Like all MMC snap-ins, the Disk Management snap-in can be opened in a number of ways. One of the most direct ways is to right-click the My Computer icon in the upper left corner of your desktop and choose Manage from the shortcut menu. This opens a local version of the Computer Management snap-in, containing the System Tools, Storage, and Services and Applications snap-ins. Click Storage to access the Removable Storage (if you have removable storage drives installed), Disk Defragmenter, Logical Drives, and Disk Management snap-ins, and then click Disk Management to open the Disk Management window. To open the Disk Management snap-in only, double-click the file Diskmgmt.msc in the %WinDir%\System32 directory. Note that the .MSC extension is probably hidden unless you've changed the default options in Windows Explorer.
When you open the Computer Management snap-in, you have the ability to manage not only the resources of the local computer but also those of remote computers. This makes it easy to manage the disks on a remote computer. If you run the Disk Management snap-in only, you'll be limited to managing disks on your local computer, unless you create a custom MMC. See Chapter 10 for information on creating and customizing MMCs.
Adding a new drive or partition to a Windows 2000 server is straightforward. First, obviously, you need to physically install and connect the drive. If you have a hot-swappable backplane and array, you don't even have to shut the system down to accomplish this task. If you're using conventional drives, however, you'll need to shut down and power off the system.
Once the drive is installed and the system is powered up again, Windows 2000 automatically recognizes the new hardware and makes it available. If the disk is a basic disk that is already partitioned and formatted, you'll be able to use it immediately. If it's a brand new disk that has never been partitioned or formatted, you'll need to prepare it first. If it's a dynamic disk or disks but is from another machine, you'll be able to use it as soon as you import it. If the disk is a basic disk that has already been formatted, you won't get prompted to upgrade it to a dynamic disk, but you should do it anyway. If the disk has never been used, you'll be prompted by the Write Signature and Upgrade Disk Wizard.
If you install a new hard drive it is recognized automatically and the Write Signature and Upgrade Disk Wizard starts automatically when you open Disk Management. To add a new disk when the logon is complete, follow these steps:
Figure 15-3. The first screen of the Write Signature and Upgrade Disk Wizard.
Figure 15-4. The Select Disks To Upgrade screen of the Write Signature and Upgrade Disk Wizard.
Figure 15-5. The main Disk Management console, showing the new disk (Disk 5).
To create a new volume (the dynamic disk equivalent of a partition), complete these steps:
Figure 15-6. The first screen of the Create Volume Wizard.
Figure 15-7. Select the type of dynamic volume you want to create.
Figure 15-8. Select the dynamic disks that will be part of this volume.
Figure 15-9. Select a drive letter or mount point for the new volume.
Figure 15-10. Set the formatting options for the new volume.
Figure 15-11. The new RAID volume being generated and formatted.
Real World
Mounted Volumes
Windows 2000 borrows a concept from the UNIX world by adding the ability to mount a volume or partition on a subfolder of an existing drive letter. A mounted volume can also have a drive letter associated with it, although it does not need to, and it can be mounted at more than one point, giving multiple entry points into the same storage.
A volume must be mounted on an empty subfolder of an existing NTFS volume or drive. FAT and FAT32 drives do not support mounted volumes. You can, however, mount a FAT or FAT32 volume at any mount point. You can mount only a single volume at a given mount point, but you can then mount further volumes on top of an existing mounted volume, with the same rules and restrictions as any other mount. The properties of a drive do not show all of the available disk space for that drive, as they will not reflect any volumes mounted on the drive.
Mounted volumes can be used to provide a mix of redundant and nonredundant storage in a logical structure that meets the business needs of the enterprise while hiding the complexities of the physical structure from the users.
You can create partitions only on basic disks, not on dynamic disks. To create a new partition, follow these steps:
Figure 15-12. The first screen of the Create Partition Wizard.
For security and performance reasons, data files should not be on the same partition as Windows. In particular, make sure the Internet Information Services (IIS) publishing folder (\Inetpub) is not in the same partition.
Figure 15-13. Select the type of partition you want to create.
Figure 15-14. Specify how much of the disk will be used by this partition.
Figure 15-15. Select a drive letter or a mount point for the new partition.
Real World
Formatting Options
Windows 2000 supports three different file system formats: FAT, FAT32, and NTFS. Before any disk or volume can be used, it must be formatted. For disks larger than 510 MB, you will probably find that FAT32 or NTFS use space most efficiently. However, only NTFS supports the more advanced features of Windows 2000.
You can choose to quick-format a drive to make it available more quickly, but this option simply removes the file entries from the disk and does no checking for bad sectors. You should select this choice only when recycling a disk that has already been formatted and when you are confident that it hasn't been damaged.
