Problem
I need to restore my data from a Compaq server running with the Smart Array 3200 (ida) to another Compaq server running with the Smart Array 431 (clad) using BackupEdge.
The server used in production is a Compaq server running with the IDA driver for the Compaq SMART Array 3200 controller with a tape device connected to the on-board Compaq-Fast-SCSI-2 controller (cha).
ida - Smart Array 3200 range clad - Smary Array 4200 range ciss - Smart Array 5300 / 6i range
However, the server I need to restore to is a Compaq server running with the CLAD driver for the Compaq SMART Array 431 controller with a tape device connected to the on-board Compaq-Fast-SCSI-2 controller (cha) or the Compaq Ultra3 SCSI Controller (ad160).
How can I restore my data from one server to another when the kernel of the production server only knows about the "ida" driver and not the "clad" driver?
Solution
BackupEdge, a product available from MicroLite, at http://www.microlite.com, is a product used for backup and emergency recovery.
Version 01.01.08b7 has Boot Time Loadable Driver (BTLD) support which previously was not available.
With previous versions, you could go from, for example, an Adaptec
controller to a new server running a Mylex controller. You need to
install the Mylex driver on the original server and create two
kernels both the Mylex and the Adaptec could be the primary boot
controller. This is be done by manually modifying the following
two files:/etc/conf/cf.d/mscsiand/etc/conf/sdevice.d/xxxwhere xxx is the name of the controller. A new kernel could be
created so, when the installation was restored to the new server, it
could boot with the appropriate kernel and have the necessary
changes made to the above files before making the final kernel build.1. Create RecoverEdge Disks on the Production ServerThis version has BTLD support, so you are able to install new BTLD from the boot: prompt at boot time.
Assuming that the product has been purchased and installed, there are a number of steps necessary to first prepare both systems for recovery.
NOTE: During the installation of BackupEdge, it is recommended to run a "Manual Check" to set the optimum threshold for the tape device.
a) Use the "RecoverEdge" program on the production server (/usr/lib/edge/bin/re2) to build the necessary recovery disks. Select the "Generate" option and you will be prompted to install two floppy disks, which will be used to generate the RecoverEdge "BTLD Boot" and "Filesytem" disks.
b) During this process, RecoverEdge may notify you that your kernel is too large to fit onto the floppy disk. A common reason for this is the Compaq Health and Wellness drivers that can be installed as part of the Compaq Extended Feature Supplement (EFS) and can increase the size of the kernel from 2MB to 3MB.
In order for your kernel to fit onto the floppy, you
will need to reduce the kernel size. The aim here is to
have a kernel that can find the hard disk and the tape
device in order for you to restore the data onto the
same server when the server is ready to go back into
production after a disaster./stand holds various copies of the kernel, including "unix.old" and "unix.safe" which were created at installation time.
You can copy one of these kernels to /kernel.edge and
restart the "re2" program. This should then fit onto
your "BTLD Boot Disk".Alternatively, you can reduce the size of the kernel. To do this, it is recommended to keep an existing copy of the kernel using the following commands:
/etc/btmnt -u
/etc/btmnt -w
cp /stand/unix /stand/unix.orig
/etc/btmnt -dTo reduce the size of the kernel, you need to remove some device drivers, for example, the networking cards from with "netconfig". Relink your kernel and re-boot, then run "re2" again.
Alternatively, copy the current 'unix' kernel to 'unix.good'
and copy 'unix.safe' to 'unix' in /stand (see btmt -w) and
use this kernel to build the Emergency Floppies. Remember to
reverse the process after "mkdev fd" has been run.Note: "re2" can only be run in Multi-User mode. You will
also see the problem of the kernel size being too big
when trying to create the normal Emergency Recovery disks
using "mkdev fd". You may get the message:"The currently installed kernel will not fit onto the
boot filesystem. Therefore, please select the option to
build a generic boot. If you used BTLD drivers, then you
will still be able to use them. Press <RETURN> to
continue."c) Once you have created these disks, re-boot your server
with these disks to ensure that the hard disk and the tape
device can be seen by RecoverEdge. See the RecoverEdge
documentation for more details.d) It is very important to ensure that you have a printout of the server's "fdisk" and "divvy" filesystem layout.
2. Preparation on the Disaster Recovery (DR) Servera) Install OpenServer 5.0.X on the Disaster Recovery (DR) server ensuring that the filesystems are the same size or bigger than the production server's layout.
To save time, you can install the "Lightweight character terminal configuration", which is 142MB.
