Installing and Configuring Devices
If your personal work area is anything like mine, you have stacks of PC catalogs sitting on shelves next to your desk. I have thick volumes from Ingram Micro, Tech Data, and Gates Arrow, as well as books from a dozen or so other wholesalers, not to mention mounds of mailers from retailers and hundreds of spam messages every day from resellers and jobbers and manufacturers.
No book could hope to detail the installation steps for all those different components. The installation steps outlined in this topic are intended to give you an idea of how the system reacts when new devices are added and where to find the tools to configure the devices after they are installed. The troubleshooting section at the end of the chapter has hints for diagnosing and correcting problems that might arise.
If you are not running a volume license or master license of Windows Server 2003, you may encounter situations where your repair work initiates the Windows Product Activation. Ordinarily, you would only expect this to happen if you replace an entire motherboard or swap a system drive to another machine (which is essentially the same thing). It can also happen if you change a certain components. See Chapter 1, "Installing and Configuring Windows Server 2003," for more details about the devices included in the WPA evaluation.
This section includes installation steps for the following:
Using Windows Updates
Adding or changing CPUs
Adding IDE hard drives
Installing SCSI adapters and drives
Adding removable media drives
Adding network adapters and configuring bindings
Using multiple displays
Using Windows Update
Keeping up with changes to applications and drivers is an ongoing problem. Windows Server 2003 enhances the Windows Update feature introduced in Windows 2000 with a simpler interface and better support.
The most common reason for requiring a patch or hotfix is to correct a security vulnerability. There should be a lot fewer of these updates in Windows Server 2003 compared to Windows 2000 and NT, but they won't disappear. There are several ways to analyze for security deficiencies. The simplest to use is the Hfnetchk utility, branded by Microsoft as the Baseline Security Analyzer.
If you want to refresh the drivers for a particular server, the simplest thing to do is run Windows Update from the Help and Support Center (HSC). The main HSC page has a hyperlink to the Windows Update site, v4.windowsupdate.microsoft.com, which is displayed within the HSC frame. The first time you touch the site, you'll be prompted to download an ActiveX control for managing updates.
The Update site walks you through scanning for new drivers based on the devices loaded on your machine. If you want to download drivers for another user's system, follow Procedure 3.1.
Procedure 3.1 Using Windows Update to Obtain Drivers for Another System
In the Windows Update window in HSC, click Personalize Windows Update.
Select the option Display The Link To The Windows Update Catalog.
Click Save Settings.
Under See Also in the left frame, click Windows Update Catalog.
In the main window, click Find Driver Updates for Hardware Devices. This opens a page with links to various device types. Figure 3.17 shows an example.
Figure 3.17. Hardware Drivers page at the Windows Update Catalog site.
You can use this procedure to download drivers to a central location on your network where desktop administrators or users can obtain them.
If you want your servers to download updates automatically, you can configure this option in the System Properties window. Select the Automatic Updates tab (see Figure 3.18).
Figure 3.18. System Properties window showing the Automatic Updates tab and the configuration settings.
All in all, Windows Update is a simple way to keep up with fixes, especially security updates, but there is no guarantee that a particular suite of updates won't make a server unstable. I would advise doing a manual scan and selection rather than automatically doing a wholesale download.
Software Update Services (SUS)
In July of 2002, Microsoft released a new service that permits centralized distribution of hotfixes and security updates from a server internal to an organization rather than going directly to the Microsoft Update site. Software Update Services (SUS) is included in Windows Server 2003.
Using SUS has several advantages:
You can evaluate each update to ensure that it does not cause problems prior to installing it on a production server or servers.
You can deploy updates to servers that are not connected to the Internet.
You can standardize your servers by specifying only those updates that should be installed. By doing this specification at the SUS server, you centrally manage your update deployment process.
Updates can be targeted to servers via group policies.
Microsoft is continually refining SUS. Check for the latest updates and procedures at www.microsoft.com.
You should be able to keep up with security patches using either Windows Update or Software Update Services (SUS). However, you may find yourself in a situation where you want to install a particular update or hotfix individually on a server for testing or evaluation. You can do this by executing the hotfix at the console of the server.
If you want to avoid multiple reboots when locally installing more than one hotfix, use the QCHAIN utility available as a download from Microsoft.
