Problem
How do I mirror my root disk to another disk using SCO Virtual Disk Manager?
A RAID 1 array is a "mirrored" pair of drives that contain identical data. It offers the fastest fault-tolerant architecture, especially when combined with other RAID-1 arrays via striping (RAID 10). With our software RAID (Virtual Disk Manager) the root filesystem is limited to RAID 1. Since the drives in the array will contain redundant data, the usable storage capacity is half the combined capacity of the drives. Software RAID allows the mirroring of disks on separate adapters. It is recommended that the adapters and disks are of the same make and model. The following example assumes two 525 megabyte SCSI disks attached to one SCSI host adapter. It also assumes that all the disk space is being used for Unix.
NOTE: As larger disks are coming on the market you may be tempted to use a portion of your root disk for a separate RAID array, one other than RAID-1. For example, you may have a 2 GB root disk and wish to configure 1 GB for your root filesystem and mirror that to your second disk. You may then decide to use the left over portion in a RAID-5 array. Raid-5 does not support the repositioning of RAID array pieces, which will become necessary should your root disk or mirrored disk fail. Therefore, it is recommended that any left over portion be used to create a RAID-1 array to be mirrored to your left over portion on your second disk. This will contain your non root filesystem, such as /u.
Solution
Virtual Disk Manager Release 1.1.0 includes a Release Supplement disk. Before proceeding, make sure it is installed. SCO OpenServer 5.0.2 and 5.0.4 have the 1.1.0 Supplement included in the product. Release 5.0.0 requires that you install this separately. It can be downloaded from our BBS in the Supplements/vdm110 directory. Virtual Disk Manager must be installed via the Software Manager and will require a license for each of the three mentioned releases.
Note: The 1.1.0 Release Supplement, among other things, allows you to mirror your system's boot division (/stand) onto a virtual disk. This will enable you to boot from the mirrored drive should your root disk fail. Again, SCO OpenServer releases 5.0.2 and 5.0.4 already have this included in the product.
Before proceeding it is important to know and understand the requirements for your system. If it is a 24X7 environment you will need to invest in hardware that supports this. See TA105386, "I am using Virtual Disk Manager but my system fails when one of my disk fails."
If you are using Adaptec adapters make sure you have bios configured correctly. Pay particular attention to the DOS drives > 1 gigabyte option. Make sure it is disabled. If you have already installed the OS with this option enabled, you cannot simply disable it. A reinstall is required, as the disk geometry has already been laid out.
NOTE: This no longer is an issue with SCO OpenServer 5.0.4.
It is also recommended that a low level format be performed to your second SCSI disk prior to configuring it in the SCO operating system.
This article also assumes that you have already installed the operating system and the Virtual Disk Manager product. We will take up where we are now adding our second disk to the system. This will become our mirrored disk. The Virtual Disk Manager refers to it as the "parity" disk.
Before proceeding, obtain the following information about the SCSI hard disk:
* host adapter type and its prefix (spad, eiad, arad, alad, and so on)
* host adapter number* bus idNote: You will be asked for the bus id. For a dual channel host
adapter, Channel A will be bus id =0 and Channel B will be bus
id=1. The host adapter number will be the same for either channel.
For example, when trying to add a SCSI disk that is attached to a
host adapter, and the host adapter is the first of its type in the
system, the host adapter number is 0, regardless of the channel the
device is attached to.* target id of the SCSI diskAfter obtaining the above information you are ready to proceed.
You must be in System Maintenance mode to continue. Type:# mkdev hdYou will see:1. Add a hard disk to IDE controller 2. Add a hard disk to SCSI controller 3. Add a hard disk to IDA controller
Choose 1, 2 or 3.Enter the prefix of the SCSI host adapter that supports
this device or press <Return> for the default of ad.Note: This may not always read 'ad' as the default.
Enter the prefix. You will see:Which xxxx SCSI host adapter supports this device?where 'xxxx' is the prefix you entered previously.
Enter the host adapter number.The following parameters will be used to configure xxxx SCSI host adapter z. Change these parameters y/n
where 'xxxx' is the prefix you entered previously and 'z' is the host adapter number.
You will see the setup parameters. If they are correct, choose 'n'.
If you need to change them, choose 'y' and enter the hardware
details about the host adapter card; confirm that you want to
save these values.What SCSI bus is this device attached to? Press <Return> to use the default:0
Enter bus id. You will see:What is the target id for this device?Note: This is the target id of the hard drive, not the host adapter.
Enter target id. You will see:What is the LUN?Enter 0. You will see:You are about to add the following SCSI device.
Host Adapter Type Device Adapter Id Lun Bus
---------------------------------------------------------
xxxx Sdsk 0 1 0 0where 'xxxx' is the prefix added previously.
The assumed host adapter is 0, target id of scsi disk is 1,
lun is always 0, and, in this case, the bus id was 0. You
will see:Update SCSI configuration? (y/n)Choose 'y' if all is as it should be. You will see:The SCSI configuration file has been updated.
