparently only the two machines involved are notified.
With gladly and thoth able to communicate, the next step was to reinstall and migrate Barbara's applications. I'd already installed IE4, Outlook 98, and Office 97, but apparently I installed Office 97 from the wrong CD. Barbara's system has the full Office 97, which I purchased as an upgrade the day it came out. This includes Access 97, which she uses to maintain various databases. The CD I installed from was apparently the Office 97 Small Business Edition, which does not include Access 97. It looks like I'll need to de-install and then re-install.
That's no great loss, actually, because the SBE version that's on there won't update to Office 97 SP1. Running SP1 just causes the machine to churn away for 30 seconds or so and then return a message that says that SP1 is unable to update this installation. At first, I thought perhaps this was due to the fact that I installed IE4 and Outlook 98 after installing Office 97. Apparently not, because I uninstalled both of those products and SP1 still wouldn't touch the Office 97 installation. Well, Barbara doesn't use Word much or Excel at all, so we'll live with this for the time being.
The next step is to get her games migrated over. She doesn't have many, and they're mostly pretty simple ones--card games, Mahjongg, Scrabble, etc.--so I'm hoping I can just copy the program folders. Yep. I copied the program folders over to the new box, and all of the games run fine. I really like programs that keep all of their required components in one place instead of scattering DLLs all over the disk and making entries in the registry.
Well, they run fine except that there's no sound. Of course there's not. The Creative SoundBlaster AWE-64 card is an ISA PNP card, and Windows NT doesn't know about those, at least until you give it some help. I already have SP3 installed on this box, and I don't want to load the drivers from the original distribution, which would require reisntalling SP3. As things stand now, NT doesn't even know the card is installed.
So, I use the /x switch to expand the SP3 .exe file to the \junk folder and display the directory listing. The file I need right now is Pnpisa.inf. I right click it and choose install to tell Windows NT to load the Pnpisa.sys driver. With that driver installed, and the system restarted, Windows NT immediately displays the Found New Hardware dialog. From there, getting the correct SB drivers installed from the SB CD is trivial. It's just getting NT to see the card that's not intuitive.
With the SoundBlaster card and drivers installed, the games use sound as expected. One problem, though. Barbara plays audio CDs from her CD-ROM drive, and when we try playing one no sound comes out of the speakers. Arggh. Of course not. The original kerby, upon whose carcasss this machine was built, had a CD-ROM drive, but no sound card. Playing audio CDs through the sound card requires a cable to connect the back of the CD-ROM drive to the sound card, and I didn't install one.
It would be nice if the SoundBlaster card came with such a cable, but it doesn't. They'd probably argue that there are three different audio connectors used on different CD-ROM drives, so they couldn't know which cable to include. That's a false argument. Octopus head cables are readily available that have the SB connector on the sound card end and a three-headed connector on the CD end. One of those cables should come with every sound card. I fished around my cable bin, but couldn't locate one. I guess this means another trip to CSO to buy a $2 cable. Geez.
Okay, Barbara says that she can live without the ability to play audio CDs for now, but what she really needs is to be able to synch her PalmPilot. I've already connected the cradle to the serial port, and I'm hoping that I can just copy the \Pilot folder from gladly. Trying that quickly establishes that it won't work, so I'll have to reinstall from scratch. That means installing the original Pilot software, then installing the Desktop-To-Go synch software she was using with Outlook 97, and finally installing the DTG upgrade to support Outlook 98.
After doing all that, I put the Pilot in its cradle and press the Sync button. Nothing happens. The Pilot informs us that it can't talk to the PC. Okay, I thought it was on COM1: but perhaps not. I reconfigured it to use COM2: with the same lack of results. Okay, back to COM1:. I notice that the speed is set to "Fastest Possible" and I seem to remember that it ran 38.4 Kbps before, so I tried reducing the speed. Still no joy.
Perhaps something is wrong with the serial port itself. The Pilot cradle has a DB9 serial connector, and this PC has one DB9 serial port and one DB25 serial port. The easiest thing to do would be to connect the Pilot to the DB25 port, but I can't find a DB9 to DB25 adapter. By this time, it's getting pretty late and I'm getting tired. Thinking about it, there's nothing undone that prevents Barbara from doing anything she needs to do, albeit with some inconvenience. If I move the Pilot cradle over to gladly, she can synch from there.
I'll put off fixing the CD audio cable and the Pilot serial port problem for another day. Barbara asked me if this was all going to be worth it. I think the answer is yes. The new box "feels" about five times faster than the old one. About half of this difference is due to the difference in clock speed--133 MHz on the old versus 300 MHz on the new. The remainder can be attributed to the greater efficiency of the Pentium II versus the Pentium (particularly when running NT) and to the fact that the new box has 128 MB of RAM, versus 64 MB on the old box.
All in all, this has been a worthwhile upgrade, and a relatively inexpensive one. Doing it today, I would probably have chosen the 333 MHz Celeron-A. It's cheaper than the Pentium II, fits the same system boards, and would probably be about as fast. It includes only 128 KB of onboard L2 cache (versus 512 KB on the PII), but that cache runs at full processor speed, rather than at 1/2 processor speed as in the PII.
Send feedback to: firstname.lastname@example.org