Adding or Changing CPUs
Adding a second CPU in prior versions of Windows NT and Windows 2000 was a chore because Setup would install a uniprocessor kernel if it discovered a single processor in a machine. This was done even if the motherboard supported multiple processors.
In Windows Server 2003, all you need to do to add a second processor is put the cartridge in the machine according to the manufacturer's instructions and start up again. The system detects the additional processor and swaps out the uniprocessor system drivers with their multiprocessor equivalents.
You can verify that the system sees the second processor by opening Task Manager. The Performance window will show graphs for each processor. If the second processor is not seen, make sure it is fully seated and that no changes are required in CMOS. Also, make sure the voltage regulators or other motherboard devices are in place and firmly seated.
The situation is not so simple if you want to change the type of kernel that was loaded by Setup. If you initially install Windows Server 2003 with the ACPI kernel and later discover that the system is unstable and you want to shift to the Standard PC kernel, you must reinstall Windows Server 2003 completely.
The reason for this is that the two different kernels store Registry information differently in HKLM | Hardware. The settings are not interoperable. If you disable ACPI support in CMOS on a machine that is running the ACPI kernel, the machine will bugcheck (crash with a Blue Screen of Death).
Adding IDE Hard Drives
The most difficult part of adding a new IDE drive is getting the right master/slave jumper combinations on the devices. If the new drive displays in CMOS with the correct size and drive geometry, Windows Server 2003 should have no problem finding it using the existing IDE bus drivers. You can see the new drive using the Disk Management snap-in in the Computer Management console. Open the Computer Management console via START | PROGRAMS | ADMINISTRATIVE TOOLS | COMPUTER MANAGEMENT. See Chapter 14, "Configuring Data Storage," for the steps to configure a new disk and allocate its storage space.
If you add another drive on the same IDE controller with the intention of striping the drives with RAID 0, you will not get a performance improvement. An IDE interface can write to only one device at a time. Put the devices on separate controllers or use an ATA/IDE RAID controller. The same applies to mirroring drives, because you can slow down the operating system by forcing it to mirror between IDE drives on the same controller.
If you install ATA devices that use UltraDMA (UDMA), make sure the system detects the correct DMA setting for the device. This is displayed in the Properties window of the IDE controller in Device Manager. Select the Advanced Settings tab. The Device Type is generally locked to Auto Detection, so you will need to work with your vendor to figure out why the system is not seeing the correct UDMA version.
Installing SCSI Adapters and Drives
You can improve server I/O performance considerably by changing from IDE to SCSI. Part of this is due to the faster SCSI bus, although ATA100 and UltraDMA 5 close this gap considerably. More importantly, SCSI uses asynchronous communication so the interface can multiplex data packets to several devices. IDE/ATA controllers can only talk to one device at a time. SCSI also consumes fewer CPU cycles per megabyte transferred, so it is more efficient and therefore gives faster overall throughput.
When you install a SCSI adapter and start the machine, Plug and Play Manager detects the new adapter and loads the appropriate drivers. If PnP recognizes that new hardware has been added but does not find drivers, you will need to supply Windows Server 2003 or Windows 2000 drivers.
If the adapter hosts the drive where you want to install the operating system, and there are no drivers for the adapter in the miniature version of Windows Server 2003 loaded by Setup, you can press F6 at the beginning of the text-based portion of Setup and load the driver.
Placing a new drive in an IA64 machine that already has an operating system installed requires no action in firmware. When you use the Disk Management console in the operating system to partition the drive, the proper GPT-based data partitions are created.
If you replace the boot drive in an IA64 machine and you intend to install Windows Server 2003, you still need to take no action in firmware. Chapter 1 discusses how to use the Extensible Firmware Interface (EFI) shell to navigate to the Windows Server 2003 CD to start the operating system installation. Setup will create the necessary EFI System Partition, Microsoft Reserved Partition, and a generic data partition for the operating system.
When installing new devices on the SCSI bus, pay particular attention to SCSI IDs and terminations. Ninety percent of all SCSI problems originate from these two issues. Modern SCSI adapters are good at determining the active terminations used by devices on the bus, but it is still a good idea to mark each device clearly as to whether it has active termination enabled or disabled.
