Re-Initialize an MCIP with Less Shelves

on

Recently I had to re-initialize an IP MetroCluster. Reason being that the aggregate layout required a certain size of partitioned disks - partitioning across 2 shelves per site but I'd initialized it with 3 shelves attached, so the root partitions were not to the customers standard. Never mind, these things happen.

Note: In AFF IP MetroCluster you can have partitioning across 4 shelves (96 disks) because we mirror from site to site (still maximum of 48 disk per pool can be partitioned.)

Preparation Steps:

1) Make sure there is no data (no data SVMs) on both sides of the MetroCluster you are going to re-initialize.

vserver show 

2) Power off the entire MetroCluster as per:

Powering off an entire MetroCluster IP configuration

Essentially (taking precautions for encryption and tie breaker):

system node halt -node * -inhibit-takeover true -ignore-quorum-warnings true

Re-Initialization Steps (going for 3 to 2 shelves):

  1. Wait for all nodes to be at the ADP menu (option 9 at the boot menu)
  2. Run 9a on all nodes
  3. Return to main boot menu and go into maintenance mode
  4. Check everything has been blown away (disk show - all disks should have been unassigned)
  5. halt the nodes
  6. disconnect 3rd shelf
  7. boot to boot menu again
  8. Run 9a and 9b on all nodes
  9. [Optional] Disable disk auto assign and hot add the 3rd shelf if desired.

Comments

Post a Comment