SUMMARY
This article provides some basic procedures and a checklist for creating
and modifying a fault tolerant set. For this article, a fault tolerant
set includes a stripe set, stripe set with parity, volume set, and disk
mirror/duplex.
Basic Procedures
- To create a fault tolerant set, perform a full backup of all data. Back
up anything that cannot be reinstalled and back up any data that you
cannot regenerate on physical media, other than the drives you are going
to use.
- Because information about fault tolerant sets is kept in the registry,
it is preferable that you back up the disk key in the system hive
separately. You can easily do this by using the Save option of the
Configuration entry under the Partition menu in Disk Administrator. This
option will copy the disk key in an uncompressed format to a floppy
disk. The disk key keeps the Disk Signature and any fault tolerant set
information of the drives in your system.
NOTE: It does not save the master boot record, boot sector, or any
partition information. You may also want to use RDISK to back up the
rest of the registry including the entire SYSTEM hive.
- All third-party services and drivers need to be stopped through Services
tool and Devices tool in Control Panel. These services may not only
be limited to virus scanners, backup agents, and network services, but
may extend to utilities that may maintain open files, or access system
services directly (low-level hardware access such as Compaq Insight
Manager). Also, applications such as Microsoft Exchange Server and SQL
Server, among others that perform low-level disk I/O, need to be
stopped.
The drive needs to be checked for file structure and physical corruption
with a utility such as CHKDSK. For CHKDSK use the /r to search for bad
sectors and /f to fix any errors it finds. If possible, use the SCSI
controller low-level media test to run a nondestructive test (consult
your hardware vendor for details and availability).
For additional information, please see the following articles in the
Microsoft Knowledge Base:
ARTICLE-ID: 100110
TITLE : Overview of Redundant Arrays of Inexpensive Disks (RAID)
ARTICLE-ID: 114872
TITLE : Restoring Disk Configuration Information
Stripe Sets
Stripe sets offer no data redundancy, unlike striping with parity. All data
is lost in the set if one drive fails in a stripe set. Better performance
usually results because of the ability to read and write to several drives
simultaneously. To do this, the drives and controller must support the read
and write request simultaneously. In general, most SCSI controllers and
drives allow this, while IDE controllers and drives only allow one device
per channel at a time to accept read and write requests. Making a stripe
set out of IDE drives will not improve performance.
Stripe sets with parity offer data redundancy. They maintain a fairly high
level of performance. Read performance matches a stripe set without
parity. When writing data, however, everything is written twice -- once to
the disk and once to the parity strip. The pagefile should not be placed on
a stripe set with parity because redundant data is written and performance
degrades. After a stripe set is created, it cannot be enlarged or
extended without backing up the data and reformatting the set. You cannot
incorporate the Windows NT system or boot files into a stripe set.
For additional information, please see the following articles in the
Microsoft Knowledge Base:
ARTICLE-ID: 113933
TITLE : Disk Striping and Disk Striping with Parity in Windows NT
ARTICLE-ID: 108082
TITLE : Windows NT Cannot Regenerate Stripe Set with Parity
Volume Set
Volume sets offer no data redundancy but allow you to span multiple drives,
and to assign them one drive letter. When you use NTFS on a volume set, it
is possible to add another drive to volume set without formatting the
drive. FAT does not allow extending the drive without reformatting and
restoring from a backup. When extending a volume set, take the same
precautions as when you create any other fault tolerant set.
For additional information, please see the following article in the
Microsoft Knowledge Base Article related to Volume Sets:
ARTICLE-ID: 113503
TITLE : Overview of Disk Volume Sets in Windows NT
Disk Mirror/Duplex
Disk mirroring and disk duplexing offer full data redundancy under Windows
NT Server. The difference between the two is that mirroring assumes you
have two hard drives and one controller, while duplexing has a controller
for each hard drive. Duplexing provides maximum protection if a controller
or a hard drive fails. The type and size of your hard drive does not
matter, but there must be enough free space available on the shadow drive.
When duplexing, drive translation can be an issue on SCSI drives. When the
BIOS is enabled on a SCSI controller, the drive translation is done by the
BIOS. For controllers not under BIOS control, Windows NT defaults to 64
heads and 32 sectors per track. This happens because Windows NT cannot read
the translation scheme from the disabled BIOS. An issue arises when there
are two identical controllers and two identical drives, with two
different sizes under Disk Administrator. In this situation, even if the
mirror can be established, recovery in the event of failure may be
extremely difficult.
For additional information, please see the following articles in the
Microsoft Knowledge Base:
ARTICLE-ID: 114779
TITLE : Overview of Disk Mirroring (RAID Level 1) in Windows NT
ARTICLE-ID: 113977
TITLE : Booting From Mirror After Primary Partition Is Lost
ARTICLE-ID: 141702
TITLE : How to Recover Mirroring Windows NT Using IDE devices
ARTICLE-ID: 161563
TITLE : How Windows NT Handles Drive Translation
ARTICLE-ID: 148425
TITLE : Duplexed Mirror Fails with Insufficient Space
ARTICLE-ID: 172912
TITLE : Windows NT May Appear to Hang When a Mirrored IDE Drive Fails