If you add a drive that changes the scan order of the bus containing the boot drive, you may end up unable to boot the machine. This is true even if the SCSI BIOS is set to boot from the correct drive. The problem occurs because the ARC path in Boot.ini references the bus devices in the order they are scanned by the SCSI controller. For instance, here is a sample ARC path:
multi(0)disk(0)rdisk(0)partition(1)\WINDOWS="Standard Edition" /fastdetect
If you change the scan order so that the boot drive is now the second (rdisk(1)) or third (rdisk(2)) drive, NTLDR will not be able to find the boot partition and you will get an Unable to Locate Ntoskrnl.exe error.
You should always low-level format new SCSI drives prior to placing them in service. This ensures that the translation table on the SCSI interface matches that used by the device. This is especially true if the drive is part of a RAID array. Some very subtle failures can occur if a drive is not matched to its controller. Low-level formatting can take quite a while, so include a couple of extra hours into your estimates for adding or replacing SCSI drives.
You can simplify analyzing problems with SCSI drives by selecting drives that support S.M.A.R.T. technology. Get details at www.pcguide.com/ref/hdd/perf/qual/featuresSMART-c.shtml and at www.storagereview.com/guide2000/ref/hdd/perf/qual/featuresSMART.shtml.
Using EFI Disk Utilities
Windows Server 2003 Setup places several disk utilities in an MSUtil folder on the EFI System Partition. You can use these utilities to create and manage drive partitions, although this is generally simpler to do from within the operating system itself rather than from the EFI shell.
These EFI-based disk utilities give you a lot of control over the disk partitioning, but that control comes at a price. You can very easily destroy one or all partitions on a disk. Chapter 2, "Performing Upgrades and Automated Installations," describes the use of the EFI Shell environment. Be very sure to use the map command in the shell to verify partition block numbers and mount aliases prior to using any of the utilities described here.
This utility operates in a similar fashion to the DISKPART utility in Windows Server 2003. You can use it to create new EFI System Partitions, new Microsoft Reserved Partitions, and new data partitions prior to installing the operating system. This is not required, because Setup is able to prepare a new drive, but the steps are included here in the event that you need to prepare a disk in a system that does not yet have an operating system loaded.
When using Diskpart.efi to create a partition, it will create the GUID automatically but you'll need to give the partition a friendly name. You'll need to specify the partition size in megabytes. For instance, size=1024 would create a 1GB partition. Follow Procedure 3.2.
SelectedDisk = 0
TypeGuid = EBD0A0A2-B9E5-4433
TypeName = MSDATA
Requested OffsetInBlocks = 0
Requested SizeInMegabytes = 1024
Attributes = 0
Resulting Size In Blocks = 2097152
Resulting Size In Bytes = 1073741824
Procedure 3.2 Using EFI DISKPART to Create a New Data Partition
Boot to the EFI shell.
Change directory to fs0:\MSUtil.
Enter List to see a list of the manageable drives in the machine.
Enter Select # where # is the ID number of the drive shown in the List command.
Enter Symbols to get a list of the partition types. There are three options: MSRES (Microsoft Reserved), EFISYS (EFI System), and MSDATA. You'll be using the MSDATA option when creating the partition.
Enter Create name=partition_name type=msdata size=1024 ver. The ver option gives a verbose listing of the result. Here is an example listing of a successful partition creation:
You can also use DISKPART to wipe a disk using the Clean command then prepare a new GUID Partition Table using the New GPT command. This results in a loss of all data on the drive, of course, so only use this option if you want to refresh a drive from another machine. For SCSI drives, you should use low-level formatting to accomplish this task and sync the translation tables between the controller and the drive at the same time.
This utility has one function. It creates an EFI System Partition on a blank disk. If you use it on a disk with existing partitions, you will lose those partitions and any data they contain.
This utility is intended solely to format the EFI System Partition created by Fdisk.efi. The only available format options are FAT and FAT32. Before using FORMAT, you must first mount the partition created by FDISK. Use the Mount command for this—for example, mount blk4 fs8.
After the partition is mounted, do a quick format using the Format command as follows: format fs8 /fs:fat /q.
If you accidentally overwrite or delete the boot menu entry in NVRAM that points at the Windows Server 2003 operating system partition, you can use Nvrboot.efi to recover the original boot menu entry. You can also use this utility as an alternative to the EFI Boot Menu Management utility to modify the entries in the boot menu.
If you want to modify the contents of the boot menu entries themselves, Nvrboot only works with entries installed by Windows products. Other entries, such as the EFI Shell entry and the PXE boot entry, can only be managed by the Boot Menu Management utility.
Nvrboot has these options:
This option shows the full NVRAM entry for the boot item.
