Recovering from Disk Failures

Unfortunately, hard disks can and do fail. This alone is a very solid reason for ensuring that you have a well thought out and practiced disaster-recovery plan in place (see Chapter 17 for more information on using the Windows Backup utility). If your data was on a basic storage disk or a simple, spanned, or striped dynamic volume, you have no choice but to replace the disk and restore the data from the last backup. If you experience failure of a disk in a mirrored volume or a RAID-5 array, fault tolerance is in place and you can recover your data.

Recovering a Failed Mirrored Drive

If one of the drives in a mirrored set happens to fail, you will be provided with a graphical indication in Disk Management. Fortunately, because mirrored volumes are fault tolerant, your server continues to operate normally, albeit with slower read performance. You can schedule your repairs at a convenient time.

The process to repair the mirror depends on whether the disk that failed was part of a mirrored set that contained the system or boot partition.

Note: Simulating an Error

The error condition in this Step by Step can be simulated by unplugging one of your physical drives if you're using dedicated hardware. If you're using a virtual product such as VMWare or Virtual PC, you can delete or disable one of your virtual drives.

If the failed disk did not contain the system or boot partition (only data), you can restore the mirror by following the procedure outlined in Step by Step 12.20.

Step by Step

12.20 Repairing a mirrored volume

1.	Open the Disk Management snap-in. You should have an error indication similar to that shown in Figure 12.21. The failed disk usually shows a status of Missing, and the volume set shows a status of Failed Redundancy. Figure 12.21. The Disk Manager snap-in indicates a problem with the mirrored set and flags the failed device.
2.	In the lower-right pane of the Disk Management snap-in, right-click the failed mirrored volume and select Remove Mirror from the shortcut menu.
3.	In the Remove Mirror dialog box, shown in Figure 12.22, select the disk that is to be removed and click Remove Mirror. Figure 12.22. After you highlight the failed drive and select the Remove Mirror button, the volume will be removed from the mirrored set.
4.	Confirm that you want to remove the mirror by clicking OK when prompted with a dialog box.
5.	Power down the computer, replace the hard disk, and restart the computer.
6.	Create a new mirrored volume using the previous mirrored disk and the replacement disk. Refer to the procedure for creating a mirror earlier in this chapter.

If the mirrored volume that contains the boot or system partition fails and it's the primary drive that has failed, you most likely will not be able to boot your server. This is because the BOOT.INI file is still configured to look for that drive. This will not affect the operation of your server because the server is able to access the files it needs to continue running from the other half of the mirror.

However, unless you have hot-swap drives and can replace the failed disk without shutting down the server, eventually you will need to power down, replace the failed disk, and then restart the server. The easiest way to reboot and repair the server is to create a Windows Server 2003 boot floppy that contains the system files and a copy of the BOOT.INI file that points to the secondary disk in the mirrored set.

Technically speaking, the floppy you are creating is not really a boot floppy because it doesn't contain the operating system. Earlier in the chapter we discussed the Windows Server 2003 boot process. When the server is started, it looks for the files in the Windows Server 2003 system partition. We will re-create the system partition on a floppy disk and edit the BOOT.INI file to point to the operating system files that are on the surviving secondary mirrored volume.

If the failed disk contained the system or boot partition, you can restore the mirror by following the procedure outlined in Step by Step 12.21.

