Chapter 11
Move Root and Swap to a Fibre Channel (Shareable) Disk

If the root and swap partitions are currently on a local disk, you can move them to a Fibre Channel disk after the system is running DYNIX/ptx V4.6 and disk labeling is enabled. (rd disks and any other disks requiring the installation of a layered product cannot be used for this purpose.)


ATTENTION

On single-node systems, disk labeling is automatically enabled on the first boot. On clustered systems, disk labeling is enabled after all nodes are booted to multiuser on V4.6 and have run the rolling_upgrade_complete script.


The procedure for moving the root and swap partitions to an FC disk is different for those systems running ptx/SVM than for those systems that are not.


ATTENTION

Booting on shareable disks attached to the Fibre Channel is only supported for IBM Enterprise Storage Subsystems (ESS), IBM FAStT200TM Storage Servers, and Pbay storage subsystems.

This feature is not supported for EMC® Symmetrix® storage subsystems or HDSTM disk arrays from Hitachi Data Systems.



Systems Running ptx/SVM

Complete the following procedure to move the root and swap partitions on a system that is running ptx/SVM:

  1. Use the diskown command to label an FC disk as node-owned. The label should also include a unique user-defined name and a unique comment.

    # diskown -w -n node_uuid [-N user-defined_name] [-C comment] diskname

    The user-defined name allows the following simple syntax to be used in the bootpath:

    nm(user_defined_name)disk()

    The comment enables you to easily identify the root disk if you boot with the -b option. This option lists the node-owned disks, including the comment from each disk label.

    For example, to label the disk sd66 as node-owned by the current host with the user-defined name shared_disk and the comment shared disk, enter the following:

    # diskown -w -n 0 -N shared_disk -C "shared disk" sd66
  2. If you need to expand the root partition on the local disk, do so as described in Appendix C before completing the next step.

  3. If necessary, use mkvtoc or crvtoc to create a new VTOC on the target FC disk(s) with the same size root and swap partitions as the local boot disk has defined..

    # /etc/crvtoc -w -b -s {-c chunk_size | -n num_parts} -d disk_name -o /etc/vtoc/disk_name
  4. If you created a new VTOC, run devbuild to build the VTOC on the FC disk.

    # devbuild sd66
  5. Initialize the new root partition, add the root and swap partitions to the rootdg disk group and then create a mirror of the current root partition and current swap partition. This example assumes that the new FC root partition is sd66s0 and the new FC swap partition is sd66s1.

    # vxdctl enable
    # vxdisk init sd66s0
    # vxdg -g rootdg adddisk sd66s0 sd66s1
    # vxassist mirror ROOTVOL sd66s0 &
    # vxassist mirror SWAPVOL sd66s1 &
  6. Verify that ROOTVOL and SWAPVOL are on the FC disk(s) you designated for the root and swap partitions:

    # vxprint -g rootdg -ht
  7. Use the mapdev -a command to return the physical specifiers for the FC disks that contain the root filesystem and the swap partition. A physical specifier can be a label name, disk serial number, or physical path. For example:

    # mapdev -a sd66s0
    nm(shared_root)disk(0,0,0)
    sn(IBM OEM,DCHS09X,6808F12D)disk(0,0,0)
    quad(0)pci(0,0)fc(0x11c00,7)scsi(4,0)disk(0,0,0)

    # mapdev -a sd66s1
    nm(shared_root)disk(1,0,0)
    sn(IBM OEM,DCHS09X,6808F12D)disk(1,0,0)
    quad(0)pci(0,0)fc(0x11c00,7)scsi(4,0)disk(1,0,0)
  8. Set the boot strings to point to the new root and swap partitions. In the bootpath, use the user-defined name you specified for the disk label. This command is shown on two lines for readability.

    # /etc/bootflags -p -c 'n0=nm(shared_root)disk(0,0,0) 
    -s nm(shared_root)disk(1,0,0) -L -v -V'
  9. Reboot the system.

