Linux LVM (Logical Volume Manager) | AKA Spanned volumes

Hi guys

We’ve a EBS instance that totals almost 1TB physical size hosted on a high end IBM server & periodically we clone the instance to insure that the cold backups are reliable for DR purposes.

Recently we’ve decommissioned one HP ML110 G6 server with single xeon processor, 8GB memory that was dedicated for obsolete bio-metric monitoring and reporting running Windows 2003. I thought of using the same server for future restorations of EBS cold backups & realized that the server doesn’t support RAID 5 & moreover the built-in RAID is categorized under “fakeRAID”, which uses the built-in RAID technology, depending upon the CPU for the crippled RAID processing.

Using the HP Pavilion Easy Setup CD, I created an array and to my total disappointment found that Linux doesn’t read the fakeRAID while an installation is attempted.

The above were attempted because the ML110 G6 had 4 numbers of 500GB SATA HDD drives and I needed 1TB on a single volume. My database instance size as on date is 493GB, which would scream lack of space on a single 500GB partition. So I started reading about software RAID, which was too complex to setup with my minimal exposure to Linux. Further readings brought me to LVM (Logical Volume Manager) using which one can create spanned volumes as like in Windows.

Before proceeding further, please be aware of the RISKS associated with spanned volumes AKA LVM with multiple drives

How to?

We’ll consider a fresh installation of CentOS6/RHEL6/OEL6 for the exercises

Source thread (Please, please read)

Hardware: HP ML110 G6, 8GB memory, 4x500GB SATA HDD

Linux installation details

Installed Linux on HDD#1 (/dev/sda), 10GB boot, 4GB Swap, 110G / & balance as extended partition

Now, I am left with 3 HDDs, which are “untouched”, ie, no partitions are made

  1. /dev/sdb
  2. /dev/sdc
  3. /dev/sdd

As I have mentioned, my requirement was to have 1TB of storage for the cloning purposes, hence I chose 2x500GB (/dev/sdb, /dev/sdc)

First I created partitions using “fdisk”, the age old command line utility, even though better structured GUI is available with latest Linux distributions

Login to terminal as “root”

$fdisk /dev/sdb

n (new parition) -> p (primary partition) -> 1 (number of partitions) -> w (Write changes)

Repeated the same for /dev/sdc

$fdisk /dev/sdc

n (new parition) -> p (primary partition) -> 1 (number of partitions) -> w (Write changes)

We’ll use the following 3 commands to create our LVM

  1. pvcreate
  2. vgcreate
  3. lvcreate

create two physical volumes

$pvcreate /dev/sdb1 /dev/sdc1

create one volume group with the two physical volumes

$vgcreate VG_DATA /dev/sdb1 /dev/sdc1

create one logical volume

$lvcreate -l 100%FREE -n DATA VG_DATA

create the file system on your new volume

$mkfs.ext4 /dev/VG_DATA/DATA  #You may use ext3, based on your Linux distribution

$mkdir -p /u01

mount the volume (mount /dev/VG_DATA/DATA /u01)

That’s all folks, I have created my1st LVM aka spanned volume in Linux.

If you are planning to create logical volumes using multiple disks, be aware of the risks. You may lose millions worth data if no proper backups are taken and recovery could be a nightmare!

Not limited to total data loss, performance issues also should be considered, especially when such a setup hosts databases which require faster I/O.

for Windows7bugs

rajesh

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.