Those of you who hang out in the #trilug channel on IRC heard me asking some questions earlier, then finding the answers and becoming quite enamored with this completely underexposed part of most modern Linux distributions.
LVM, or Logical Volume Management, is a software tool that allows you to abstract and manage your hard disk(s). It has undeniable perks for small standalone servers with a single IDE disk as well as large enterprise servers with multi-terabyte SAN arrays.
First you need to wrap your brain around how LVM organizes your disks. First you take entire hard disks, or partitions of hard disks, and tag them as a Physical Volume (PV).
Next you create a Volume Group (VG) which can contain one or more PV's. You can think of the VG as one large virtual hard disk.
The next step is to create Logical Volumes (LV). An LV is like a partition of a hard disk. One or more LV's are contained within a VG. You don't have to use all of your available space, and indeed you probably shouldn't use up all of your available space when creating LV's.
The LV's are formatted with the filesystem of your choice (usually ext3) just like a normal hard disk partition. They can then be mounted like a normal filesystem.
But then it gets neat.
Let's say your /home partition gets close to filling up. Remember how you didn't allocate all of your available space when making up the LV's? Well now you can extend your LV that contains /home, and extend the filesystem itself to consume the additional space in the LV that you just appended.
What if you *did* use up all of your available space when you defined your LV's? Not a problem. Chances are one of your other filesystems has free space. Let's say /usr has 2GB of free space. /usr doesn't change much. So let's shrink the LV that contains /usr by 1800MB and then reallocate that space to /home. Not a problem. You can shrink filesystems just as easily as you can grow them.
For you RDBMS admins check this out; you can create filesystem snapshots using unallocated space. So you can stop the database, create a snapshot, and start the database again in a second or two. Then just back up from the snapshot, and disable the snapshot when you're done. Total RDBMS downtime is so fast you might miss it if you blink.
You can read the LVM HOWTO at http://tldp.org/HOWTO/LVM-HOWTO/index.html but really the only way to get an appreciation for LVM is to slap an extra old drive or two (or three) into your Linux box and play with it. I did just that today and came away salivating from the practical applications of this. And this is someone who has been using Linux professionally for over five years.
The commands are very easy to understand. If you have figured out how to edit a file in vi or emacs, then you have enough brain cells to understand the simple management commands for LVM.
For you install fest folks, you might want to give serious though to making use of LVM in your installs. It will give the owner of the box a lot more control over disk layout down the road when they understand the dynamics of their machine better and can re-allocate space as needed (or append new drives even).
I know that Red Hat 9 has LVM support in the installer, and I am told Mandrake has had this for some time. Older versions of Red Hat don't seem to have full support for this, and by extension some of the Red Hat offshoots don't seem to yet. The HOWTO describes, well, HOW TO install LVM on systems that don't already have it.
--
C. Magnus Hedemark http://trilug.org/~chrish PGP Key fingerprint = 984D 9A88 3D60 016F BE01 1506 60FB 85E1 9ABD 96F6
PGP.sig
Description: PGP signature