Ensure that the filesystem layout is not mapped to
another directory in your "divvy" table. For example,
if your divvy table has a /u filesystem, but is actually
mounted on another directory, RecoverEdge will mount
on /u. This means that the data restored will be restored
into the root filesystem and could result in the root
filesystem running out of disk space.Within the "customize" selection when building the filesystem layout, you can give the name of the filesystem as "u", for example, and the mount point can be a directory of your choice (/data, for instance).
b) Run "mkdev tape" and add the tape device on the appropriate device. You may wish to move the device from the "ad160" to the on-board Compaq-Fast-SCSI-2 controller (cha) so that the device is on the same controller as the production server. Otherwise, you will need to load the BTLD for the "ad160" at boot time.
This will require a link of the kernel and re-boot the operating system to test the tape using the following commands:
tape status
tape xstatus - removed in tape v4.3 - v5.0.7
tape reset
tape getdensity - removed in tape v4.3 - v5.0.7cd /; find etc -follow -print | cpio -ovcBL -O/dev/rStp0
cd /tmp
cpio -idcvmBu -I/dev/rStp0
rm -r /tmp/etcc) Install BackupEdge on the DR server and re-create the
RecoverEdge "BTLD Boot" and "Filesystem" disks. The
kernel should be small enough to fit onto the floppy
assuming that the Compaq's EFS Health & Wellness
software has not been installed.NOTE: During the installation of BackupEdge it is recommended to run a "Manual Check" to set the optimum threshold for the tape device.
d) You should also be able to use the RecoverEdge disks created from the production server to boot on your disaster recovery server. However, the kernel on these disks will believe that the "ida" controller should contain the root disk, but in this example a "clad" controller is being used.
At the "boot:" prompt, you can enter boot string commands as BackupEdge Version 01.01.08b7 has support for this.
For example:Boot: defbootstr link="clad ad160" Sdsk=clad(0,0,0,0)
Stp=ad160(0,0,6,0)Please note that you will also be able to enter this info from the SCO OpenServer Installation Media to gain access to a connected tape drive at installation time or during an Emergency recovery situation.
However, it could be possible that there is not enough
memory in the RAMDisk to load these additional drivers.
You will be prompted to enter the "BTLD" disk for the
"clad" and "ad160" drivers. These are available from
the "BTLD" disk of Compaq EFS available from
https://www.hpe.comIf an "Out of Memory" message is displayed or the server "freezes" during the boot of the kernel after the RecoverEdge "Filesystem" disk has been inserted, that is, while loading "ramdisk.Z", then it will not be possible to load these additional drivers.
If, however, you are using a small enough kernel where you can load additional BTLD drivers and specify the controller and root disk, as in the above defbootstr example, then it could still be possible that at boot time you see the following "no root controller found" type message.
3. Using the Recovery Disks on the Disaster Recovery ServerYou should now be able to boot from the RecoverEdge floppies created on the same disaster recovery server, using the drivers for the "clad" or "ida" adapters and restore your data successfully using the following instructions:
a) Go into Utilities | Disk Utilities and check to ensure
that a UNIX FDISK partition is present and ACTIVE.
There should also be a Compaq SmartStart partition which
is TYPE 18.b) Check the "Divvy" option to ensure that the necessary filesystems are present and are at equal or greater size to the production server's filesystem layout.
c) Run the "Check FileSystem" option to check that all the filesystems are valid and can be mounted using the "fsck -ofull" option.
d) Use the "Restore" option to restore all the filesystems.
e) Select the "Write MasterBoot" to write the Master Boot Record back to the hard disk to ensure it will boot.
f) Check the Tape Device, in Utilities | Tape Utilities |
Test_Tape_Drv. You can interrupt the listing with the
interrupt key once you have established the tape device
is working. If it does not, it may be possible that the
tape device is on a different controller host adapter
port.For example, the production server may have the tape device on:
tape - - - type=S ha=2 id=6 lun=0 bus=0 ht=chabut the disaster recovery server may have the tape device on:
tape - - - type=S ha=1 id=6 lun=0 bus=0 ht=chaIf it is not possible to move it, you can specify the location at boot time with the following command:
defbootstr Stp=cha(1,0,6,0)Or select "Automatic" to automatically restore your
data.g) Once complete, re-boot your server.4. Re-configure the kernel so it knows that the root disk is on the "clad" controller
When you have finished restoring, remember that the kernel on the hard drive does NOT have a device driver for the host adapter you've used. Follow the procedures below, to boot into single user mode and permanently link the BTLD driver into the boot kernel.