Step by Step

12.21 Repairing a mirrored boot/system volume

1.	Open the Disk Management snap-in. You should have an obvious error indication. The failed volume usually shows a status of Missing.
2.	Determine which mirrored disk failed. If the secondary disk failed (the disk that contains the mirrored data), you can replace it as outlined previously. If the primary disk failed (the disk that contains the original data), you must proceed with these steps.
3.	If you have another Windows Server 2003 computer available (it must have a similar disk configuration) or if your server is still up and running, format a 3 1/2 inch floppy disk. (This disk must be formatted in Windows Server 2003 so that it has the correct MBR.) Copy the NTLDR, NtdETECT.COM, and BOOT.INI files to the floppy disk. You will need to edit the BOOT.INI file as shown in step 6 to reflect the location of your secondary drive.
4.	If you do not have another Windows Server 2003 computer available or if your server is down, copy the NtdETECT.COM file from the I386 folder on a Windows Server 2003 Setup CD-ROM to a blank, formatted 3-1/2 floppy. You also need to expand the NTLDR file from this same location by entering the following command from the command line: expand Ntldr._ Ntldr
5.	If your server uses SCSI drives that have the BIOS disabled (not common), you must include the driver ntbootdd.sys on your boot floppy. If used, this file is specific to your SCSI controller and will be present on your server.
6.	Finally, create a BOOT.INI file that points to the secondary drive in your broken mirror. (For more help on working with ARC paths, see http://support.microsoft.com/default.aspx?scid=kb;en-us;Q102873.) Following is an example: boot loader] timeout=30 default=multi(0)disk(0)rdisk(0)partition(1)\WINNT [operating systems] multi(0)disk(0)rdisk(0)partition(1)\WINNT="Microsoft Windows Server 2003"/fastdetect
7.	Power down the computer, replace the hard disk, and restart the computer.
8.	Use your boot disk to start your Windows Server 2003 computer.
9.	Remove and re-create the mirror as discussed in the previous procedure.

Prebuild the Recovery Floppy

When you initially configure your boot/system drives in a mirrored configuration, that's when you should build your recovery boot floppy. You should configure the boot floppy and then use it for a few test boots to make sure you have the paths configured correctly in the BOOT.INI file. Then make two or more copies of it and put them away in a safe place.

When the server has crashed and the phones are ringing off the hook is not the proper time to learn how to build and test a recovery floppy.

Recovering a Failed RAID-5 Drive

Recovering from a disk failure in a RAID-5 array is a fairly simple process thanks to the fault tolerance provided by the array. Remember that RAID-5 arrays can provide fault tolerance for only one failed disk, so be sure to replace the failed disk as soon as possible. While the disk is failed and not replaced, you can still use the RAID-5 array; however, I/O performance will be severely degraded because the missing data must be re-created from the parity information. Again, you should replace a failed disk in a RAID-5 array as soon as you can by performing the procedure outlined in Step by Step 12.22.

Step by Step

12.22 Repairing a RAID-5 volume

1.	Open the Disk Management snap-in. The failed volume usually shows a status of Failed Redundancy.
2.	Power down the server, if necessary. Replace the failed disk. Restart the server, if necessary. If the disk is hot swappable, rescan the disks.
3.	In the lower-right pane of the Disk Management snap-in, right-click one of the volumes in the failed RAID-5 set (it will be marked as Failed Redundancy) and select Reactivate Volume from the shortcut menu.
4.	When the confirmation prompt appears, click Yes to confirm that you want to reactivate the failed volume.
5.	When prompted, select the new disk you installed and click OK to begin the rebuilding process.

Importing Foreign Disks

Dynamic disks can be moved from any Windows 2000 computer or later to Windows Server 2003. When a dynamic disk is moved from one computer to another, it is marked as a Foreign Disk. This is because it's not a member of the disk group on the receiving computer. Before the disk can be used, it must join the disk group on the new computer by importing it. After the disk is imported, you can access any volumes that are on it.

Importing a foreign disk can be accomplished by using the diskpart utility or via the Disk Manager MMC. When moving a single disk, or the entire disk group to another computer, the process is fairly straightforward. Just right-click the entry for the foreign disk in the Disk Manager MMC and click Import.

Note: Rescan Disks

It is imperative that you rescan disks after you move hard disks between computers. When Disk Management rescans disk properties, it scans all attached disks for changes to the disk configuration. It also updates information about removable media, CD-ROM drives, basic volumes, file systems, and drive letters. Without a rescan, your new disks probably won't be recognized.

However, the situation becomes more complicated when fault tolerant or nonredundant volumes that span multiple disks are moved. For example, if you move one volume of a mirror, and then move the other volume at a later date, they will both appear to be in synch, but in reality they could possibly have different contents.

A second example is if you move one of the volumes of a RAID-0 array. This volume will appear as disabled, even after you attempt to import it. It can be brought back online after you move the remaining volumes over, or you can just delete it and reuse it.