Summary: Moving cluster root and boot disk

From: Raymond Pang <wmpang_at_se.cuhk.edu.hk>
Date: Mon, 05 Aug 2002 15:13:35 +0800

Hi,

Sorry for this late summary. I finally finished the migration last Saturday.

Thanks to:

"Hagan, Stephen" <Stephen.Hagan_at_hp.com>
david.ross_at_cantire.com
alan_at_nabeth.cxo.cpqcorp.net
Matthew Wild <wild_at_wdcc1.bnsc.rl.ac.uk>

Especially Stephen.Hagan_at_hp.com, he provided me the detailed steps of
how this could be done. Please email me if anybody is interested.

Anyway, below was what I did.

Member boot disk migration
==========================
clu_get_info -m 2 - determine members is down
# backup the boot partition
mount root2_domain#root /mnt
vdump -0Df /var/cluster/members/member2/boot_part_vdump /mnt
umount /mnt
disklabel -z dsk147
disklabel -rw -t advfs dsk147 hsg80
disklabel -e dsk147 # partition the disk
/usr/sbin/clu_bdmgr -c dsk147 2 # config dsk147 as member2 boot disk
Ignore "Bad disk label" error
mount root2_domain#root /mnt
vrestore -xf /var/cluster/members/member2/boot_part_vdump -D /mnt
/usr/sbin/clu_bdmgr -h dsk147 /mnt/etc/clu_bdmgr.conf

file /dev/disk/dsk147h # get major/minor
cd /mnt/etc
cp sysconfigtab sysconfigtab-bu
vi sysconfigtab
  swapdevice=/dev/disk/dsk147b
  clubase:
  cluster_seqdisk_major=19 - from "file /dev/disk/dsk147h"
  cluster_seqdisk_minor=2767 - from "file /dev/disk/dsk147h"
umount /mnt

Boot from new disk
>>> show dev - find the correct boot disk
>>> set bootdef_dev dkd0
>>> boot dkd0

Cluster common disk migration
=============================
  # cluster root duplication
  mkfdmn /dev/disk/dsk145b cluster_target_root
  mkfset cluster_target_root root
  mkdir /cluster_target_root
  mount cluster_target_root#root /cluster_target_root
  clonefset cluster_root root clone_root
  mkdir -p /clone/clone_root
  mount cluster_root#clone_root /clone/clone_root
  vdump -0f - /clone/clone_root | (cd /cluster_target_root; vrestore -xf -)
  umount /clone/clone_root
  rmfset cluster_root clone_root
  rmdir /clone/clone_root

  cd /cluster_target_root/etc/fdmns
  cd cluster_root
  rm *;ln -s /dev/disk/dsk145b
  cd ../cluster_usr
  rm *;ln -s /dev/disk/dsk145g
  cd ../cluster_var
  rm *;ln -s /dev/disk/dsk145h
  cd
  umount /cluster_target
  rmdir /cluster_target_root

  # cluster usr duplication
  rm -r /etc/fdmns/cluster_target_root
  mkfdmn /dev/disk/dsk145g cluster_target_usr
  mkfset cluster_target_usr usr
  mount cluster_target_usr#usr /cluster_target_usr
  clonefset cluster_usr usr clone_usr
  mkdir -p /clone/clone_usr
  mount cluster_usr#clone_usr /clone/clone_usr
  vdump -0f - /clone/clone_usr | (cd /cluster_target_usr; vrestore - xf -)
  umount /clone/clone_usr
  rmfset cluster_usr clone_usr
  rmdir /clone/clone_usr
  rmdir /cluster_target_usr

  # cluster var duplication
  rm -r /etc/fdmns/cluster_target_usr
  mkfdmn /dev/disk/dsk145h cluster_target_var
  mkfset cluster_target_var var
  mount cluster_target_var#var /cluster_target_var
  clonefset cluster_var var clone_var
  mkdir -p /clone/clone_var
  mount cluster_var#clone_var /clone/clone_var
  vdump -0f - /clone/clone_var | (cd /cluster_target_var; vrestore -xf -)
  umount /clone/clone_var
  rmfset cluster_var clone_var
  rmdir /clone/clone_var
  rmdir /cluster_target_var


 file /dev/disk/dsk145b
           /dev/disk/dsk145b: block special (19/2723)

Shutdown the whole cluster and boot each member with:
>>> boot -fl "ia"

'vmunix':vmunix cfs:cluster_root_dev1_maj=19 \
  cfs:cluster_root_dev1_min=2723[Return]

Regards,
Raymond


On Thu, Jun 20, 2002 at 05:32:42PM +0800, Raymond Pang wrote:
> Hi Guru,
>
> We have a 3-member 5.1A cluster. All member boot and cluster root disks
> are sitting in HSZ40. Each is protected by individual mirror set (2x4G).
> Now, we are planning to move all of this to a mirror set (2x36G
> 4-partition) in HSG80.
>
> Could anybody give me some advices on how this could be done safely
> and without interruption to the cluster services?
>
> On the other hand, we are thinking about the redundance of these critical
> disks. We have another HSZ70. So, one option is to put each member boot
> disk on different HSX, and use LSM to mirror the cluster root to other
> HSZ. In case if the whole HSG goes down, cluster still have sufficient
> votes to survive. Does it work or is there any better solution?
>
> Thanks in advance.
>
> Regards,
> Raymond
>
Received on Mon Aug 05 2002 - 07:14:15 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:43 NZDT