An Introduction to the Logical Volume Manager

To use LVM, a disk must be first initialized into a physical volume (also called an LVM disk).

Once you have initialized one or more physical volumes, you assign them into one or more volume groups. If you think of all of your physical volumes as forming a storage pool, then a subset of disks from the pool can be joined together into a volume group.

A given disk can only belong to one volume group. The maximum number of volume groups that can be created is determined by the configurable parameter maxvgs. See "Reconfiguring the Kernel" in Chapter 1 for information on modifying system parameters.

A volume group can contain from one to 255 physical volumes.

Disk space from the volume group is allocated into a logical volume, a distinct unit of usable disk space. A volume group can contain up to 255 logical volumes.

A logical volume can exist on only one disk or can reside on portions of many disks.

The disk space within a logical volume can be used for swap, dump, raw data, or you can create a file system on it.

In Figure 3-1 “Disk Space Partitioned into Logical Volumes ”, logical volume /dev/vg01/lvol1 might contain a file system, /dev/vg01/lvol2 might contain swap space, and /dev/vg01/lvol3 might contain raw data. As the figure illustrates, a file system, swap space, or raw data area may exist within a logical volume that resides on more than one disk.

Figure 3-1 Disk Space Partitioned into Logical Volumes

If a logical volume spans multiple physical volumes, it is not required that each disk be of the same interface type except in the case of HP-IB disks; however, having the same interface type will result in better performance. See "Using Disk I/O Interfaces" later in this chapter for more information on interface types and limitations.

LVM divides each physical disk into addressable units called physical extents. Extents are allocated to disks sequentially starting from the beginning of the disk with address zero, and incrementing the address by one for each unit. Physical extent size is configurable at the time you form a volume group and applies to all disks in the volume group. By default, each physical extent has a size of 4 megabytes (MB). This value can be changed when you create the volume group to a value between 1MB and 256MB.

The basic allocation unit for a logical volume is called a logical extent. A logical extent is mapped to a physical extent; thus, if the physical extent size is 4MB, so will be the logical extent size. The size of a logical volume is determined by the number of logical extents configured.

When LVM allocates disk space to a logical volume, it automatically creates a mapping of the physical extents to logical extents. Logical extents are also allocated sequentially, starting at zero, for each logical volume. Therefore, regardless of where the actual physical data resides for a logical volume within a volume group, LVM will use this mapping to access the data. Commands are provided for you to examine this mapping; see pvdisplay(1M) and lvdisplay(1M).

Except for mirrored or striped logical volumes, each logical extent is mapped to one physical extent. For mirrored logical volumes, either two or three physical extents are mapped for each logical extent depending upon whether you are using single or double mirroring. For example, if one mirror copy exists, then each logical extent maps to two physical extents, one extent for the original and one for the mirror copy. (See Chapter 7 for more information on mirroring.) For information on striped logical volumes, see Chapter 8.

Figure 3-2 “Physical Extents and Logical Extents ” on the following page shows an example of several types of mapping available between physical extents and logical extents within a volume group.

Figure 3-2 Physical Extents and Logical Extents

As can be seen in Figure 3-2, the contents of the first logical volume are contained on all three physical volumes in the volume group. Since the second logical volume is mirrored, each logical extent is mapped to more than one physical extent. In this case, there are two physical extents containing the data, each on both the second and third disks within the volume group.

By default, LVM assigns physical extents to logical volumes by selecting available physical extents from disks in the order you added physical volumes to the volume group. As a system administrator, you can bypass this default assignment and control which disks are used by a logical volume (see "Spanning Disks With File Systems" in this chapter).

If a logical volume is to be used for root, primary swap, or dump, the physical extents must be contiguous. This means that the physical extents must be allocated with no gaps on a single physical volume. On non-root disks, physical extents that correspond to contiguous logical extents within a logical volume can be non-contiguous on a physical volume or reside on entirely different disks. As a result, it becomes possible for a file system created within one logical volume to reside on more than one disk.