Ask Dr. Bob Your NT Questions - 01 Apr 1999

SEND US YOUR TIPS AND QUESTIONS.
You can also visit Bob Chronister's online Tricks & Traps at http://www.winntmag.com/ forums/index.html.

Q: How do I control synchronization between my Primary Domain Controllers (PDCs) and Backup Domain Controllers (BDCs) to minimize superfluous network activity?

Windows NT 4.0 offers only a partial solution to this problem. The Netlogon service tries to optimize domain controller synchronization by considering factors such as the amount of network traffic and choosing a replication interval that it deems best to optimize domain controller synchronization; however, you can change a default Registry value to reduce the frequency of replication. Go to the HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services \Netlogon\Parameters Registry key, and change the value of the ReplicationGovernor entry (type REG_DWORD). The default value is 100, and lowering the value to 50, for example, reduces the replication interval by half. Lowering this value reduces the amount of replication traffic over the network but also compromises domain controller synchronization.

Q: When I use an unattended Windows NT installation script with the new 3Com 3C905B NICs that ship with Dell systems, the cards won't install and the installation hangs. Do you know what the problem is?

The oemsetup.inf file that Dell supplies with the cards doesn't support an unattended NT installation. A problem in the file causes an incorrect path search, which results in the hang-up. Dell has rewritten the .inf file to fix the problem. You can obtain the rewritten file from Dell's technical support.

Q: When I try to download Service Pack 4 (SP4), the file appears to be downloading but never transfers to my computer. What's the problem?

I have the same problem. You're probably using an Internet Service Provider (ISP) that doesn't allow large-file transfers. The solution is to find an ISP in your area that allows large-file transfers.

Q: How do I swap a hard disk to recover a crashed Windows NT system?

If you can't repair the disk, you can replace the hard disk if you have a backup of your NT system and a 3.5" boot disk. If you have more than one hard disk, use the following instructions for your recovery:

1. Install a bare NT installation on the second hard disk.
2. Remove the damaged disk, and install a new disk.
3. Boot from your 3.5" boot disk, and choose the duplicate version of NT in the boot.ini menu.
4. Open Disk Administrator, and partition and format the drive.
5. Insert the tape of your latest backup, and restore the C drive.
6. Repair the boot sector of the C drive with the installation disks, if necessary.

If you have only one hard disk, use the following set of instructions:

1. Remove the damaged hard disk, and install the new one.
2. Run the NT installation, and install a minimal system. Be sure to add the backup driver.
3. Run the Backup command to restore the disk from your tape backup. In this situation, you won't have to repair the boot sector.

These procedures will work whether you are running FAT or NTFS.

Q: I'm installing Service Pack 4 (SP4) on a critical server. Do you have any advice?

First, decide whether you need SP4. Users often feel that they need the latest fixes, but you need to carefully consider this decision for a critical server.

The following steps outline how to install a service pack in a critical environment. This process assumes that you have accounted for redundancy and that the server can be offline for several hours.

1. Update all security settings and the Registry with the RDISK /S command.
2. Perform a complete server backup, and be sure that the tapes work and that you can easily restore the files. I recommend doing two backups and taking one tape offsite.
3. Disable any unnecessary drivers (e.g., a UPS monitoring applet) and all OEM drivers, if possible. (Screen 1 shows a disabled Wacom driver.) Consider selecting a generic driver (e.g., VGA) as your video driver.
4. Check to see whether you need new versions of OEM drivers.
5. Reboot the machine, and check NT Event Viewer to be sure the system is running without any problems.
6. I strongly suggest that you obtain the service pack on a CD-ROM. You can download SP4 from the Internet (you need Internet Explorer—IE—3.02 or later for this installation to work properly), but the installation will fail if you lose your connection during the download. The CD-ROM has an autoplay feature that starts your Web browser and creates an Install Service Pack option; select this option.
7. Answer yes when the system asks to save or open spsetup.bat.

If your Web browser doesn't open, go to \I386\Update or \Alpha\Update (depending on your machine type) on the CD-ROM. Click Update, and follow the onscreen instructions.

If you decide to undo the installation, you must accept the uninstall option the first time the system offers it. If you have a problem updating a file, you need to uninstall from the Add/Remove Programs Control Panel applet or run spuninst.exe from the System Root directory. SP4 requires a minimum of 80MB of free space.

Heed one very serious warning: Uninstall won't remove the changes an SP4 installation makes to the Security Accounts Manager (SAM). If you reinstall Service Pack 3 (SP3), you will have to click No in the Confirm File Replace dialog box, which asks whether you want to overwrite samsrv.dll and winlogon.exe. If you overwrite the new SP4 files with older SP3 files, you won't be able to log on to the system. Also, if you reinstall SP4, you'll need to create a new uninstall folder (so you'll have to move the old uninstall folder). As you can see, installing SP4 isn't a simple task!

