PRIMECLUSTER Global Disk Services Configuration and Administration Guide 4.2 (Linux)

6.1.4 Restoring (When the System Cannot Be Booted)

1) Turn on the power of the node, and insert the OS installation CD into the CD-ROM drive.

 

2) From boot devices displayed in the boot option selector window of the EFI boot manager, select the CD-ROM device, and boot the system in rescue mode.

With Red Hat Enterprise Linux AS (v.4 for Itanium), boot the system using the following procedure.
For details, see the OS manual.

EFI Boot Manager ver 1.10

Please select a boot option

Root1
Root2
DVD/Acpi(PNP0A03,0)/Pci(1D|1)/Usb(0,0) Select
...

Use and to change option(s). Use Enter to select an option

• When "ELILO boot:" is displayed, enter "linux rescue."

• In the Choose a Language window, select "English."

• In the Keyboard Type window, select "us." However, it may be necessary to choose another option depending on the keyboard.

• In the Setup Networking window, select "Yes" to perform network configuration, or select "No" not to. If "Yes" is selected, the IP address setup window will appear. Following the instructions of this window, specify the IP address.

• In the Rescue window, select "Skip."

   

3) Check the restore destination physical slice names.

Check the restore destination physical disk names.

# ls -l /sys/block/sd*/device | grep 0000:06:02.0 | grep 0:0:0
lrwxrwxrwx  1 root root 0 Jun  1  2005 /sys/block/sda/device ->\
../../devices/pci0000:02/0000:02:1f.0/0000:06:02.0/host2/\
target0:0:0/0:0:0:0

# ls -l /sys/block/sd*/device | grep 0000:06:02:0 | grep 0:2:0
lrwxrwxrwx  1 root root 0 Jun  1  2005 /sys/block/sdb/device ->\
../../devices/pci0000:02/0000:02:1f.0/0000:06:02.0/host2/\
target0:0:2/0:0:2:0

For the grep command arguments, specify physical disk information on the restore destination disks (Root1 and Root2 in this example) confirmed as described in "Checking Physical Disk Information and Slice Numbers."

In this example, the physical disk names are sda and sdb.

By combining the physical disk names and the slice numbers confirmed as described in "Checking Physical Disk Information and Slice Numbers," you can get the physical slice names.

In this example, the restore destination physical slice names are as follows.

Use	Physical slice name
/	sda1	sdb1
/var	sda2	sdb2
/usr	sda3	sdb3
/boot	sda4	sdb4
/boot/efi	sda5	sdb5

When Using the System Volume Snapshot Function
Restore also the joined proxy volumes of system volumes. It is not necessary to restore the parted proxy volumes.
If the proxy volume status is unknown, all the proxy volumes of system volumes should be restored.

 

4) Restore backup data on a tape medium back to one of the two slices.

The following example shows the procedure for restoring the root file system using data backed up with the dump(8) command. In this example, the file system type is ext3, and a temporary mount point is the /work directory.

# mkdir /work
# mkfs.ext3 /dev/sda2
# mount -t ext3 /dev/sda2 /work
# cd /work
# restore rf /dev/st0 .
# cd /
# umount /work

For the mkfs.ext3(8) command and mount(8) command arguments, specify the device special file for one of the two slices confirmed in step 3).

• If restoration to a slice fails due to an I/O error and so on, restore data to the other slice.

• Do not perform restoration using data backed up before system disk mirroring.

For details on the restore methods, see the manuals of file systems to be restored and used commands.

 

5) Copy data from the slice restored in step 4) to other slices.
The following example shows the procedure for copying data from sda2 to sdb2.

# dd if=/dev/sda2 of=/dev/sdb2 bs=1M

If the mirroring multiplicity is n, among n mirrored slices, copy data to all "n-1" slices other than the slice restored in step 4).

Specify a correct option referring to the dd(1) command manual.

 

6) Exit the rescue mode, and boot the system.

With Red Hat Enterprise Linux AS (v.4 for Itanium), exit the rescue mode using the following command.
For details, see the OS manual.

# exit