When A Raid Controller Module Is Replaced Or Removed; Raid Controller Failover Modes - Dell PowerVault MD3000 Hardware Owner's Manual

When a RAID Controller Module Is Replaced or Removed

When one RAID controller module fails, or is removed from the RAID enclosure, failover software on the
host directs the alternate RAID controller module to assume all controller responsibilities for the storage
array. The virtual disks owned by the failed or removed RAID controller module are failed over to the
surviving controller and no data is lost.
Since host-based multi-path software is configured for automatic failback, once the RAID controller
module is replaced, its data paths and virtual disks are restored to their original RAID controller module.
Automatic failback occurs under the following conditions:
• Data is protected by dual redundant paths to the host.
• The host has multi-path failover software installed, and it is configured for auto failback.
• Failover occurred due to either RAID controller module lockdown, replacement, or path failover.
For more information on failover, see "Cabling the Enclosure" on page 20.

RAID Controller Failover Modes

Failover provides redundant access to the virtual disks owned by either RAID controller module on the
storage array. Failover happens automatically (and transparently). As a result, the host(s) never loses
access through the preferred controller which owns the virtual disk.
Two failover modes are supported:
• Host-based Multi-Path I/O (MPIO) for Windows or MPP for Linux
• Controller-based Auto Virtual Disk Transfer (AVT)
Host-based uses a multi-path driver installed on the host server to access the storage array. If required,
the multi-path driver will issue an explicit command to transfer ownership from the RAID controller
module that owns the virtual disk to its peer RAID controller module.
AVT is used only in single-port cluster solutions (see Figure 2-6). I/O access to the Logical Unit Number
(LUN) of a virtual disk causes failover. In AVT mode, firmware transfers ownership of a virtual disk to the
online RAID controller module. The alternate or backup node in a cluster takes over and issues I/O to the
peer controller and moves the virtual disk to itself.
NOTE:
Clustering requires simultaneous access from cluster nodes to shared storage. If you have clustering
software installed on the host, automatic failback should be disabled to prevent "ping-pong" with single-path
failure.
NOTE:
If setting up a cluster host, the MD3000 Stand Alone to Cluster.reg file entry must be merged into the
registry of each node. If re-configuring a cluster node to a standalone host, the MD3000 Cluster to Stand Alone.reg
file must be merged into the host registry. These registry files, which set up the host for correct failback operation,
are in the windows\utility directory of the MD3000 Resource CD.
44
Using Your RAID Enclosure