2장. Persistent naming attributes

Universally unique identifier (UUID)

UUIDs are primarily used to uniquely identify file systems on storage devices. Each file system instance has its own UUID assigned automatically, and this identifier remains constant even if the file system is unmounted, remounted, or the device is detached and reattached.

Label

Labels are user-assigned names for file systems. While they can be used to identify and reference file systems, they are not as standardized as UUIDs. Since a user assigns the file system label, its uniqueness depends on their choice. Labels are often used as alternatives to UUIDs to specify file systems in configuration files.

When you assign a label to a file system, it becomes part of the file system metadata. This label remains associated with the file system even if it is unmounted, remounted, or the device is detached and reattached.

Universally unique identifier (UUID)

UUIDs can be used to identify storage block devices. When a storage device is formatted, a UUID is often assigned to the device itself. Such UUIDs are usually generated and assigned to virtual block devices layered on top of other block devices, where the real devices are at the bottom level. For example, device-mapper (DM) based devices and their related subsystems, such as Logical Volume Manager (LVM) and crypt, use UUIDs for device identification, such as Logical Volume UUID (LVUUID) and crypt UUID.

Also, multiple-device (MD) based devices have UUIDs assigned. The virtual devices usually also mark the underlying devices with component UUIDs. For example, LVM marks its underlying devices with a Physical Volume UUID (PVUUID) and MD marks its underlying devices with an MD component UUID.

These UUIDs are embedded within the virtual block device metadata and they are used as persistent naming attributes. It allows you to uniquely identify the block device, even if you change the file system that resides on it. UUIDs are also assigned for device partitions.

Such UUIDs can coexist with other device IDs. For example, an sda device at the bottom of the device stack which is identified by its vendor/model/serial_number combination or WWID, can also have a PVUUID assigned by LVM. This is then recognized by LVM itself to build up the Volume Group (VG) or Logical Volume (LV) in a layer above it.

World Wide Identifier (WWID)

WWID encompasses a family of identifiers which are globally unique and they are associated with storage block devices or storage components in general. They are commonly used in enterprise-level Storage Area Networks (SANs), like Fibre Channel (FC), to identify a storage node - World Wide Node Name (WWNN) or actual port/connection to a storage device on the node - World Wide Port Name (WWPN). WWIDs ensure consistent communication between servers and SAN storage devices and help manage redundant paths to storage devices.

Other types of devices may use a form of WWIDs too, like NVME devices. These devices do not necessarily need to be accessed over a network or SAN, and they do not have to be enterprise-level devices either.

The WWID format does not follow a single standard. For example, SCSI uses formats such as NAA, T10, and EUI. The NVME uses formats such as EUI-64, NGUUID, and UUID.

Serial string

The serial string is a unique string identifier assigned to each storage block device by the manufacturer. It can be used to differentiate among storage devices and may be used in combination with other attributes like UUIDs or WWIDs for device management.

Within udev rules, and consequently in the /dev/disk/by-id content, the 'short serial string' typically represents the actual serial string as reported by the device itself. Whereas, 'serial string' is composed of several components, usually <bus_type>-<vendor_name>_<model_name>_<short_serial_string>.

WWIDs and serial strings are preferred for real devices. For virtual devices, UUIDs or names is preferred.