Troubleshooting Computers

This page is for frustrated computer users. Get any ideas to fix your computer?

Sound: Frustrated woman

Topics this page:

Start or Stop, Haiku Error Messages 
Your comments???

Site Map  
About this site  

The Windows Boot Sequence

After power-up, the BIOS goes thru a POST (Power on Self Test). The BIOS reads the master boot record (MBR)-the first sector on the first hard disk. The MBR reads the boot sector —the first sector of the active partition.

Then, Windows goes through this boot sequence:

1. Ntldr loads the Advanced RISC Computing (ARC) pathname specified in Boot.ini to find the boot partition.
2. Kernel Load: The “Starting Windows” progress bar is shown at the bottom of the screen when Ntldr loads the Windows 2000 OS kernel: Ntoskrnl.exe and
3. HAL.DLL Hardware Abstraction Layer for the processors (I386 or Alpha) Machines with more than one CPU run a different HAL kernel. Uniprocessor computers contain simpler PIC (Programmable Interrupt Controller) hardware, referred to as a “Standard PC” in Device Manager.
4. SYSTEM configuration settings for devices and services
5. Service startup: When a window with Please Wait in the title bar, Ntoskrnl.exe is loading the Session Manager (Smss.exe), which in turn loads the high-order services-those necessary for a user to log on.
6. Windows loads into memory generic ATAPI drivers to read from most CD's and DVD's.
7. Ntoskrnl.exe takes over and launches Winlogon.exe, which in turn starts Lsass.exe (Local Security Administration), the program that displays the Welcome To Windows dialog box.
8. When you see the desktop, you should hear the “Microsoft sound” (composed by Brian Eno of U2 fame). If you don't check your cables, external speakers, etc.
9. Applications assigned to a computer are installed automatically before user logon.
10. User Logon

Troubleshooting Process Steps

Collect (1. detect a failure & 2. diagnose cause)

1. Identify the caller. If you're on the phone, get the user's name and phone number where he or she is calling from. Write this down on a template.
2. Identify the context. Some people are better than others at summarizing their problem. What machine. How long they've had the machine.

   Describe the approach you're taking. Get permission to proceed.

3. Identify the symptoms. Tip: Ask open-ended questions to help identify symptoms. Instead of asking “can you log in to the network?”, ask “what happens when you try to log in?”

   Use a systematic approach to eliminate possible causes. Some start with the most obvious cause and work back through other causes. But I personally think that it's better to avoid jumping to conclusions with an accusatory tone and work to layers of the system architecture.

   Prioritize the likely causes of the problem and the solutions for each. Prioritize based on: relative ease of solution, time required to implement solution, likelihood of success, and cost of solution.

   Swap parts out.

4. Re-create the problem. Start from the source (of power and information), such as whether each component is plugged in or input data is actually available.
5. Isolate the cause. To prevent the problem from re-occuring in the future.

   Never blame the user for the problem, even if it's really his/her fault. The goal is to have a happier, more productive user at the end of the call.

3. Determine possible solutions

6. Formulate a correction. Depending on the cause you've isolated, identify all the possible corrective actions before taking action. Confer with colleagues to come up with possible corrections. This prevents another call on what seems like a similar situtation.

   For example, if you've isolated the cause as a corrupt spreadsheet file, one correction would be to restore the file from a backup. Review other spreadsheets that may also be corrupted.

   Consider both costs and benefits of each possible action. Identify side-effects of each potential action. Prioritize possible actions according to their likelihood of success and ease of implementation.

4. Take Corrective Action

7. Implement the correction. This step is where you actually fix the problem. Before you do so, however, you need to ensure that productivity doesn't suffer and that downtime is minimized. For example, you might need to provide a "loaner" workstation to a user whose machine needs to be rebuilt.
8. Verify solution. Make sure that the solution that you've implemented actually solved the problem and didn't cause any new ones. Use several options and situations to conduct your tests; for instance, try the task yourself, then have the user try the task while you observe the process, or test the workstation both before and after it's connected to the network. Sometimes you'll need to test over time to ensure that your solution is the correct one. Remember to verify that the user agrees that the problem is solved before you proceed.
9. Document the problem and the solution. Take notes as you go so people won't have to say “I already told you”
10. Follow-up promptly. Make sure that promises are kept.
11. Verify user satisfaction. Ask whether the person feels the problem is solved to his/her satifaction.

