Tag Archives: pvs

LVM – How to

Intro

LVM is a very powerful technology, and can really help the Sysadmin’s life.
However, this is something that we generally setup at the beginning (most of the time now it’s automatically setup during the installation process), and it’s well know… when we stop using something, we tend to forget how to use it.

This is why I’m writing this how to, mostly to keep track of the major features and commands, in case I will need them again in the future šŸ˜‰

Before proceeding, please digest the following journey of this poor physical device that gets abstracted up to usableĀ pieces.

 

 

Prepare partions

First of all, we need to find which device(s) we want to setup for LVM

fdisk -l

We can see 3 md devices, probably RAID devices. These are the ones that we are going to use for our LVM exercise.

Now, let’s create an LVM partition.

fdisk <device>Ā => n , p , 1 , (enter) , (enter) , t , 8e , w

Do the same for all the devices that you want to use for LVM. In my example, I’ve done this for /dev/md1, /dev/md2 and /dev/md3.

Shortcut (risky but quicker) šŸ™‚

All seems now good to go: we have Linux LVM partitions!

Time to start to configure LVM

Configure LVM

First of all, we need to make these Linux LVM partition able to be part of a group (vg). I always find tricky to remember the logic behind. Let’s imagine that the device itself now is just labelled “Linux LVM” but we need toĀ initiateĀ it in somehow.

pvcreate <dev>

Now theseĀ guys are ready to be part of a group. In this case a Virtual Group (vg).
Let’s check that it’s actually true:
pvs

Time to create a group with these devices (this could be done also with just a single device):

vgcreate <lvmgroupname> <dev> <dev> …

Now, let’s check again with pvs and vgs

Now pvs shows the VG group no longer empty but withĀ mylvmvg. And vgs tells us that the VG is about 14GB in size, fully free with no LV in it.

Good! Now, let’s make some LVs (logical volumes). These will be the newĀ “partitions/disks” that we will be actually able to format, mount and use! šŸ™‚

lvcreate -n <name> -L xGB <vg_group_name>

Some checks to verify:

A new LV appears inĀ vgs andĀ lvs shows the 2GB volume that we have created.

Let’s create another one, but this time, using the full remaining space (using -l 100%VGĀ option instead of -L xGB)

Magic!

Now, we have two devices, both ‘a’ -> active and ready to be formatted:
mkfs.ext4 <device>

I’ve done this for /dev/mylvmvg/part1 and /dev/mylvmvg/part2.

Let’s create the mount points and mount them:

As you can see, the devices are appearing now asĀ /dev/mapper/mylvmvg-partX. You can use eitherĀ /dev/mylvmvg/partX orĀ /dev/mapper/mylvmvg-partX. Theoretically, theĀ mapper one is recommended (my bad!).

Now the 2 devices are ready to be used as a typical disk/partition formatted with ext4 filesystem.


Resize Logical Volume

Now, imagine thatĀ part1 is too small, and you need more space. And luckily, yourĀ part2 volume has plenty. Is there any way to “steal” some space fromĀ part2Ā and give it toĀ part1?
Ooohh yesss! šŸ™‚

How?

  1. shrinkĀ part2 logical volumeĀ AND its filesystem
  2. expandĀ part1 logical volume AND its filesystem

Here the comments inline:

 

Move logical volume onto a new RAID array

Now, let’s imagine that one of the 3 initialĀ md devices are having problems, or simply we want to move on a faster/bigger raid array.
TheĀ magic of LVM is that we can actually do this with NO DOWNTIME!

How?

In this example we assume that a new /dev/md10 device is attached to our server and we want to removeĀ /dev/md2 device.

  1. We need to take the new device and go through all the previous steps:
    1. fdisk
    2. pvcreate
  2. After that, we need to add thisĀ initialised device in the existing volumeĀ group (vg)
  3. Move whatever is stored on the physical device
  4. Shrink the volume group
  5. Remove the device

Now where the new bits are starting:
pvmove, vgreduce, pvremove

 

In this example we have left LVM to decide where to put the data that was stored onĀ /dev/md2 device.
Just for reference, we could have specified the destination physical device (e.g. if we were thinking to remove more devices and make sure that the data was ending up on the new RAID and not sprat across the other disks):

Or, in case we just wanted to move a specific logical volume, let’s sayĀ part1

 

…happy LVM’ing! šŸ˜‰