This option permits you to change the LoadIdentifier and OsLoadOptions variables in the entry. The two load paths cannot be modified.
This option permits you to make a backup copy of one or all NVRAM options. This is a terrific way to keep a copy of the entries on-hand in case something goes wrong. You can put a copy of the exported file on a floppy then copy the file back from the floppy in the event that the NVRAM entries are overwritten.
This option permits you to import one or more NVRAM entries from a file into NVRAM.
Use this option to make a copy of an entry prior to making changes to the main entry.
Use this option to delete an entry you no longer want displayed on the boot menu. Use caution. The system will permit you to erase all entries by entering an asterisk (*).
Adding Removable Media Drives
Windows Server 2003 has a new service called the Removable Storage Manager (RSM). The RSM service simplifies managing removable media if you have large CD-ROM jukeboxes or a robotic tape library, but it can make simple interfaces such as small CD disc changers more difficult to handle.
In classic NT, each disk in a jukebox or disc changer got a separate drive letter. This caused much frustration when configuring big CD libraries because there aren't enough letters to accommodate a 100-disc library. RSM solves this problem by assigning a single drive letter to the device itself and manipulating the CDs in the background.
But for small 4x and 6x CD changers, the separate drive letters actually added to the convenience of using the product. A user who wants a particular CD can go to a drive letter and find it. Higher-end changers cache the CD contents so that the drives can be browsed quickly without causing the changer to move them around. This is not possible with RSM. Each disk must be mounted by name using the RSM Mount command. This requires that you know the logical media ID assigned to the CD by RSM. This information is available in the RSM snap-in within the Computer Management console.
Removable media drives come in three general types:
Fast oversized floppy disk drives:
Zip, Sony, and so on
Slow undersized hard disk drives:
Jaz, Shark, and so on
Fussy plastic burners:
CD-R and CD-RW drives and magneto-optical drives
Most of these drives have parallel port and SCSI models. All major vendors have models with Windows 2000 drivers, which should work with Windows Server 2003. The Plug and Play Manager detects the drive during enumeration and loads the appropriate driver. It then turns control over to RSM. The drive is presented in Explorer with an icon that represents a removable media drive.
If you have a SCSI removable media drive, be sure to select a SCSI ID for it that doesn't conflict with other devices on the bus. Also, remember that the results of the SCSI bus scan determine the relative disk location for the ARC path in Boot.ini. If the removable storage device has a SCSI ID lower than that of the primary drive, it inserts itself into the scan list ahead of the boot drive and you get an Unable to Locate Ntoskrl.exe error.
Media in removable media drives cannot be partitioned. This includes LS120/240 superfloppies, DVD-RAM drives, Castlewood Orb drives, and other removable media drives with large capacities.
Drive letter assignments can become a challenge when you have removable media drives on a server. During initial Setup, drives and drive partitions are scanned in this order:
System partition on the boot drive is assigned drive C.
Boot partition, if in a different partition, is assigned drive D.
Logical drives in extended partitions and removable media drives come next.
Basic partitions on fixed drives come last.
This scan order means that removable media drives will get letters ahead of partitions on fixed drives, which you may not prefer. You can change the drive letter after Setup has completed.
If you install a removable media drive after initial setup, it will not disturb the existing drive letters. It adopts the next available drive letter starting from the top of the alphabet. For example, if you have three existing partitions lettered E, F, and M, the removable media drive will take drive letter G.
If you have two or more removable media drives, you can sometimes cause problems because they can swap drive letters if you disable and reinstall them.
If you have a network drive that uses the first letter after the existing hard drives, the removable media drive steps on the network drive and takes the letter. LanMan Server attempts to use the letter then displays an error saying The Local Device Name Is Already In Use. You should configure the system to break this network connection and remake it using another letter.
Adding Network Adapters
Even if you disregard all the other features in PnP, the way it simplifies adding and configuring network adapters makes it worth the price of admission.
When selecting server adapters, stick to high-end models from reputable manufacturers. Microsoft has dropped support for many desktop-quality and whitebox network adapters. Also, you may get performance improvements if the adapter supports Network Driver Interface Specification (NDIS) 5.0 offload capabilities. See www.microsoft.com/hwdev/tech/network/taskoffload.asp for more information.
