[[http://tmade.de|Home tmade.de]]

[[http://wiki.tmade.de|Home Wiki]]

===== LVM =====

==== LVM Layout ====

Basically LVM looks like this:

{{:wiki:lvm_layout.png|}}

You have one or more physical volumes (/dev/sdb1 - /dev/sde1 in our example), and on these physical volumes you create one or more volume groups (e.g. fileserver), and in each volume group you can create one or more logical volumes. If you use multiple physical volumes, each logical volume can be bigger than one of the underlying physical volumes (but of course the sum of the logical volumes cannot exceed the total space offered by the physical volumes).

It is a good practice to not allocate the full space to logical volumes, but leave some space unused. That way you can enlarge one or more logical volumes later on if you feel the need for it.

In this example we will create a volume group called fileserver, and we will also create the logical volumes /dev/fileserver/share, /dev/fileserver/backup, and /dev/fileserver/media (which will use only half of the space offered by our physical volumes for now - that way we can switch to RAID1 later on (also described in this tutorial)). 


To create a LVM on commandline you have to:

  1. Create a Partition (e.g. fdisk /dev/sdb - Hex code (type L to list codes): <-- 8e = Linux LVM)
  2. Execute "pvscan"
  3. Execute "pvcreate /dev/sdb1"
  4. Execute "vgcreate data /dev/sdb1"
  5. Execute "lvcreate -l 100%FREE -n lv-data data"                # lv-data = name logical volume, data = name volume group
or  
  lvcreate -L 5G -n lv-data data
  lvcreate --name lv-data --size 30G data                          #data = name vg, lv-data = name lv
  lvcreate --name home --size 5G system                            #system = name vg, home = name lv
  6. Format disk (e.g. mkfs.ext3 -Ldata /dev/data/lv-data)
  
Note: TAB (automatic completion) does NOT work here and you have to enter device info/path manually!!

To extend an existing LV (logical Volume) 

  1.   Create a Partition (e.g. fdisk /dev/sdb)
  1.1  partx -u /dev/sdb (reread kernel partition table to omit reboot)
  2.   Execute "pvscan"
  3.   Execute "pvcreate /dev/sdb1"
  4.   vgextend data /dev/sda4 /dev/sdn1           #Tries to extend the existing volume group "data" by the new physical volumes (see pvcreate) "/dev/sdn1" and /dev/sda4"
  5.   lvextend -L+100G /dev/data/lv-data          #Extends "/dev/data/lv-data" 100GB - e. g. from 100 to 200 GB
  5.1. lvextend -L 10G /dev/data/lv-data           #Extends "/dev/data/lv-data" to 10GB
  5.2. lvextend -l +100%FREE /dev/data/lv-data     #Extends to max. available space
  6.   resize2fs /dev/data/lv-data                 #Resize FS (ext3/4)
  
**Note:**
To expand a **xfs** filesystem run

  xfs_growfs /mountpoint
  xfs_info /mountpoint
  
==== Resize LVM-Disk ====

http://www.techrepublic.com/blog/smb-technologist/extending-partitions-on-linux-vmware-virtual-machines/

or (if no data on Disk!!):

  1.  Enlarge vdisk (on VMware/ HyperV or whatever)
  1.1 Check on system, if enlarged disk has been recognised/ updated (cat /proc/partitions) - if not, see next chapter "LUN" to get it updated
  2.  fdisk /dev/device , delete Partition, create Partition with new blocksize
  3.  restart system (depends on OS)
  4.  pvresize -v -d /dev/device
  5.  lvextend -l +100%FREE /dev/vgname/lvname
  6.  resize2fs /dev/vgname/lvname
  
**Note:** To execute 5 and 6 at once (add "-r"):

  lvextend -r -l +100%FREE /dev/vgname/lvname
  lvextend -r -L 10G /dev/vgname/lvname
  
==== Backup & Recovery Partition Table ====
  
  sfdisk -d /dev/sda > partinfo_sda
  sfdisk -d /dev/sdb > partinfo_sdb                        #backup partition table   
  sfdisk -d /dev/sdc > partinfo_sdc

  sfdisk --force /dev/sdb < partinfo_sdb                   #restore partition table 
  
==== LUN ====

A „LUN“ (or virtual Disk) can be recognised in a VMware Virtual Maschine (Linux) with the following command:
  
  echo "- - -" > /sys/class/scsi_host/host0/scan
  echo "- - -" > /sys/class/scsi_host/host1/scan
  echo "- - -" > /sys/class/scsi_host/host2/scan  

"host1" represents the controller ID.