I'm sorry, Dave. I'm afraid I can't do that. —Computer HAL in Arthur C. Clarke's film 2001: A Space Odyssey

“Prescription without Diagnosis is Malpractice”

“For Every Obstacle There is a Solution - over, under, around, or through" —Dan Zadra

Results! Why, man, I have gotten a lot of results. I know several thousand things that won't work.
—Thomas Edison

Diagnostic Tools

• Microsoft's Diagnostic Solution Guide for Windows NT and 2000 helps you find Technet articles.

• Internetweather.com illustrates the bottlenecks causing your ISP to have routing problems in California

• HP's user forums is where HP engineers and other users gather to solve problems, exchange ideas, and learn lessons related to HP's products (servers, peripherals, OpenView, etc. for Microsoft, Linux, and HP-UX)

Windows 2000 Debugging

Proactive Steps for a rainy day

• Store your data on a different physical hard drive than the one used by the operating system.
• Copy the Windows 2000 Install CD onto a hard disk.
• Create a new Emergency Repair Disk (ERD) after every time you apply a Windows Service Pack.
• Backup data to a different media, before you install or upgrade a new application
• Use a (RAID-1) mirror drive for redundancy on the boot partition.
• Install Recovery Command Console [Q216417] so that it's an option presented from Boot.ini. Example:
  X:\ENGLISH\WIN2000\PRO\I386\winnt32.exe /cmdcons

  Stuck in boot-up Text screen

• Boot-up from CD-ROM or diskette created from Windows installation CD's.
• ntdetect.com
• Press F6 and load third-party SCSI drivers

  Responses to problems

• Look at the ntldr.txt
• If your system does not boot, at the black screen, press F8 and choose Last Known Good
• If your system boots, boot up in Safe Mode to start the operating system with basic drivers and services

Safe Mode Advanced Options

• Enable Boot Logging
• Enable VGA mode
• Directory Services Restore mode
• Debugging mode

Uses for Safe Mode

• Incompatible driver or setting for the video card
• Corrupted Windows Explorer shell
• Run msinfo32

  In the Recovery Console

• Almost all previous DOS commands are available.
• FIXBOOT to restore a damaged boot sector.
• FIXMBR to repair a Master Boot Record.
  Or from a boot up disk: FDISK /MBR
• Replace damaged system files, drivers, and registry hives from floppy or CD.
• Remove Active Directory from Domain Controller
• For security reasons, the Recovery Console allows reading from but NOT writing to diskettes.

Understanding Windows Safe Mode options, Aug 21, 2001

Windows 2000 Installation lists the proactive steps to make troubleshooting easier.

The Windows SDK provides the WDEB386 program for Kernel Debugging (WKD). 386SWAT is a similar tool. They use a low-level debugging interface to obtains control early in the Windows startup sequence. During Windows startup, while still in RM/VM, Windows tells the WKD as it loads the real mode portion of all of the VxDs to be used during a session.

MSDN subscribers receive Checked builds that contain separate copies of various system DLLs containing more messages and checks on Windows execution.

Make Network Boot Diskette

You'll need a Windows 2000 bootup diskette access to recover from a failed system volume that is mirrored. This takes the place of your system partition. After formatting the diskette with “/S” system files, copy these files to the diskette:
• The Boot.ini text file displayed by the Boot Loader menu during boot-up. It identifies the physical location of the operating system files. So, it must adhere to ARC Path conventions Some values in this file can be changed from GUI: Computer properties, Advanced tab, “Startup and Recovery” dialog.
• The Ntldr which reads Boot.ini. It displays the “Starting Windows” progress bar at the bottom of the screen. It calls file Ntoskrnl.exe, the Windows 2000 OS kernel that manages all remaining boot functions. Note: Ntldr takes the place of file BOOTSECT.DOS which loads non-Windows 2000 OSs. Ntldr uses interrupt (INT) 13 BIOS calls to locate Ntoskrnl.exe and other files that it loads at startup.
• NTDETECT.COM, which builds a (“Plug and Pray”) hardware list for Ntldr to add to the Registry.
• NTBOOTDD.SYS, a device driver for SCSI disk controllers if BIOS SCSI is disabled.
• If you want to make a network boot-up diskette [Q142857]
  • Create on the diskette a Net folder
  • Copy the driver file (such as epos.dos), System.ini, and Protocol.ini files obtained from the manufacturer of your NIC card. They may also be in the Server installation CD folder Clients\Wdl\Update
  • Make sure the Protocol.ini file refers to the name of the driver file. For example, if driver file Epos.dos is used, the file should contain “[EPRO$]” and “drivername=EPRO$”.
  • Make sure the System.ini file refers to the name of the driver file. For example, if driver file Epos.dos is used, the file should contain “netcard=Epro.dos” under the “[network drivers]” section.