On an NTFS volume or partition, you can specify the allocation unit size. This option lets you tune the disk for a particular purpose, depending on the disk's size and intended function. A database storage volume that will contain large database files that are managed by the database program might lend itself to large allocation units (also called clusters), whereas a disk that must hold many small files is a candidate for smaller clusters. However, the default sizes are an excellent compromise for most situations and should be modified only with caution and a clear understanding of the consequences for your environment.
You can also choose to enable disk and folder compression on NTFS volumes and partitions. This causes all files and folders on the volume to be compressed, as opposed to individual files or folders that you select. Compression can minimize the amount of hard disk space used by files but can have a negative impact on performance, especially on busy file servers. Given the cost of hard drive space today, using compression is a false economy, particularly for frequently updated data.
If you've created a new extended partition, the next step is to create logical drives in the partition. You can assign one or more logical drives in an extended partition, and each of those logical drives can be assigned a drive letter and/or one or more mount points. Each of the logical drives can be formatted with any of the supported file systems, regardless of the format of other logical drives. To create a logical drive, follow these steps:
Figure 15-16. The Select Partition Type screen of the Create Partition Wizard.
Figure 15-17. Adjusting the size of the logical drive.
Figure 15-18. Assigning a drive letter or mount point for the logical drive.
Deleting a partition, deleting a logical drive, and deleting a volume are essentially the same task, with one important exception. When you delete a logical drive, you end up with free space on the partition, but other logical drives on the partition are untouched. When you delete a partition or volume, the entire volume or partition is deleted. You cannot, however, delete an extended partition until all of the logical drives in the partition have first been deleted. You can directly delete a primary partition or a volume.
In all cases, when you delete a volume, logical drive, or partition, you'll end up with free or unallocated space and no data on the volume, drive, or partition when you're done, so make sure you've got a good backup if there's a chance you might later need any of the data. To delete a partition, logical drive, or volume, follow these steps:
Figure 15-19. Confirmation message for deleting a volume.
Once the volume or partition has been completely deleted, the space it occupied is unallocated. Space that is unallocated on dynamic disks can be used to create mirrors, extend an existing volume, create a RAID array, or otherwise manage the storage on your server. Space that is unallocated on basic disks can be partitioned.
The advantages of dynamic disks are substantial. Even if you use hardware RAID controllers and hot-swappable disks to manage your hard disks, you'll probably find it a good idea to use dynamic disks. There is a caveat, however. Because you can't boot from or even see a dynamic disk from another operating system, and even Windows Setup doesn't have complete support for dynamic volumes, you might want to consider leaving at least your boot drive as a basic drive. Doing so makes working with it somewhat easier. If you need to provide for redundancy on that drive, and if hardware RAID is an option, use RAID level 1 to make recovery from a failed hard disk or other disaster as painless as possible. To convert a basic disk to a dynamic disk, follow these steps:
Figure 15-20. You can select more than one disk to upgrade.
If there are any open files on the disk to be upgraded, you might experience data loss. You should only perform disk upgrades during quiet times when no users are logged on or using the server.
You can add space to a volume without having to back up, reboot, and restore your files if the volume is on a dynamic disk and if it is a simple volume or a spanned volume. You do this by converting the volume to a spanned or extended volume that incorporates unallocated space on any dynamic disk. Unfortunately, you can't increase the size of a RAID-5 or RAID-0 (striped) volume simply by adding disks to the array, unless you're using a version of hardware RAID that supports this functionality. To extend a volume, follow these steps:
Figure 15-21. Selecting the disks to use to extend the volume.
It's important to remember that a spanned (extended) volume is actually less reliable than a simple disk. Unlike a mirror or RAID-5 volume, in which there is built-in redundancy, a spanned or striped volume will be lost if any disk in the volume fails.
Real World
Extending—Administrator's Friend or Foe?
Most administrators have wished at some point that they could simply increase the users' home directory space on the fly without having to bring the system offline for several hours while the entire volume is backed up and reformatted to add the other hard disks, the backup is restored, and the share points are re-created. Fun? Hardly. Risky? Certainly. And definitely a job that means coming in on the weekend or staying late at night—in other words, something to be avoided if at all possible.
All this makes Windows 2000's ability to create additional space on a volume without the need to back up the volume, reformat the disks, and re-create the volume a seductive feature. However, if you're using conventional hard disks without hardware RAID, you might want to think twice before jumping in. Only dynamic volumes allow you to add storage on the fly, and only if the volume is either a spanned or striped volume, or a simple volume that was created as a dynamic volume (not upgraded). Because spanned volumes provide no redundancy, using them exposes your users to the risks of a failed drive. Yes, you have a backup, but even under the best of circumstances, you'll lose some data if you need to restore a backup. Further, using spanned volumes actually increases your risk of a hard disk failure. If any disk used as part of the spanned volume fails, the entire volume is lost and must be restored from a backup.
