Introduction to device mapper multipathing
Note:
This documentation has moved to a new home! Please update your bookmarks to the new URL for the up-to-date version of this page.
Device mapper multipathing (which we will refer to as multipath) allows you to configure multiple input/output (I/O) paths between server nodes and storage arrays into a single device. These I/O paths are physical storage area network (SAN) connections that can include separate cables, switches, and controllers.
Multipathing aggregates the I/O paths, creating a new device that consists of those aggregated paths. This page provides an introduction and a high-level overview of multipath.
Benefits of multipath
Multipath can be used to provide:
-
Redundancy
Multipath provides failover in an active/passive configuration. In an active/passive configuration, only half the paths are used at any time for I/O. If any element of an I/O path (the cable, switch, or controller) fails, multipath switches to an alternate path. -
Improved performance
Multipath can be configured in active/active mode, where I/O is spread over the paths in a round-robin fashion. In some configurations, multipath can detect loading on the I/O paths and dynamically re-balance the load.
Storage array overview
It is a good idea to consult your storage vendor’s installation guide for the recommended multipath configuration variables for your storage model. The default configuration will probably work but will likely need adjustments based on your storage setup.
Multipath components
Component | Description |
---|---|
dm_multipath kernel module |
Reroutes I/O and supports failover for paths and path groups. |
multipath command |
Lists and configures multipath devices. Normally started up with /etc/rc.sysinit , it can also be started up by a udev program whenever a block device is added, or it can be run by the initramfs file system. |
multipathd daemon |
Monitors paths; as paths fail and come back, it may initiate path group switches. Provides for interactive changes to multipath devices. This daemon must be restarted for any changes to the /etc/multipath.conf file to take effect. |
kpartx command |
Creates device-mapper devices for the partitions on a device. It is necessary to use this command for DOS-based partitions with multipath. The kpartx is provided in its own package, but the multipath-tools package depends on it. |
Multipath setup overview
Setting up multipath is often a simple procedure, since it includes compiled-in default settings that are suitable for common multipath configurations. The basic procedure for configuring your system with multipath is as follows:
-
Install the
multipath-tools
andmultipath-tools-boot
packages -
Create an empty config file called
/etc/multipath.conf
-
Edit the
multipath.conf
file to modify default values and save the updated file -
Start the multipath daemon
-
Update initial RAM disk
For detailed setup instructions for multipath configuration see DM-Multipath configuration and DM-Multipath setup.
Multipath devices
Without multipath, each path from a server node to a storage controller is treated by the system as a separate device, even when the I/O path connects the same server node to the same storage controller. Multipath provides a way of organising the I/O paths logically, by creating a single device on top of the underlying paths.
Multipath device identifiers
Each multipath device has a World Wide Identifier (WWID), which is guaranteed to be globally unique and unchanging. By default, the name of a multipath device is set to its WWID. Alternatively, you can set the user_friendly_names
option in multipath.conf
, which causes multipath to use a node-unique alias of the form mpathn
as the name.
For example, a node with two host bus adapters (HBAs) attached to a storage controller, with two ports, via a single un-zoned FC switch sees four devices: /dev/sda
, /dev/sdb
, /dev/sdc
, and /dev/sdd
. Multipath creates a single device with a unique WWID that reroutes I/O to those four underlying devices according to the multipath configuration.
When the user_friendly_names
configuration option is set to ‘yes’, the name of the multipath device is set to mpathn
. When new devices are brought under the control of multipath, the new devices may be seen in two different places under the /dev
directory: /dev/mapper/mpathn
and /dev/dm-n
.
-
The devices in
/dev/mapper
are created early in the boot process. Use these devices to access the multipathed devices. -
Any devices of the form
/dev/dm-n
are for internal use only and should never be used directly.
You can also set the name of a multipath device to a name of your choosing by using the alias
option in the multipaths
section of the multipath configuration file.
See also:
For information on the multipath configuration defaults, including theuser_friendly_names
andalias
configuration options, see DM-Multipath configuration.
Consistent multipath device names in a cluster
When the user_friendly_names
configuration option is set to ‘yes’, the name of the multipath device is unique to a node, but it is not guaranteed to be the same on all nodes using the multipath device. Similarly, if you set the alias
option for a device in the multipaths
section of /etc/multipath.conf
, the name is not automatically consistent across all nodes in the cluster.
This should not cause any difficulties if you use LVM to create logical devices from the multipath device, but if you require that your multipath device names be consistent in every node it is recommended that you leave the user_friendly_names
option set to ‘no’ and that you do not configure aliases for the devices.
If you configure an alias for a device that you would like to be consistent across the nodes in the cluster, you should ensure that the /etc/multipath.conf
file is the same for each node in the cluster by following the same procedure:
-
Configure the aliases for the multipath devices in the in the
multipath.conf
file on one machine. -
Disable all of your multipath devices on your other machines by running the following commands as root:
systemctl stop multipath-tools.service multipath -F
-
Copy the
/etc/multipath.conf
file from the first machine to all other machines in the cluster. -
Re-enable the
multipathd
daemon on all the other machines in the cluster by running the following command as root:systemctl start multipath-tools.service
Whenever you add a new device you will need to repeat this process.
Multipath device attributes
In addition to the user_friendly_names
and alias
options, a multipath device has numerous attributes. You can modify these attributes for a specific multipath device by creating an entry for that device in the multipaths
section of /etc/multipath.conf
.
For information on the multipaths
section of the multipath configuration file, see DM-Multipath configuration.
Multipath devices in logical volumes
After creating multipath devices, you can use the multipath device names just as you would use a physical device name when creating an LVM physical volume.
For example, if /dev/mapper/mpatha
is the name of a multipath device, the following command (run as root) will mark /dev/mapper/mpatha
as a physical volume:
pvcreate /dev/mapper/mpatha
You can use the resulting LVM physical device when you create an LVM volume group just as you would use any other LVM physical device.
Note:
If you try to create an LVM physical volume on a whole device on which you have configured partitions, thepvcreate
command will fail.
When you create an LVM logical volume that uses active/passive multipath arrays as the underlying physical devices, you should include filters in the lvm.conf
file to exclude the disks that underlie the multipath devices. This is because if the array automatically changes the active path to the passive path when it receives I/O, multipath will failover and fallback whenever LVM scans the passive path if these devices are not filtered.
For active/passive arrays that require a command to make the passive path active, LVM prints a warning message when this occurs. To filter all SCSI devices in the LVM configuration file (lvm.conf
), include the following filter in the devices section of the file:
filter = [ "r/block/", "r/disk/", "r/sd.*/", "a/.*/" ]
After updating /etc/lvm.conf
, it’s necessary to update the initrd
so that this file will be copied there, where the filter matters the most – during boot. Perform:
update-initramfs -u -k all
Note:
Every time either/etc/lvm.conf
or/etc/multipath.conf
is updated, theinitrd
should be rebuilt to reflect these changes. This is imperative when denylists and filters are necessary to maintain a stable storage configuration.