\[Editor's Note: Share your Windows 2000 and Windows NT discoveries, comments, problems, solutions, and experiences with products and reach out to other Windows 2000 Magazine readers (including Microsoft). Email your contributions (400 words or less) to [email protected] Please include your phone number. We edit submissions for style, grammar, and length. If we print your submission, you'll get $100.\]
Another LPR Problem for NT Server 4.0
As I mentioned in Reader to Reader: "NT Print Server Fix" (January 2000), Windows NT Server 4.0 has had problems as a print server. For example, users report that their print jobs just sit in the queue and don't print. When I investigated the Application log of an NT print server that had this problem, I uncovered the following entry:
Event ID 2004: Printer RAW on host <IP address of printer> is rejecting our request. Will retry until it accepts the request or the job is cancelled by the user.
This entry is only a warning message. To temporarily correct the problem, I reset the printer's power.
Later, I discovered a fix in the Microsoft article "Spooler Service LPR Monitor Leaks Memory with Each Event 2004 Error" (http://support.microsoft.com/support/kb/articles/q223/7/91.asp). The fix is an updated Line Print Remote (LPR) monitor. After I obtained and applied this fix, users no longer experienced hung print jobs.
Setting Services' Startup Order
In Reader to Reader: "Startup Sequencer," October 1999, Paul Bruesch provides a solution to a services startup-sequence problem. I work for a software development company, and we have also experienced problems with the order in which Windows NT starts services. A problem occurred in which the server application that my company wrote (called Service 1) was starting before the SQL Executive service.
To solve this problem, I used regedt32 to add the DependOnService subkey of type REG_MULTI_SZ to the Service1 subkey of the HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSetServices Registry key. Then, I added SQL Executive as the service that Service1 depends on.
In the DependOnService string, you must use the registered service name, which isn't usually the name that the Control Panel Services applet displays. To find a service's registered service name, look in the HKEY_LOCAL_MACHINE\CurrentControlSet\Services Registry key.
Tin-Pinned PPGA Cards
If your system uses PPGA cards and displays strange symptoms such as power problems, system freeze ups, and ghost reboots, the card's tin (i.e., silver-colored) pins might be the culprit. The tin oxidizes over time, which results in contact problems. If you replace tin-pinned PPGA cards with gold-pinned cards, the problems will disappear.
Free Remote Control Software
I recently discovered AT&T Laboratories' platform-independent, free remote control software, Virtual Network Computing (VNC). This program is small and simple, and you can run the Win32 viewer (which is about 150KB) directly from a 3.5" disk. The server-side installation takes only about 1 minute to set up and configure. You can display a desktop running on a Linux machine, on a PC, on a Solaris machine, or on several other architectures. For more information and to download this product, go to http://www.uk.research.att.com/vnc/index.html.
Avoiding Problems with Diamond Monster Products
In response to Reader to Reader: "Using a Diamond Monster AGP Video Card on NT 4.0" (Winter 1999/2000), you don't have to modify the Registry to install a Diamond Monster product. My experience has proved that to install any Diamond Monster product on a Windows NT 4.0 system, you must reset your video display to Standard VGA before you try to install Diamond Monster's drivers.
However, if you use the Diamond Monster AGP G460 card on an NT 4.0 system that has an Intel 440BX-2 motherboard, you must use the Intel native driver for the I740 graphics chip embedded in the card. Otherwise, the system will crash.
SMS 1.2 to SMS 2.0 Problem
If you're considering upgrading from Microsoft Systems Management Server (SMS) 1.2 to SMS 2.0, plan, test, and test again. Despite my company's attempts to thoroughly test the upgrade before deploying it, we still encountered several problems.
One of the major problems we faced had to do with the remote control portion of the client installation. To automatically start the remote control client on SMS 1.2 desktops, we added C:\ms\sms\bin\wuser.exe to the HKEY_LOCAL_MACHINE\SOFTWARE\RunServices\RemoteControl Windows 95 Registry subkey. We felt this method was the best and safest way to automatically start the SMS 1.2 remote control client without end-user involvement. Unfortunately, when we tried to upgrade the SMS 1.2 remote client to SMS 2.0, the installation wouldn't successfully complete. When we went to the Components tab of the client's Control Panel SMS applet, we discovered that the upgrade installed all client components except the remote control component. The tab listed this component's installation status as pending. In most circumstances, SMS 2.0 will continue trying to complete the client installation on subsequent reboots of the client machine. However, in this circumstance, we had to remove the Registry entry and any instance of the wuser.exe file. Then, we manually repaired the installation using the Repair Installation option on the Components tab of the client's Control Panel SMS applet.
STA Causes IP Address Renewal Problems
My company was experiencing problems obtaining IP addresses from our DHCP servers. If users left their computers (i.e., Windows NT Workstation 4.0 and Windows 9x systems) off for a period of 3 or more days, the DHCP server wouldn't renew their IP address. I could go to a DOS prompt and manually renew their IP address without a problem.
I tried to extend the lease expiration from 1 to 3 days, then to 5 days, but users still experienced problems even when their computers were off only for a weekend. Random computers experienced this problem. Searching the Microsoft site for a solution didn't return any helpful information.
The NT Event Viewer on one of the NT Workstation 4.0 systems provided the following information about the problem:
DHCP failed to obtain a lease. The semaphore timeout period has expired.
This entry provided a clue about the cause of the problem, and I searched the Microsoft site for information about the semaphore timeout period. The search returned the Microsoft article "DHCP Renewal Failures on Switched Networks" (http://support.microsoft.com/support/kb/articles/q168/4/55.asp). This article looked promising because I had just upgraded all our 10Mbps hubs to Cisco Systems' 100Mbps switches.
According to the article, if the switch uses the spanning tree algorithm (STA), the system can lose your DHCP request packet. Although STA provides useful functionality (e.g., it blocks loop conditions on ports, enables backup loops between switches in case of cable or port failure, provides some fault tolerance for incorrect wiring, helps switches discover the best path), in my case this algorithm was the culprit.
I checked the Cisco documentation, which revealed that the company's switches use STA. I used HyperTerminal to assign the switches an IP address, then I connected to each switch over the Internet and used Cisco's Visual Switch Manager to disable STA. The method to disable STA varies, so check your switch documentation for the specifics. You need to disable STA only for the ports connecting the switches and not for the ports that clients use, but the Cisco models I have only let you enable or disable STA for the entire switch.
Certifications + Experience = Fellowship
I agree with Robert Ian Clark in Reader to Reader: "MCSE Paths" (December 1999), but I propose the IT community go one step further and develop a fellow certification. After a set number of years of experience, an IT professional who has up-to-date certifications is either chosen by or applies to an international board for a fellow certification, a situation that is similar to training in medicine. This requirement should probably be at least 10 years of proven IT industry work experience and at least two major certifications. After a given number of years being a fellow, the IT professional can reach a fellowship status of some type. These fellows would have an association, and the big players (e.g., IEEE, Microsoft, Novell, Linux, Cisco Systems, 3Com) would hear their voices about how things really work in the trenches vs. lab production environments.
This process would curb the paper-certification phenomenon by eliminating the three problems inherent to the current certification system. Potential fellows couldn't cheat from a brain dump, attend an expensive boot camp, or take an exam seven or eight times until they memorize the questions and answers. In addition, the fellowship would be a great incentive for certified and experienced IT professionals.