Q: How do I set uniqueness database files (UDFs) for an unattended installation of Windows NT 4.0?

When you set up a large-scale unattended NT installation, adding a unique ID for every user is difficult because unattended.txt refers to one user. However, you can use UDF to merge into or replace sections of the answer file (unattended.txt in this example) for the GUI section of a software installation. UDFs are ASCII files that you can easily modify using a simple text editor such as Edit, which you run at a command prompt. You can break down a UDF into a UniqueIDs section and a section for each UniqueID parameter that the file specifies. The UniqueIDs section identifies each unique ID you want the answer file to include and the sections of the answer file that the UDF will merge into or replace. The UniqueID parameters sections define the specific data that the UDF will place in the answer file.

For example, I created UDF.txt, which Listing 1, page 196, shows. UDF.txt is a UDF with two unique IDs: Bob1 and Bob2. Bob1 and Bob2 have three sections—Userdata, GuiUnattended, and Network—that will merge into or replace the corresponding sections in unattended.txt. First, UDF.txt defines the unique IDs and unattended.txt sections that it will provide replacement data for. Next, UDF.txt provides the replacement data. Each UDF.txt section begins with the unique ID, followed by a colon and the unattended.txt section name that the UDF.txt section's data refers to. The UDF merges these parameters into unattended.txt, rather than replacing unattended.txt data, because these parameters aren't in the unattended.txt file. If the parameters already existed in the unattended.txt file, which Listing 2, page 196, shows, the UDF parameters would replace the unattended.txt's existing parameters.

Merging this UDF file and the unattended.txt file requires command-line switches. The following switches work with winnt.exe or winnt32.exe:

• /u specifies the name of the answer file
• /s defines the NT source
• /udf determines the UniqueID and the local name of the UDF

The following example shows how to start an unattended NT installation using a UDF and the unattended.txt file:

winnt32 /u:unattend.txt /s:h:\ /udf:Bob1,h:\udf.txt

As you can see, a UDF contains all the parameters that are unique to an individual. The unattended.txt file contains statements common to every user.

Q: When I attempt an unattended Windows NT installation, the system sends me the following error message: Setup is now ready to guide you through installation of Windows NT Networking. If you want to review or change any settings before continuing, click Back. To begin installing Windows NT Networking, click Next. I checked my adapter, protocol, and service sections, and they're all fine. Where is the problem?

I've received this message, and the solution is simple. I forgot to assign a computer name in the uniqueness database file (UDF). After I entered the computer name, the system stopped sending the error message, and I proceeded with the unattended installation. (For more information about setting up a UDF, see the previous question, page 194.)

Q: How are distributed systems and parallel systems different from each other, and how do they relate to clients, servers, and mainframes?

A distributed computing environment is a network of computers working together to complete a common task. For example, in a manufacturer's distributed computing environment, separate computers handle each phase of the manufacturing, but all computers communicate with one another and notify one another of the status of the manufacturing. The advantage of this setup is that you can place resources on the systems that need them and base the network's resource allocation on resource need. You can build each job-related task on a client/server model but maintain an overall organization to the environment—all tasks support selling and delivering the product.

In contrast to a distributed environment, mainframe environment's centralize all of a network's processing in one or several mainframe computers. Unlike PCs, mainframes have a reputation for being reliable and durable. The mainframe is the hub of all tasks in an organization. The problem with these systems is that they're expensive and they create one point of failure. However, mainframe environments are far from dead, and I don't foresee their disappearance any time soon.

Parallel processing uses a different processing algorithm from distributed systems. In this processing design, a system breaks down a large data query into smaller queries and passes these smaller queries to nodes in the environment. The nodes process these queries in parallel. When the nodes finish processing, the system joins the smaller queries' responses to form the answer to the major query. This setup offers several advantages:

• Parallel processing optimizes processing time, particularly with large-database handling that is CPU-intensive.
• A properly managed database can scale massively.
• You don't have to lock files; parallel processing is concurrent, so users can't write to a file and corrupt it while another system is processing it.
• The network automatically recovers from the failure of any node.

Parallel processing is possible on the old Tandem ServerNet systems and cluster nodes of PCs running Microsoft Cluster Server (MSCS) over Windows NT, Enterprise Edition. However, parallel processing represents a very small segment of computing. Parallel processing treats all the systems on the network as equals and treats all applications the same.