Troubleshooting Windows under Safe Mode

Disk Management

Type		Technology	Purpose	Limits	Recovery Process
Basic (from NT4)	-	-	No fuss	-	Restore from backup
	RAID-1	disk mirroring sets	Redundancy (faster parallel data transfer)	50% overhead. Can't be extended.	Resynchronize Mirror. Regenerate Mirror.
Dynamic (Windows 2000 only)	Simple	-	Fast	-	Restore from backup
	RAID-0	disk striping volumes. No redundancy / fault-tolerance.	Faster transactional access.	2-32 drives	Restore from backup. Create a new striped volume.
	RAID-1	disk mirroring volumes	Redundancy	50% overhead. Can't be extended.	Reactivate Mirror. Remove mirror, replace hardware, Add Mirror. Regenerate Mirror.
	RAID-3	spanning extensions across disks (Volume sets)	Bulk transfer of big files spread across several hard disks	Coordinated transactional reads slow	Restore from backup
	RAID-5	striped volume with parity	Up-time and Redundancy	-	Reactivate Disk, Repair Volume

RAID-6 protects data from the failure of two concurrent hard drives (such as during a rebuild) — an increasingly more likely with larger disk array sizes, increasing disk densities, and use of Serial ATA (SATA) drives (which some say have lower reliability). RAID 6 can rebuild bad data blocks discovered during RAID rebuilds because it stores a second independent distributed parity stripe. Until recently, it been available only in proprietary devices.

About FAT

In Windows 2000, the Disk Management MMC is used for both basic and dynamic disks. Dynamic disks are unique to Windows 2000. RAID sets upgraded from NT4 remain as "basic disks". So a new basic drive is needed to repair such sets. New RAID sets must be dynamic.

Only basic disks can be bootable or partitioned (to a limit of 4) -- a legacy of NT4. Dynamic disks (not available on removeable media). "Revert Back to Basic Disk" requires a backup, disk deletion, and restore.

NTFS Formatting: Only NTFS supports extending volumes. Logical drives and System and boot partitions are simple volumes in Windows 2000. However, Windows 9x on the same machine can't read dynamic disks.

After installing hot swap hardware, click Action and Rescan Disk.

Managing Windows 2000 Disks, Backup and Restore

Disk Recovery

Solve Disk Problems Before Volume

On mirrored (RAID-1) drives:
1. When the status of a disk is Online(Errors) it indicates I/O errors have occured. If any disk within a volume is this way, the volume has a status of Healthy(At Risk).
2. To fix this, right-click on the disk to Reactivate Disk.
3. If this doesn't return the disk to Online and the volume to Healthy, replace the drive and Reactivate Volume or, for fault-tolerant volumes, Reactivate Volume.

If a RAID-5 fails, don't Delete Volume or you'll delete the entire RAID array and its data.

1. After physically replacing the drive,
2. Rescan Disks to write a signature on the drive.
3. Next, upgrade the basic disk to a dynamic disk so that its status becomes Unknown.
4. Next, Repair Volume to change the status to Regenerating during formatting.
5. Finally, logically Remove Disk on the failed disk.

Proactive Coding Practices

What programmers can do to prevent problems:

Differentiate the type of messages issued:

• Notification of what triggered the message (what happened).
• Explanation of what was expected.
• Solution - the corrective action to correct the root cause of the problem.

  Store detected errors for future analysis

• Store detected errors with sufficient detail:
  • The caller or thread ,
  • ADO error collection HRESULT structure defined in Winerror.h and translated through FormatMessage() API.
  • Designate a Severity level:
    • SQL_SUCCESS_WITH_INFO
    • SQL_ERROR
    • Fatal error which causes disconnection between client and sever.
• Store detected errors in a central location (across different releases, users, etc.) to correlate related errors.
• Store detected errors in the Windows Application Event Log so that system (COM+ or network) errors are in the same place.
• Store SQL Server database to correlate the application's other data.