A new kernel must be built and rebooted before disk configuration can continue. Would you like to relink at this time (y/n)
Choose 'y' to add the disk.NOTE: SCO OpenServer 5.0.4 has added some SCSI scanning features. If you had your second disk added to the bus during installation you may not have the option to relink. You may be just brought into fdisk.
You will be given the option to relink if the second disk has been low-level formatted.
You will then see:Do you want the kernel to boot by default?
Choose 'y'. You will see:Do you want the kernel environment rebuilt?Choose 'y'.Reboot the system and enter System Maintenance mode.
You must run 'mkdev hd' again and enter the same parameters as you did previously in order to finish configuring the hard drive. After adding the same parameters for the second drive you will see:
During installation you may choose to overwrite all or part of the present contents of your hard disk. Continue y/n
Choose 'y'. You will see:The hard disk installation program will now invoke /etc/fdisk. Entering 'q' at the following menu will exit /etc/fdisk, and the hard disk installation will continue.
1. Display Partition Table 2. Use Entire Disk for Unix 3. Use Rest of Disk for Unix 4. Create Unix Partition 5. Activate Partition 6. Delete Partition 7. Create Partition
Enter choice or q to quitNote: For this example, the entire disk for Unix will be used for the disk.
Choose "2. Use Entire Disk for Unix".
Note: If you previously had data on the disk, you will see this message:
Warning! All data on your disk will be lost! Do you wish to continue (y/n)
Choose 'y'. You will see:Total disk size .....
Press <return> to continue1. Scan entire Unix partition 2. Scan a specified range of blocks 3. Scan a specified filesystem 4. List current bad block table 5. Add entries to bad block table 6. Delete entries from bad block table 7. Clear bad block table 8. Re-allocate bad blocks
Enter choice or q to quitNOTE: If you performed a low level format of your scsi disk you can bypass this step and choose 'q' to quit.
Choose "1. Scan entire Unix partition". After scan completes you
should see:0 bad blocks have been found. Enter the number of bad blocks to allocate space for or press <Return> to use the existing value of 511. WARNING: Changing the value will remove all the filesystems in this partition.
After accepting the default you will see :
The virtual disk driver is installed. Do not create a division table on partitions which will be used for virtual disk pieces.
Choose one of the following options below or quit.
`y' to create a division table on the active partition`n' to avoid creating a division tableEnter `y', `n', or enter `q' to quitChoose 'q' to quit.IMPORTANT NOTE: YOU MUST CHOOSE Q TO QUIT HERE AS PREVIOUS NOTICE WARNS YOU! CREATING DIVISION TABLES ON PARTITIONS BEING USED FOR MIRROR WILL PREVENT THEM FROM BEING CONFIGURED IN THE ARRAY.
When you have completed this for your second drive, enter
multiuser mode and bring up the Virtual Disk Manager. To do
this, click on the System Administration ==>Filesystems ==>
Virtual Disk Manager.You will see:+----------------------------------------------------------------+
| Disk Piece Database Boot View Options |
| |
+----------------------------------------------------------------+Click on 'Boot'. A list will appear. Click on 'Mirror'.
This will bring up the following window:+----------------------------------------------------------------+
| Warning |
| |
| Mirror the root, swap or boot divisions. This operation will |
| relink the kernel. After this operation has completed, the |
| system must be rebooted immediately so that the new mirrors |
| are enabled |
| |
| OK Cancel |
+----------------------------------------------------------------+Choose OK.+----------------------------------------------------------------+
| |
| Mirror Root/Swap/Boot |
| |
| +---Select Divisions------------------------------------+ |
| | | |
| | Vdisk1 to mirror root: Yes No | |
| | | |
| | Vdisk2 to mirror swap: Yes No | |
| | | |
| | Vdisk3 to mirror boot: Yes No | |
| +-------------------------------------------------------+ |
| |
| Continue Cancel Help |
+----------------------------------------------------------------+Click on 'Continue'.NOTE: If you DO NOT choose Continue and choose NO on Vdisk1, 2 or 3, you will not be able to boot from the mirrored drive. Thus, DO choose Continue if you expect to be able to boot from the mirrored drive.