Why, then, would anyone use spanning? Because they have hardware RAID to provide the redundancy. This combination offers the best of both worlds—redundancy provided by the hardware RAID controller and flexibility to expand volumes as needed, using Windows 2000 disk management. This is yet another compelling argument for hardware RAID, if you need one.
Windows 2000 uses the terms "extended" and "spanned" nearly interchangeably when describing volumes. Technically, however, a spanned volume must include more than one physical disk, whereas an extended volume can also refer to a volume that has had more space added to the original simple volume on the same disk.
When your data is mission-critical and you want to make sure that no matter what happens to one of your hard disks the data is protected and always available, you should consider mirroring the data onto a second drive. Windows 2000 can mirror a dynamic disk onto a second dynamic disk so that the failure of either one does not result in a loss of data. To mirror a volume, you can either select a mirrored volume when you create the volume (see "Creating a Volume," earlier in this chapter) or you can add a mirror to an existing volume. To add a mirror to an existing volume, follow these steps:
Figure 15-22. The Add Mirror dialog box.
Regeneration is both CPU- and disk-intensive. When possible, you should create mirrors during slack times or during normally scheduled downtime. This goal should be balanced, however, by the equally important goal of providing redundancy and failure protection as expeditiously as possible.
Figure 15-23. A newly created mirrored disk in the process of regeneration.
To improve your overall data security and reliability, you should mirror your volumes onto disks that use separate controllers whenever possible. This process is known as duplexing, and it eliminates the disk controller as a single point of failure for the mirror while actually speeding up both reading and writing to the mirror, as the controller and bus are no longer potential bottlenecks.
If one of the disks in a mirrored volume fails, you'll continue to have full access to all of your data without loss. Windows 2000 sends an alert to the console (Figure 15-24), marks the failed disk as missing, and takes it offline, but it continues to read and write from the other half of the mirrored volume as though nothing had happened. Be warned, however: you no longer have any fault tolerance on that volume, and any additional failure will result in catastrophic data loss.
Figure 15-24. The warning message that displays when a disk that's part of a mirrored volume fails.
Once you've replaced the failed disk or corrected the problem and reactivated it, the mirror automatically starts regenerating. If the problem can be solved without powering down the system, you can regenerate the mirror on the fly. To reactivate the failed disk, follow these steps:
Figure 15-25. Reactivating a failed disk that's part of a mirrored volume.
Figure 15-26. Data being regenerated on a reactivated mirrored disk.
If you need to make additional disk space available on your system and you have no additional disks available, you can remove the mirror from a mirrored volume. When you remove a mirror, the data on one of the disks is untouched, but the other disk becomes unallocated space. Of course, you will have lost all redundancy and protection for the data, so you'll need to take steps to restore the mirror as soon as possible, and until then you might want to modify your backup schedule for the remaining disk. To remove a mirror, follow these steps:
Figure 15-27. The Remove Mirror dialog box.
If a disk fails and you can't replace it with an identical one, you should break the mirror until a replacement becomes available. Breaking a mirror severs the connection between the two disks, allowing the remaining disk to continue to function normally until a replacement disk is available. You might also find it useful to break a mirror even when both disks are still functioning, as you then end up with two identical copies of the same data. One of the halves of the broken mirror continues to have the same drive letter and/or mount point, whereas the second half of the broken mirror is assigned the next available drive letter. To break a mirror, follow these steps:
Real World
Backing Up Active Files
One of the most difficult tasks faced by the system administrator is to get a reliable, sure backup of a file that is in constant active use, such as a data file for a database such as SQL Server or Oracle. The Break Mirror command can be used to get around this problem. You can momentarily stop the database, break the mirror, and restart the database. Now you have a copy of the data file that is no longer in active use and can be safely and effectively backed up. Once the backup has completed, you can delete the broken volume and re-create the mirror. Note, however, that the creation of a mirror involves substantial overhead and might adversely affect system performance until the regeneration is completed.
You can convert a volume or partition from the FAT or FAT32 file system to the NTFS file system without losing data or interrupting the availability of the rest of the server. However, Windows 2000 offers no graphical way to do this—you'll have to run a command-line utility. To convert a volume or partition, open a command window and type
convert <volumename | mountpoint | driveletter:> /fs:ntfs [/v].