  Logic:

• Make sure that errors in a sub-module are passed up to the calling module.

Developing Windows Error Messages by Ezzell, Ben. O'Reilly and Associates, 1998.

Instead of a simple "Syntax error", this author suggests a less ambiguous message text:
Syntax Error - invalid (non-numeric) characters found in entry: A numerical integer is expected here. You may enter a positive or negative value but your entry must not contain punctuation or alphabetical characters other than a plus (+) or minus (-) sign. To correct this error, re-enter the number in the highlighted cell.

Windows DNA Error Handling 101 by Paul Bates

Windows 2000 Error Messages

Code Complete

Jim Carlton reports on stupid computer stories in the Wall Street Journal

Computer Problem Report Form

From Cyberhighway Humor

Your Name: __________________________ Your Login Name: ____________________

The date? __/__/__  The date the problem first occurred if different? __/__/__

1. Describe your problem:
   __________________________________________
 
2. Now, describe the problem accurately:
    __________________________________________
 
3. Speculate wildly about the cause of the problem:
    __________________________________________
 
    __________________________________________
 
4. Problem Severity:
    A. Minor__
    B. Minor__
    C. Minor__
    D. Trivial__
 
5. Nature of the problem:
    A. Locked Up__
    B. Frozen__
    C. Hung__
    D. Shot__
 
6. Is your computer plugged in? Yes__ No__
 
7. Is it turned on? Yes__ No__
 
8. Have you tried to fix it yourself? Yes__ No__
 
9. Have you made it worse? Yes__
 
10. Have you read the manual? Yes__ No__
 
11. Are you sure you've read the manual? Yes__ No__
 
12. Are you absolutely certain you've read the manual? No__
 
13. Do you think you understood it? Yes__ No__
 
14. If `Yes' then why can't you fix the problem yourself?
     _________________
 
15. How tall are you?  Are you above this line? __________________
 
 
16. Is the equipment unexpectedly noisy?    Yes___  No ___

17. If 'Yes", what sort of noise?
     Grinding ___    Rattling ___
         Whirring ___    High Pitched Whine ___
     Sound of disk head scouring disk ___
     Strange, out of tune whistling or humming ___
         
18. Is there a smell of burning?            Yes___  No ___

19. If "Yes", is the equipment on fire?     Yes___  No ___

20. Is the fault repeatable?                Yes___  No ___
 
21. What were you doing with your computer at the time the problem
     occurred?
     ______________________________________________________
 
22. If `nothing' explain why you were logged in.
     ________________________________________________________________
 
23. Are you sure you aren't imagining the problem? Yes__ No__
 
24. How does this problem make you feel? ____________________________
 
25. Tell me about your childhood. ___________________________________
 
26. Do you have any independent witnesses of the problem? Yes__ No__
 
27. Which are appears to be at fault?
        Communications ___      Disk ___        Base Unit ___
        Network ___             Keyboard ___    Screen ___
        Mouse ___               Everything ___  Don't Know ___
 
28. Can't you do something else, instead of bothering me? Yes_

Is Your Computer a Male or Female?

As you are aware, sailing ships have long been characterized as being female (e.g. "Steady as she goes", or "She's listing to starboard, Captain").

Recently, a group of computer scientists (all male) announced that computers should be referred to as female. Their reasons for drawing this conclusion follow:

Five reasons to believe computers are female:

1. No one but the Creator understands their internal logic.
2. The native language they use to communicate with other computers is incomprehensible to everyone else.
3. The message "BAD COMAND OR FILE NAME" is about as informative as, "If you don't know why I'm mad at you, then I'm certainly not going to tell you."
4. Even your smallest mistakes are stored in long-term memory for later retrieval.
5. As soon as you make a commitment to one, you find yourself spending half your paycheck on accessories for it.

However, another group of computer scientists (all female) think that computers should be referred to as male. Their reasons follow:

Five reasons to believe computers are male:

1. They have a lot of data, but are still clueless.
2. They are supposed to help you solve problems, but half the time they are the problem.
3. As soon as you commit to one you realize that, if you had waited a little longer, you could have obtained a better model.
4. In order to get their attention, you have to turn them on.
5. Big power surges knock them out for the rest of the day (or night).

“To err is human. To really foul things up requires a computer.”

This sort of thing has cropped up before....and it has always been due to human error. —Computer HAL in Arthur C. Clarke's film 2001: A Space Odyssey