  10. Edit the boot file to point to the new swap partition on the FC disk:

    # cd /etc/system
    # vi boot

    The line in the file should look similar to the following:

    set primaryswap = sd66s1
    

    This step is necessary because the primary swap definition in the bootflags command does not update the /etc/system/boot file. Once booted, the swap designation in the bootflags command is removed and the system relies upon the */etc/system/boot file to define primary swap space.

  11. Examine the system boot parameters for the root and swap partitions to ensure that both the root and swap partitions are on the FC disk:

    # /etc/bfget -r -s
    root_dev=sd66s0:0:0
    swap_dev=sd66s1:0:0

Systems Not Running ptx/SVM

Complete the following procedure to move the root and swap partitions on a system that is not running ptx/SVM:

  1. Use the diskown command to label an FC disk as node-owned. The label should also include a unique user-defined name and a unique comment.

    # diskown -w -n node_uuid [-N user-defined_name] [-C comment] diskname

    The user-defined name allows the following simple syntax to be used in the bootpath:

    nm(user_defined_name)disk()

    The comment enables you to easily identify the root disk if you boot with the -b option. This option lists the node-owned disks, including the comment from each disk label.

    For example, to label the disk sd66 as node-owned by the current host with the user-defined name shared_disk and the comment shared disk, enter the following:

    # diskown -w -n 0 -N shared_disk -C "shared disk" sd66
  2. If you need to expand the root partition on the local disk, do so as described in Appendix C before completing the next step.

  3. If necessary, use mkvtoc or crvtoc to create a new VTOC on the target FC disk(s) with the same size root and swap partitions as the local boot disk has defined..

    # /etc/crvtoc -w -b -s {-c chunk_size | -n num_parts} -d disk_name -o /etc/vtoc/disk_name
  4. If you created a new VTOC, run devbuild to build the VTOC on the FC disk.

    # devbuild sd66
  5. Copy the root filesystem to the new root partition on the Fibre Channel disk by using dump and restore.

    # /etc/newfs /dev/rdsk/sd66s0
    # /etc/mount /dev/dsk/sd66s0 /mnt
    # sync
    # /etc/dump 0bdsf 32 54000 10000 - /dev/rdsk/sd0s0 | (cd /mnt : /etc/restore xbf 2 -)

  6. Check the consistency of the root filesystem:

    # fsck -y /dev/rdsk/sd66s0
  7. Use the mapdev -a command to return the physical specifiers for the FC disks that contain the root filesystem and the swap partition. A physical specifier can be a label name, disk serial number, or physical path. For example:

    # mapdev -a sd66s0
    nm(shared_root)disk(0,0,0)
    sn(IBM OEM,DCHS09X,6808F12D)disk(0,0,0)
    quad(0)pci(0,0)fc(0x11c00,7)scsi(4,0)disk(0,0,0)

    # mapdev -a sd66s1
    nm(shared_root)disk(1,0,0)
    sn(IBM OEM,DCHS09X,6808F12D)disk(1,0,0)
    quad(0)pci(0,0)fc(0x11c00,7)scsi(4,0)disk(1,0,0)
  8. Set the boot strings to point to the new root and swap partitions. In the bootpath, use the user-defined name you specified for the disk label. This command is shown on two lines for readability.

    # /etc/bootflags -p -c 'n0=nm(shared_root)disk(0,0,0) 
    -s nm(shared_root)disk(1,0,0) -L -v -V'
  9. Reboot the system.

  10. Edit the boot file to point to the new swap partition on the FC disk:

    # cd /etc/system
    # vi boot

    The line in the file should look similar to the following:

    set primaryswap = sd66s1
    

    This step is necessary because the primary swap definition in the bootflags command does not update the /etc/system/boot file. Once booted, the swap designation in the bootflags command is removed and the system relies upon the */etc/system/boot file to define primary swap space.

  11. Examine the system boot parameters for the root and swap partitions to ensure that both the root and swap partitions are on the FC disk:

    # /etc/bfget -r -s
    root_dev=sd66s0:0:0
    swap_dev=sd66s1:0:0