This command converts the volume or drive from either FAT or FAT32 to NTFS. If you use the /v command-line switch, the conversion is fairly noisy, listing the name of each file and directory that's converted. If someone has a file open on the volume and the program cannot gain exclusive access to it, you are offered the opportunity to schedule the conversion for the next time you reboot. This option is acceptable if you have a planned maintenance reboot coming anyway, but otherwise you probably shouldn't schedule the conversion, as it might end up taking a fairly long time if the drive is large and contains a lot of files.
Real World
NTFS Conversion Planning
If you schedule a conversion to NTFS for the next reboot before you're actually ready to reboot the server and your server needs to be rebooted unexpectedly, you simply have to wait while the conversion happens. There is no way to bypass the conversion once you've committed to it. This can turn a minor downtime into a major headache if it happens in the middle of the production day. Don't commit to a conversion unless you are sure you can afford the downtime if Windows 2000 decides it needs to reboot.
Before a partition, logical drive, or volume can be used, it must be formatted. Formatting lays down the necessary structure to support the file system you choose for the volume. You must format a volume or drive when it is first created, and at any later point if you want to clear it off. You can also use Format to change the type of file system on a drive, partition, or volume, but all data on the target is deleted during the formatting. (The command-line Convert command—which allows conversion of FAT and FAT32 targets to the NTFS format only—preserves any data on the target.) The file systems that are supported by Windows 2000 are FAT, FAT32, and NTFS.
In general, we recommend that you use NTFS unless you have a compelling reason not to. One case in which you would not use NTFS is when log files will reside on the volume or partition. A FAT or FAT32 volume tends to be faster and more appropriate for large files that grow constantly in small increments, as log files do. However, even with log files, you should use the FAT or FAT32 file system only when security or quotas are not an issue, because these file systems do not support quotas or the security features of NTFS.
Another reason to use FAT or FAT32 is to allow the computer to support dual booting into other operating systems. NTFS is not visible or accessible from other operating systems, whereas FAT can be used by a variety of operating systems and FAT32 can be used by Microsoft Windows 95 and Microsoft Windows 98. To format a logical drive, partition, or volume, follow these steps:
Figure 15-28. The Format dialog box.
Real World
Assigning Volume Names
The name you assign to a volume, partition, or drive should tell you something about it rather than simply mimicking the drive letter. A volume name like "120GBSCSI" tells you pretty conclusively that it's that big new SCSI drive you just bought, unless, of course, you already have half a dozen of them on your server, in which case you're going to need to come up with a more effective name. On the other hand, a volume name of C_DRIVE is just about useless, because the drive letter is available from anywhere that the volume name is.
Real World
Optimum Cluster Size
A discussion of the best cluster size for a particular application or need is beyond the scope of this chapter and can be fairly heated, with an enormous amount of unsubstantiated and generally false statements being bandied about. Suffice it to say that the defaults are good for almost all situations and that decisions to choose something other than the default allocation size should be made only when there is a specific, clearly understood reason and a compelling need to do so. You can choose cluster sizes from 512 bytes up to 256 KB. You cannot enable file and folder compression on NTFS if you go beyond 4 KB for your allocation unit size.
If you seriously think you need to change the cluster size for a particular volume, we suggest you first make the change in a controlled lab environment and perform confidence and performance tests to ensure that the change meets your needs and doesn't cause unintended side effects.
You can change the drive letter of a volume or partition at any time, and you can even have multiple paths to a given drive. In addition, unlike Windows NT, Windows 2000 allows you to change the drive letter of a removable drive, such as a Jaz or Zip drive. To change a drive letter, follow these steps:
Figure 15-29. Changing the drive letter and path of a logical drive.
Figure 15-30. Message warning that a drive has a dual identity.
Windows 2000 adds a new feature to the disk and storage management process. You can mount a dynamic volume—or any partition or logical drive in an extended partition—on any empty directory that resides on a drive that is both NTFS formatted and nonremovable. The mounted volume can be formatted as FAT, FAT32, or NTFS and appears to users as a simple directory. This feature makes it possible to create larger file systems that use multiple hard disks without the inherent risks of using spanned volumes, as the failure of any one of the mounted volumes affects only the directories that were part of that volume. You can also easily support multiple formats from a single drive letter. To mount a volume, follow these steps:
Figure 15-31. The Add New Drive Letter Or Path dialog box, used to mount a volume.
It's actually easy to get yourself into trouble with this new feature. Disk Management lets you make multiple levels of mounted volumes, including ones that are recursive. We think you're well advised to mount volumes only at the root level of a drive. Trying to mount below that point can lead to confusion and make management and documentation difficult.