18.8 About Device Multipathing

Multiple paths to storage devices can provide connection redundancy, failover capability, load balancing, and improved performance. Device-Mapper Multipath (DM-Multipath) is a multipathing tool that allows you to represent multiple I/O paths between a server and a storage device as a single path.

You would be most likely to configure multipathing with a system that can access storage on a Fibre Channel-based storage area network (SAN). You can also use multipathing on an iSCSI initiator if redundant network connections exist between the initiator and the target.

Figure 18.2 shows a simple DM-Multipath configuration where two I/O paths are configured between a server and a disk on a SAN-attached storage array:

  • Between host bus adapter hba1 on the server and controller ctrl1 on the storage array.

  • Between host bus adapter hba2 on the server and controller ctrl2 on the storage array.

Figure 18.2 DM-Multipath Mapping of Two Paths to a Disk over a SAN

The diagram shows a simple DM-Multipath configuration where two I/O paths are configured between a server and a disk on a SAN-attached storage array. DM-Multipath creates a single multipath device, /dev/mapper/mpathN, that subsumes the underlying devices, /dev/sdc and /dev/sdf.


Without DM-Multipath, the system treats each path as being separate even though it connects the server to the same storage device. DM-Multipath creates a single multipath device, /dev/mapper/mpathN, that subsumes the underlying devices, /dev/sdc and /dev/sdf.

You can configure the multipathing service (multipathd) to handle I/O from and to a multipathed device in one of the following ways:

Active/Active

I/O is distributed across all available paths, either by round-robin assignment or dynamic load-balancing.

Active/Passive (standby failover)

I/O uses only one path. If the active path fails, DM-Multipath switches I/O to a standby path. This is the default configuration.

Note

DM-Multipath can provide failover in the case of path failure, such as in a SAN fabric. Disk media failure must be handled by using either a software or hardware RAID solution.