How to create partitions and file systems on DM-Multipath devices?

Updated: 22 October 2026 · Applies to: Device-Mapper Multipath (DM-Multipath) on Linux

DM-Multipath combines multiple physical I/O paths between a server and a storage array (separate cables, switches and controllers) into one aggregated device, for redundancy and throughput. These devices appear as /dev/mapper/mpathN, where N is the multipath group number.

Create a partition on a DM-Multipath device

Use fdisk, with n for a new partition and p for primary:

fdisk /dev/mapper/mpath0
Command (m for help): n
Command action
   e   extended
   p   primary partition (1-4)
p

Give the partition number, the first cylinder (the default is usually fine) and the last cylinder or a size:

Partition number (1-4): 1
First cylinder (1-1017, default 1):
Using default value 1
Last cylinder or +size or +sizeM or +sizeK (1-1017, default 1017):
Using default value 1017

Write the partition table to disk with w:

Command (m for help): w
The partition table has been altered!
Calling ioctl() to re-read partition table.

Register the new partitions under /dev/mapper:

kpartx -a /dev/mapper/mpath0

List the partitions on the device to confirm:

kpartx -l /dev/mapper/mpath0
mpath0p1 : 0 2295308 /dev/mapper/mpath0 61

Create a filesystem on the new partition

mkfs.ext4 /dev/mapper/mpath0p1

Then create a mount point and mount it:

mkdir /data
mount /dev/mapper/mpath0p1 /data

Which device name to use

DM-Multipath exposes the same device under three different naming schemes:

1. /dev/dm-N
2. /dev/mpath/mpathN
3. /dev/mapper/mpathN

Never use /dev/dm-N: it is reserved for DM-Multipath's own internal use. Never use /dev/mpath/mpathN either: at boot time, udev may not create these device nodes quickly enough, which can cause the mount to fail intermittently.

Always use /dev/mapper/mpathN: these names are persistent and created automatically by device-mapper early in boot, so they are the names you should actually reference. In a clustered configuration, /dev/mapper/mpathN names are persistent on one machine across reboots, but there is no guarantee every node in the cluster assigns the same name to the same disk; use udev's own persistent-naming facility if you need a name that matches across every node.