Replace one of the physical volumes in an LVM volume group
Tested on |
Debian (Lenny, Squeeze) |
Ubuntu (Hardy, Interpid, Jaunty, Karmic, Lucid, Maverick, Natty, Oneiric, Precise) |
Objective
To replace one of the physical volumes within an existing LVM volume group
Scenario
Suppose that /dev/vg0
is a volume group composed of a single physical volume located on a hard drive with a capacity of 100GB. You wish to replace this with a new hard drive having a capacity of 200GB. You have been able to temporarily connect both drives to the system, such that the old drive is presented as /dev/sdb
and the new drive as /dev/sdc
.
Method
Overview
The method described here has five steps:
- Optionally, initialise the new storage device as a physical volume.
- Add the new physical volume to the volume group.
- Migrate all data located on the old physical volume to the new physical volume.
- Remove the old physical volume from the volume group.
- Optionally, wipe the label from the old storage device to prevent it from being detected as a physical volume.
The corresponding sequence of commands is as follows:
pvcreate /dev/sdc vgextend vg0 /dev/sdc pvmove /dev/sdb /dev/sdc vgreduce vg0 /dev/sdb pvremove /dev/sdb
Be aware that any operation of this nature carries some risk of data loss. This is unlikely if the hardware is in good working order, but it would be prudent to make a backup of the relevant volume group before starting.
Optionally, initialise the new storage device as a physical volume
With older versions of LVM it was necessary for physical volumes to be explicitly initialised using pvcreate
before being added to
a volume group:
pvcreate /dev/sdc
As of version 2.02.54 this is no longer necessary because initialisation will occur automatically if required. Prior initialisation may still be desirable in order to deviate from the default settings used by pvcreate
, obtain better diagnostics by proceeding one step at a time, or retain compatibility with older versions of LVM.
Add the new physical volume to the volume group
A physical volume can be added to a volume group using the vgextend
command:
vgextend vg0 /dev/sdc
The first argument is the name of the volume group to be extended. This can be written as a pathname if you prefer (/dev/vg0
). Subsequent arguments are the physical volumes to be added.
If successful you should see a response of the form:
Volume group "vg0" successfully extended
If the physical volumes have not previously been initialised using pvcreate
then there will be some additional diagnostic messages, for example:
No physical volume label read from /dev/sdc
Physical volume "/dev/sdc" successfully created
Volume group "vg0" successfully extended
You can check that the physical volume has been added using the pvs
command:
pvs
If successful then you should see a response of the form:
PV VG Fmt Attr PSize PFree /dev/sdb vg0 lvm2 a- 100.00g 0 /dev/sdc vg0 lvm2 a- 200.00g 200.00g
The VG
column indicates which volume group each physical volume is a member of (if any). In this instance it shows (as expected) that both /dev/sdb
and /dev/sdc
are members of vg0
.
Migrate all data located on the old physical volume to the new physical volume.
Data can be transferred from one physical volume to another using the pvmove
command:
pvmove /dev/sdb /dev/sdc
The first argument is the physical volume to be emptied. The second argument is the physical volume to which the content should be moved. For large storage devices the transfer can take a considerable amount of time, however the machine should remain usable during this period so pvmove
can be left to run in the background. It is safe to continue using filing systems that are wholly or partly located within the physical volume that is being moved.
pvmove
periodically reports the progress it has made as a percentage, returning control when the transfer is complete. If it is interrupted for any reason then the transfer can be resumed by executing pvmove
again with no arguments.
You can check that the migration was successful using the pvs
command again. The response should show that the source physical volume contains no data (and therefore that PFree
is equal to PSize
):
PV VG Fmt Attr PSize PFree /dev/sdb vg0 lvm2 a- 100.00g 100.00g /dev/sdc vg0 lvm2 a- 200.00g 100.00g
Remove the old physical volume from the volume group.
A Physical volume can be removed from a volume group using the vgreduce
command:
vgreduce vg0 /dev/sdb
The first argument is the name of the volume group. Subsequent arguments are the names of physical volumes to be removed. If successful you should see a response of the form:
Removed "/dev/sdb" from volume group "vg0"
You can check that the physical volume has been removed using the pvs
command again. The response should show that the source physical volume is no longer a member of any volume group:
PV VG Fmt Attr PSize PFree /dev/sdb lvm2 a- 100.00g 100.00g /dev/sdc vg0 lvm2 a- 200.00g 100.00g
Optionally, wipe the label from the old storage device to prevent it from being detected as a physical volume
LVM will continue to recognise the old storage device as a physical volume (albeit an empty one) unless you take explicit action to wipe the label that was written by pvcreate
. This can be done using the pvremove
command:
pvremove /dev/sdb
If successful you should see a response of the form:
Labels on physical volume "/dev/sdb" successfully wiped
Leaving the label in place is not necessarily harmful, but it can cause confusion in some circumstances. For example, repartitioning a hard drive can result in LVM discovering physical volumes that are the wrong size for the drive layout. Use of pvremove
is therefore recommended unless there is a reason not to.
Troubleshooting
See Troubleshooting LVM.
Errors
Physical volume not found in volume group
An error of the form:
Physical volume "/dev/sdc" not found in Volume Group "vg0"
in response to pvmove
indicates that the source and destination physical volumes do not belong to the same volume group. The purpose of pvmove
is to move data physically without affecting how it is presented logically. For this reason it can only move data within a volume group, not between volume groups.
If the new physical volume has merely not been added to the correct volume group then you can correct this using vgextend
, as described above. If it has been added to the wrong volume group then you will first need to remove it using vgreduce
.
See also
- Increase the capacity of an LVM volume group
- Increase the size of an LVM logical volume
- Increase the size of an ext2, ext3 or ext4 filesystem
Tags: lvm