The NDIS drivers that support Windows networking have a well-deserved reputation for being difficult to interface with. PnP, coupled with a properly prepared INF and a driver written by someone who really knows how to code a WDM device driver, makes installing a new network adapter a brief and event-free operation.
Plug and Play Manager discovers the new device during bus enumeration after you start the machine. If you have a server that permits hot-swapping of PCI adapters and a network adapter that is hot-swap enabled, PnP will see the device as soon as you release the software lock on the PCI bus.
PnP assigns resources to the device based on information it obtains from ACPI, which obtained the information from the PCI header of the device. PnP will also discover legacy network adapters, although these are becoming rare and should be avoided. If resources claimed by a legacy device conflicts with a PnP device, the PnP Manager will attempt to assign a different resource to the PnP device. Errors in the Event log will warn you of these sorts of conflicts.
By default, new network interfaces are configured to use Dynamic Host Configuration Protocol (DHCP). To change the TCP/IP settings to static mappings, or to install additional transports, clients, or network services, open the Properties of the connection from the Network Connections windows.
If you do not have a DHCP server, the new interface will obtain a network address using Automatic Private IP Addressing, or APIPA. This mechanism assigns an address from the 169.254.0.0/16 address space. Microsoft owns this address space and has assigned it for this purpose.
An APIPA client selects a random address from the 65,535 it has available in this space, then pings the address to see if some other client got there first. It makes three attempts before giving up and putting an error on the console. After the client has obtained an address via APIPA, it continues to look for a DHCP server once every five minutes.
If you configure two LAN interfaces to talk to the same network segment, you will get a Duplicate Computer Name error. This is because the second adapter will attempt to register its NetBIOS name after the first adapter has already laid claim to it. When installing multiple adapters, you should use either different IP subnets or the Network Bridge option.
Diagnosing Network Problems
The Windows Server 2003 Task Manager has an additional tab for monitoring network interface traffic. This provides a quick-and-dirty check on interface performance.
The interface status (available from the Interface icon in the Notification Area or from the Network Connections window) has a Repair option that gets a new IP address from the DHCP server (if applicable) and performs other routines designed to fix network communications.
If you have previously uninstalled a network adapter on a server, you may get an error message when you attempt to reinstall another adapter. This is because the interface has not been completely removed from the system but might be hidden from the Device Manager. Open the Device Manager console via Devmgmt.msc and show the hidden devices via View | Show Hidden Devices. Then remove the outdated network device.
When problems get trickier, one handy way to get a quick look at all the network settings for a server is to use the Network Diagnostics window in Help and Support Center. This support page uses WMI parameters to show you the current status of the network interfaces on a server.
There is also a valuable command-line tool, NETDIAG, that performs the same checks and reports them to the console. This is handy for creating scripts and batch files that can take periodic system snapshots.
A new switch was added to the Netsh command in Windows Server 2003 that also helps in troubleshooting. It is the -diag switch. Using it in conjunction with -diag show, you can get a quick listing of current settings and statistics for all network interfaces and network services such as DHCP and WINS.
Using Multiple Displays
If you are a CAD operator or graphic designer, you probably either have a big 21-inch monitor or dual monitors, or both. Both Windows 2000 and Windows Server 2003 support dual displays, but Windows Server 2003 and XP also support DualView adapters, which permit putting different portions of the same image on the two screens.
Only a limited number of vendors produce DualView-compatible adapters. On desktops, the Matrox "G" line includes its Dual Head technology, which can deliver full-frame DVD in one monitor and business graphics in the other, just in case you want to watch Shrek while you do spreadsheets. On laptops, Trident Microsystems and SiS are the major suppliers of DualView video controllers.
If you are shopping around for a second video adapter to put in a system that already has an Accelerated Graphics Port (AGP) adapter, be cautious about using the same model PCI adapter because the drivers are different but might have the same name.
You can use the same steps to configure a system with two video adapters or one with a single, DualView adapter and two monitors. The monitors are configured in Display Properties, which is most easily accessed by right-clicking anywhere on the desktop and selecting PROPERTIES from the flyout menu. The Settings tab shows the two monitors.
The Primary display determines which monitor will display the START menu and status bar. This does not make changes to the underlying system BIOS, so you may see the other monitor display the POST results and initial text phase portion of the Windows boot sequence.
The default orientation puts the logical monitors side by side. You can move the display monitors to change that orientation. You can select different resolutions and refresh rates, but this might cause images to become distorted as you move them between monitors.