Another solution:

  rescan-scsi-bus
  
Note: The package "scsitools" is required (apt-get install scsitools)!

Last solution (this works always even on older linux-derivates):

  echo 1 > /sys/class/scsi_device/0:0:1:0/device/rescan 
  echo 1 > /sys/class/scsi_disk/0:0:2:0/device/rescan
  echo 1 > /sys/class/scsi_disk/1\:0\:2\:0/device/rescan
  
Update Partition:

  partprobe
  cat /proc/partitions
  kpartx -a /dev/sdc
  cat /proc/partitions
  partx -a /dev/sdc

What the kernel actualy see´s:

  cat /sys/block/sdc/sdc1/start
  cat /sys/block/sdb/sdb1/size

  
==== Usefull Commands ====

  pvscan
  vgscan
  lvscan
  
  pvdisplay
  vgdisplay 
  lvdisplay
  
  pvs
  vgs
  lvs 
  
  vgcfgrestore vgname
  vgchange -ay vgname
  pvcreate --uuid "LxgpTz-iIqL-wpz5-sbuN-9Dn8-VOWF-cNZgSM" --restorefile /etc/lvm/archive/x /dev/sdb2
  
  /etc/lvm/archive/ 
  
  lvextend -L+1G /dev/myvg/homevol
  lvm lvremove /dev/system/lv-system                #Delete LV "lvm lvremove /dev/system/lv-system"
  lvcreate -n swaplv -L 500M volg1                  #Create LVM-Swapfile "swaplv" on volumegroup "volg1" with size 500M
  vgremove volume_group_name                        #Delete volumegroup (execute "vgscan" to get an overview of groups and replace the name)
  vgreduce --removemissing --test volume_group      #Removes missing disks (uuid) on volumegroup "volume_group", but just as a test
  vgreduce --removemissing volume_group             #Removes missing disks (uuid) on volumegroup "volume_group"
  lvchange -an /dev/system/lv-system                #Disable LV "/dev/system/lv-system"
  lvchange -ay /dev/system/lv-system                #Enable LV "/dev/system/lv-system"
  vgchange -an system                               #Disable VG "system"
  vgchange -ay system                               #Enable VG "system"
  pvcreate /dev/sdb1
  pvresize -v -d /dev/sdb1
  vgextend system /dev/sdc1                         #add "/dev/sdc1" to vg "system"
  vgreduce system /dev/sdb1                         #remove "/dev/sdb1" from vg "system"
  lvcreate --name backups --size 30G server1
  mkfs.ext3 /dev/mapper/server1-backups
  
  pvmove
  
  /dev/server1/root = /dev/mapper/server1-root
  
==== Snapshot ====

  lvcreate -L10G -s -n rootsnapshot /dev/vg_name/lv_name
  mkdir -p /mnt/backup
  mkdir -p /backup
  mount /dev/vg_name/lv_name /mnt/backup
  dd if=/dev/vg_name/lv_name of=/mnt/backup/root.dd
  tar -pczf /backup/root.tar.gz /mnt/backup
  umount /mnt/backup
  lvremove /dev/vg_name/lv_name
  
=== Restore Snapshot ===

  dd if=/mnt/backup/root.dd of=/dev/vg_name/lv_name

=== MySQL Backup ===

To backup MySQL via LVM-sanpshot:

<code>
#!/bin/bash

snapshot_create() {
	exec 3> >(/usr/bin/mysql) && \
	echo "flush tables with read lock;" >&3 && \
	LVM_SUPPRESS_FD_WARNINGS='y' /sbin/lvcreate -L100%FREE -s -n mysql-backup /dev/vg00/root >/dev/null && \
	echo "unlock tables;" >&3 && \
	exec 3>&- && \
	mount /dev/vg00/mysql-backup /mnt && \
	mount --bind /mnt/srv/mysql /srv/mysql-backup
}

snapshot_cleanup() {
	umount /srv/mysql-backup
	umount /mnt
	lvremove -f /dev/vg00/mysql-backup >/dev/null
}

snapshot_create || { snapshot_cleanup; exit 1; }
res=$(/usr/bin/rdiff-backup --preserve-numerical-ids --exclude-sockets --exclude-globbing-filelist /etc/rdiff-backup-exclude / /backup/bastet/ 2>&1)
ret=$?
snapshot_cleanup

if [ $ret -ne 0 ]; then
	[ "$res" ] && echo "$res"
	exit $ret
fi

[ "$res" ] && echo "$res" | grep -Ev 'UpdateError var/log/(auth.log|btmp) Updated mirror temp file .* does not match source'
exit 0
</code>