a) Boot from the RecoverEdge BTLD Boot Disk (not from the
hard drive). At the Boot: prompt enter:boothd link=clad Sdsk=clad(0,0,0,0)This will load the /stand/unix.re.Z kernel from your hard drive and prompt you for the BTLD disk when needed. If for any reason the unix.re.Z file was removed or doesn't load because of the "clad" driver freezing, enter the following instead:
bootfdThis will load the kernel from the floppy, which is slower but accomplishes the same thing.
b) You MUST type the root password at the <Ctrl> D prompt
and go into Single User mode. At the command prompt:Mount the BTLD floppy diskette with the commands:
mount -r /dev/fd0135ds18 /mntRun the BTLD installer:btldinstall /mntThe floppy should then be read and should display a list of available drivers that can be installed.
At this point, enter the name of the package (driver) you used, eg. "clad".
Normally, after the driver has been installed you should see:
Do you want to make the (Driver) adapter your default boot adapter?
Answer "yes" if you want to boot from the (Driver)
You should then get the following message:
"File /etc/conf/cf.d/mscsi has been modified to boot
from (Driver). Please note that a new kernel must be
built for the changes to take effect. Installed (Driver)"However, the "clad" and "ida" drivers do not operate in the same way and you do not get the above message for the "clad" driver.
c) Check the /etc/conf/cf.d/mscsi file.Make sure you change all of the devices to reflect the correct host adapter driver.
Ensure that the file contains:clad sDsk 0 0 0 0You will not find the "ida" entry here as this operates
differently from other SCSI controllers. You would
normally run "mkdev ida" to configure "ida" devices.Optionally, you may turn off the old SCSI driver to save
space in the kernel. Do this by editing
/etc/conf/sdevice.d/XXX*, where XXX is the old driver
name.For example, /etc/conf/sdevice.d/ida. Change field 2 in
each line from a Y to an N. This will disable the driver
in subsequent kernel builds.d) Then rebuild the kernel with the following commands:
cd /etc/conf/cf.d
./link_unix -yWhen complete, un-mount the floppy disk with:
cd /
umount /mntThen shut the server down and reboot.
Please Note: you still may see the following message during the next boot:
WARNING: apic: no BIOS info found for IRQ x WARNING: hd: no root disk controller was found hd: a Boot-Time Loadable driver maybe required H iinit PANIC: smountfun - Error 19 mounting root dev (1/42) Unable to freeze processor 2, proceeding ... Error 19 opening dumpdev(1/41)
Dump not complete** Safe to power off ** or ** Press any key to reboot **
The server displays this message regarding "sdmountfun"
and "No root controller" when detecting the "clad"
driver, as described above. Although loaded into the
kernel, it did not correctly create the necessary
/etc/conf/node.d entry. This means that although the
kernel can now detect the "clad" driver, it is unable to
see the device node for it.You will be able to boot into Multi-User mode using the
"bootfd" solution, but you will be unable to create a new
device, such as a different network interface card (nic)
using "netconfig". The kernel will link but the device
will not be present at the next boot because the kernel
is being loaded from the floppy disk.You will need to ensure that the "clad" driver has the following:
Ensure the /dev directory has a "clad" file:
crw------- 1 root root 51,0 cladEnsure that the /etc/conf/node.d/clad file exists and
contains:clad clad c 0 root root 600Once you have done this re-link and re-boot your server.
These problems have been reported to Microlite support.
ALTERNATE SOLUTIONS:a) Having followed the above steps and are still unable to boot
from the hard disk, you still can re-install OpenServer
5.0.6 over your disaster recovery installation and preserve
the existing layout. You must first ensure that all your
applications, data, user, host and printer information is
copied to an additional partition that can be restored after
the installation.b) Replace the Compaq "clad" SMARTArray 431 controller with a Compaq "ida" SMARTArray 3200 controller and use the original RecoverEdge floppies and tape to restore. This should preserve your existing logical drive definitions.
You would only need to follow steps 1, to prepare and 3, to restore (using the RecoverEdge disks from the production server), above.
NOTES: Once restored you will need to ensure that the following devices
are correctly installed:a) The tape device, configured with "mkdev tape"b) The networking, configured with "netconfig" (Check /etc/hosts for server name duplication)
See also
ta105937, "SCO OpenServer installation bootstring options."ta112505, "Additional Notes on Emergency Recovery for SCO OpenServer 5.0.6."
ta114161, "Kernel is too large to fit on a boot diskette, which causes the creation of emergency boot and root floppies to fail."
http://www.microlite.com