You will see:+------------------------------------------------------------------+
| Mirror Root/Swap/Boot |
| +---Piece Allocation---------------------------------------+ |
| | | |
| | | |
| | | |
| | Automatic Piece Allocation Yes No | |
| +----------------------------------------------------------+ |
| |
| |
| +--------------+ +--------------+ +--------+ |
| | Continue | | Cancel | | Help | |
| +--------------+ +--------------+ +--------+ |
+------------------------------------------------------------------+Click Continue.+------------------------------------------------------------------+
| Mirror Root/Swap/Boot |
| |
| +---Vdisk parameters-------------------------------------+ |
| | | |
| | Size of disk cluster 32 | |
| | | |
| +--------------------------------------------------------+ |
| |
| CONTINUE CANCEL HELP |
+------------------------------------------------------------------+Note: "Size of disk cluster" - A cluster consists of a set of
contiguous blocks of data written to a physical hard disk
within a disk array. In other words, it is the size of the
data chunks written to the pieces in an array. The cluster
size is the parameter that has the most impact on performance
of array systems. Improperly matched I/O and cluster sizes can
adversely affect performance. By default, you will see 32. This
figure is an approximation based on test results, but may not be
ideal for all applications. Modifying cluster size will require
reconfiguration. That will mean backing up the data, reconfiguring
the array with the new cluster size and restoring from backups.Click on 'CONTINUE'.+------------------------------------------------------------------+
| New Root/Swap/Boot Mirrors |
|------------------------------------------------------------------|
| Virtual Device: /dev/dsk/vdisk1 |
| +----------------------------------------------------------+ |
| |Piece1: /dev/dsk/0s1 Offset:13016 Length:663084 | |
| |Piece2: Not Allocated parity | |
| +----------------------------------------------------------+ |
| |
| Virtual Device: /dev/dsk/vdisk2 |
| +----------------------------------------------------------+ |
| |Piece1: /dev/dsk/0s1 Offset:32016 Length:98000 | |
| |Piece2: Not Allocated parity | |
| +----------------------------------------------------------+ |
| |
| Virtual Device: /dev/dsk/vdisk3 |
| +----------------------------------------------------------+ |
| |Piece1: /dev/dsk/0s1 Offset:2016 Length:30000 | |
| |Piece2: Not allocated parity | |
| +----------------------------------------------------------+ |
| |
| Allocate pieces... |
| |
| Create Cancel Help |
+------------------------------------------------------------------+As you can see, Piece 1 for each of the Virtual Devices has been allocated. That is, the root, swap, and boot on the root disk has been allocated. We must now allocate Piece 2 for our mirrored drive. So select Allocate pieces. In fact, Create is shaded out at this point as we are not yet ready to Create the mirror.
Should you see the error:Error. There is not enough free space available on the system to complete this configuration, it is all being used for division and/or vdisk pieces.
As a check that everything is OK at this point then you can run divvy against the additional disk , with the command "divvy /dev/hd1a" and this should indicate that no divvy table is present on this disk. If not then you can run "divvy -D x " where x is 0 through to 6 to remove any divisions.
Additionally, you can run Disk -> Examine Disk Map under VDM and ensure that VDM sees the second disk.
If this all looks fine then you could compare the output of:
fdisk -f /dev/rdsk/0s0 fdisk -f /dev/rdsk/1s0 (assuming you are mirroring to the second disk)
The sizes of the two active Unix partitions should be the same (or the second should be no smaller than the first).
The next window will allow you to select the device for the pieces for the mirror to reside. If you only have the second disk added to the system and no others you will see /dev/dsk/1s1. Follow the default settings and choose OK. You will get back to the screen as shown above but this time Allocate pieces will be shaded out. Now click Create.
+--------------------------------------------------------+
| Information |
+--------------------------------------------------------+
|Successfully mirrored root on vdisk1 |
| |
|Successfully mirrored swap on vdisk2 |
| |
|Successfully mirrored boot on vdisk3 |
| |
|The system has been successfully relinked with the new |
|configuration. It must be rebooted for the new |
|configuration to take affect. |
+--------------------------------------------------------+To be sure you have configured your Virtual disks correctly, have a look at the resulting /etc/dktab file. This file depicts the configuration you just created. An example follows:
/dev/dsk/vdisk1 mirror 2 32/dev/dsk/0s1 130016 663084
/dev/dsk/1s1 130016 663084/dev/dsk/vdisk2 mirror 2 32/dev/dsk/0s1 32016 98000
/dev/dsk/1s1 32016 98000/dev/dsk/vdisk3 mirror 2 32/dev/dsk/0s1 2016 30000
/dev/dsk/1s1 2016 30000Let's analyze the entries for vdisk1 so that we might understand the contents of this file. The first line shows the device name /dev/dsk/vdisk1. This is the device used to mirror root. It consists of 2 pieces with a cluster size of 32 (16K).
The second and third lines show the offset (where the data begins) and the length of the piece. Notice the offsets and lengths match. As long as the lengths are the same the Virtual Disk Manager will allow you to configure the mirror with different offsets but YOU WILL NOT BE ABLE TO BOOT FROM THE MIRRORED DISK. So, make sure your OFFSETS ARE THE SAME for each respective mirrored set.
You are now ready to relink the kernel and shut down. Upon reboot you will see activity on the disks as the data is being mirrored to the mirrored disk. The following command will show the process:
ps -ef | grep dkconfigUpon completion all data written to your root disk will simultaneously be written to your mirrored disk.
NOTE: It is highly recommended that the /etc/dktab file be backed
up to floppy or a log book. Should this file become corrupted, a
backup might then be accessible, or the file can be recreated. Be
sure to keep current backups of this file should any changes be
made. See also
SCO OpenServer 5.0 System Administration Guide, Chapter 8, Virtual Disk Manager 1.1.0 Release Notes and other related articles in this database.