Skip to content

Latest commit

 

History

History
86 lines (70 loc) · 2.92 KB

aix_disk_mig.md

File metadata and controls

86 lines (70 loc) · 2.92 KB
Raw

Disk Migration

Migration Scenario 1

A fairly complicated scenario, with 2 node cluster mirrored to a remote site. Scenario involves migration disks from one storage server type to another using LVM mirroring. Remote mirroring splits and re-syncs can be inserted into these procedures as appropriate.

Beforehand, cfgmgr -vl fcsxx for every fcs adapter, check new disks all present and correct.

On all 4 nodes:

  • stop the cluster (all nodes and resource groups)
  • unmount filesystems and varyoff vgs

Nodes 1 & 2:

  • remove devices & driver software
  • install new driver software & cfgmgr new devices
  • check new disk device settings as per standards/recommendations with lsattr

Node 2:

  • varyonvg vgname
  • extendvg vgname new_disk...
  • mirrorvg -s vgname, (-s means don't start the sync yet)
  • syncvg -P32 -v vgname, uses max parallel LPs
  • unmirrorvg vgname old_disk...
  • reducevg vgname old_disk...
  • For heartbeat disks, there may be custom disk methods for different storage vendors, break reservations. See cllscustomdisk command to view/change these.

Node 1:

  • exportvg old_vgs
  • chdev -l new_disk -a pv=yes, to get new pvid on shared disks
  • importvg -V majornum -y vgname new_disk, using major numbers from node 2 vgs

Node 2:

  • add new heartbeat disks to HA config via smitty
  • remove old heartbeat disks from HA config
  • sync cluster

Nodes 1 & 2:

  • use dhb_read to check hb disks communicating properly
  • start cluster & carry out tests
  • cleanup reboot if necessary

Nodes 3 & 4, once disk mirroring in sync (if it was split before the maintenance started):

  • Replicate the Nodes 1 & 2 steps to delete the old devices & driver software and install/configure the new ones.
  • Replicate the Node 1 steps to import the vgs from the new disks, using the same major numbers.

Node 4:

  • Replicate the steps to remove the old HB disks from the HA config and add the new ones.
  • Sync the cluster.
  • Start the cluster and carry out some testing.

Migration Scenario 2

Scenario: client lpar with SAN disks presented via vscsi. The SAN disks are cloned by storage server and the copies need to be presented to the client instead of the originals.

  • bootlist -m normal -ov
  • note the boot device path, e.g. /vdevice/v-scsi@30000002/disk@8400000000000000:2
  • shutdown client lpar
  • delete mappings and disks from VIO
  • clone disks
  • configure new disks on VIO, check attributes
  • map new disks to client lpar
  • lsmap -all | grep -p {boot_disk}
  • note the LUN location 0x8100000000000000
  • boot lpar to open firmware prompt
  • use the old path, substituting the long LUN number found in lsmap
  • boot /vdevice/v-scsi@30000002/disk@8100000000000000:2
  • check disk attrs on lpar
  • check and fix bootlist in lpar:
    • bootlist -o -m normal and compare with before
    • bootinfo -v to get blv, e.g. hd5
    • lslv -l hd5 to get blv hdisk
    • bootlist -m normal {hdisk} blv=hd5 # watch for multiple blvs
  • remove defined disks