Logical Volume Manager and MirrorDisk/UX

Increased logical volume size: LVM now supports logical volumes up to 16 terabytes (16TB) in size, an increase from 2TB in previous releases. Logical volumes larger than 2TB may not be usable on previous releases; see the compatibility note below.

The tunable parameter maxvgs has been obsoleted; LVM can now create up to 256 volume groups dynamically.

LVM Device Online Replacement (OLR): The pvchange command has a new option, -a, that can be used to temporarily stop LVM accessing a device or device special file, and to reenable access. A white paper (see “Documentation” below) explains how OLR simplifies the process of replacing or isolating disks. (This feature was introduced in HP-UX 11i Version 2 June 2006.)

Volume Group Quiesce/Resume: Two new options, -Q and -R, have been added to the vgchange command. These options temporarily quiesce an activated volume group and resume it, respectively, and are discussed in detail on the vgchange(1M) manpage.

Using the -Q option, you can quiesce both read and write operations to the volume group, or just write operations. While the volume group is quiesced, the vgdisplay command will report the volume group’s status as “quiesced.” The indicated I/O operations will be queued until the volume group is resumed, and commands that would modify the volume group configuration will return an error. This keeps the volume group disk image in a stable state suitable for using disk management utilities to perform snapshots of all the disks in the volume group, without deactivating the volume group.

Note that the entire volume group is quiesced. Individual logical volumes or physical volumes cannot be quiesced using this feature. To disable or replace a physical volume or PVLink, use the pvchange command. To provide a stable image of a particular logical volume, in order to back up the data there, use the lvsplit command.

Striping with Mirroring: In previous releases, LVM supported a limited extent-based striped mirror functionality, as described in the lvcreate(1M) manpage. This type of striped mirror required the stripe size to be a multiple of extent size. In HP-UX 11i v3, LVM supports mirroring of striped logical volumes with the entire range of stripe sizes. The lvcreate command options -m and -i/-I can be used together, and the -m option of the lvextend and lvreduce commands can be applied to striped logical volumes. Note that logical volumes configured with such striped mirrors cannot be imported on releases before HP-UX 11i v3; see the compatibility note below.

To maximize data Integrity for striped mirrors, LVM enforces a strict allocation policy; that is, mirrored physical extents must be allocated on different physical volumes. This forces the physical extents used by all the replicas of a strip onto different physical volumes.

Mass Storage Stack: LVM supports the next generation mass storage stack, as described in the Mass Storage Stack section of this document. In particular, LVM supports the use of both legacy and persistent device special files within the same volume group. New options to the vgscan and vgimport commands, described below, affect how LVM creates its configuration.

By default, vgscan recovers the LVM configuration information (the /etc/lvmtab file) using legacy device special files. If the new -N option is specified, then vgscan will use persistent device special files. If the new -B option is specified, then vgscan will populate the /etc/lvmtab file using both legacy and persistent device special files.

By default, when importing a volume group in shared mode, vgimport will populate the /etc/lvmtab file using legacy device special files. If the new -N option is specified, then vgimport will use persistent device special files.

Multi-pathing and Alternate Links (PVLinks): Management of multi-pathed devices is available outside of LVM using the next generation mass storage stack. By default, the mass storage stack balances the I/O load across all available paths to a disk. However, the new scsimgr command can be used to emulate LVM’s PVLink functionality, and offers additional options for handling LUN failure and load balancing.

HP recommends converting volume groups with multi-pathed disks to persistent device special files. This can be done by running the /usr/contrib/bin/vgdsf, vgscan -N, or vgimport -s -N commands.

SLVM Single Node Online Volume Reconfiguration (SNOR): A new option, -x, has been added to the vgchange command. This option allows an administrator to make configuration changes to a shared volume group while keeping the volume group activated on one cluster node. A white paper, described below, explains the SNOR functionality.

Dynamic Volume Group Modification: A new command, vgmodify, is available to dynamically modify a volume group’s characteristics. In previous releases, the number of physical volumes, the number of logical volumes, and number of physical extents per disk were set when a volume group was created; the vgmodify command allows these parameters to be modified without recreating the volume group.

Dynamic LUN Expansion: If the administrator increases the size of a LUN, the vgmodify command can be used to incorporate that additional space into the volume group without recreating the volume group.

Boot resiliency: If during boot time the LVM subsystem detects an inconsistency between the firmware boot path and the LVM root volume group configuration, LVM will scan all the disk devices to find the physical volumes belonging to the root volume group and will continue the boot sequence. In previous releases, the administrator had to boot in LVM Maintenance Mode to resolve this inconsistency.

Display enhancements: The lvdisplay, pvdisplay, vgdisplay, and vgscan commands all support the long hostnames described in “Long hostname, uname, and setuname”. Those commands also support a new -F option to print in a format more easily parsed by user scripts. The pvdisplay command has a new -l option to display if a disk is under LVM control; the existing -d option displays if a physical volume is a bootable physical volume.

vgscan enhancements: In addition to support of the mass storage stack, vgscan has new -f and -k options.

By default, the vgscan command does not modify or supplement the /etc/lvmtab file entries for volume groups that already have entries. The new -f option forces an update of the existing entries for the specified volume group.

By default, the vgscan command scans the I/O configuration searching for LVM physical volumes, which can be a time-consuming operation. The new -k option reads the LVM data structures in kernel memory and populates the /etc/lvmtab file based on that data.