Hot-removing storage from a MetroCluster FC configuration

You can hot-remove drive shelves—physically remove shelves that have had the aggregates removed from the drives—from a MetroCluster FC configuration that is up and serving data. You can hot-remove one or more shelves from anywhere within a stack of shelves or remove a stack of shelves.

Before you begin

Your system must be a multipath HA, multipath, quad-path HA, or quad-path configuration.
In a four-node MetroCluster FC configuration, the local HA pair cannot be in a takeover state.
You must have already removed all aggregates from the drives in the shelves that you are removing.
Attention
If you attempt this procedure on non-MetroCluster FC configurations with aggregates on the shelf you are removing, you could cause the system to fail with a multidrive panic.
Removing aggregates involves splitting the mirrored aggregates on the shelves you are removing, and then re-creating the mirrored aggregates with another set of drives.
Disk and Aggregate Management Guide
You must have removed drive ownership after removing the aggregates from the drives in the shelves that you are removing.
Disk and Aggregate Management Guide
If you are removing one or more shelves from within a stack, you must have factored the distance to bypass the shelves that you are removing.
If the current cables are not long enough, you need to have longer cables available.

About this task

This task applies to the following MetroCluster FC configurations:

Direct-attached MetroCluster FC configurations, in which the storage shelves are directly connected to the storage controllers with SAS cables
Fabric-attached or bridge-attached MetroCluster FC configurations, in which the storage shelves are connected using FC-to-SAS bridges

Verify the operation of the MetroCluster configuration in ONTAP:
1. Check whether the system is multipathed: node run -node node-name sysconfig -a
2. Check for any health alerts on both clusters: system health alert show
3. Confirm the MetroCluster configuration and that the operational mode is normal: metrocluster show
4. Perform a MetroCluster check: metrocluster check run
5. Display the results of the MetroCluster check:metrocluster check show
6. Check for any health alerts on the switches (if present): storage switch show
Set the privilege level to advanced: set -privilege advanced
Verify that no mailbox drive is on the shelves: storage failover mailbox-disk show

Remove the shelf according to the steps for the relevant scenario.

Scenario 1: To remove an aggregate when the shelf contains either unmirrored, mirrored, or both types of aggregate...

Steps:

Use the storage aggregate delete -aggregate aggregate name command to remove the aggregate.
Use the standard procedure to remove ownership of all drives in that shelf, and then physically remove the shelf.
Follow the instructions in the Service Guide for DM120S, DM240S, and DM600S to hot-remove expansion enclosures.

Scenario 2: To remove an aggregate when the shelf contains either unmirrored, mirrored, or both types of aggregate...

Steps: To remove a plex from a mirrored aggregate, you need to unmirror the aggregate.

Identify the plex that you want to remove by using the run -node local sysconfig -r command.

In the following example, you can identify the plex from the line Plex /dpg_mcc_8020_13_a1_aggr1/plex0. In this case, the plex to specify is plex0.

dpgmcc_8020_13_a1a2::storage aggregate> run -node local sysconfig -r
*** This system has taken over dpg-mcc-8020-13-a1
Aggregate dpg_mcc_8020_13_a1_aggr1 (online, raid_dp, mirrored) (block checksums)
 Plex /dpg_mcc_8020_13_a1_aggr1/plex0 (online, normal, active, pool0)
  RAID group /dpg_mcc_8020_13_a1_aggr1/plex0/rg0 (normal, block checksums)
   RAID Disk Device                          HA  SHELF BAY CHAN Pool Type  RPM  Used (MB/blks)    Phys (MB/blks)
   --------- ------                          ------------- ---- ---- ---- ----- --------------    --------------
   dparity   mcc-cisco-8Gb-fab-2:1-1.126L16  0c    32  15  FC:B   0   SAS 15000 272000/557056000  274845/562884296
   parity    mcc-cisco-8Gb-fab-2:1-1.126L18  0c    32  17  FC:B   0   SAS 15000 272000/557056000  274845/562884296
   data      mcc-cisco-8Gb-fab-2:1-1.126L19  0c    32  18  FC:B   0   SAS 15000 272000/557056000  274845/562884296
   data      mcc-cisco-8Gb-fab-2:1-1.126L21  0c    32  20  FC:B   0   SAS 15000 272000/557056000  274845/562884296
   data      mcc-cisco-8Gb-fab-2:1-1.126L22  0c    32  21  FC:B   0   SAS 15000 272000/557056000  274845/562884296

 Plex /dpg_mcc_8020_13_a1_aggr1/plex1 (online, normal, active, pool1)
  RAID group /dpg_mcc_8020_13_a1_aggr1/plex1/rg0 (normal, block checksums)

   RAID Disk Device                          HA  SHELF BAY CHAN Pool Type  RPM  Used (MB/blks)    Phys (MB/blks)
   --------- ------                          ------------- ---- ---- ---- ----- --------------    --------------
   dparity   mcc-cisco-8Gb-fab-3:1-1.126L37  0d    34  10  FC:A   1   SAS 15000 272000/557056000  280104/573653840
   parity    mcc-cisco-8Gb-fab-3:1-1.126L14  0d    33  13  FC:A   1   SAS 15000 272000/557056000  280104/573653840
   data      mcc-cisco-8Gb-fab-3:1-1.126L41  0d    34  14  FC:A   1   SAS 15000 272000/557056000  280104/573653840
   data      mcc-cisco-8Gb-fab-3:1-1.126L15  0d    33  14  FC:A   1   SAS 15000 272000/557056000  280104/573653840
   data      mcc-cisco-8Gb-fab-3:1-1.126L45  0d    34  18  FC:A   1   SAS 15000 272000/557056000  280104/573653840

Use the storage aggregate plex delete -aggregate aggr_name -plex plex_name command to remove the plex.
plex defines the plex name, such as plex3 or plex6.
Use the standard procedure to remove ownership of all drives in that shelf, and then physically remove the shelf.
Follow the instructions in the Service Guide for DM120S, DM240S, and DM600S to hot-remove expansion enclosures.

Give documentation feedback