Raid1 with LVM from scratch

In this manual will be created RAID1 with LVM. In this tutorial, disk are mounted in system as /dev/sdX and /dev/sdY

Prerequisites

  • 2 empty HDDs with same capacity
  • Kernel with LVM
  • lvm2
  • parted

Software

Install lvm2 package

root #emerge sys-fs/lvm2

Install parted

(parted)emerge parted

Disk partitions create

Warning
Data on /dev/sdX and /dev/sdY will be lost. Be carefull with disk names

Create partitions on both disks with parted

Start parted for /dev/sdX disk

root #parted -a optimal /dev/sdX

Set units to mib

(parted)unit mib

Create GPT table on disk

(parted)mklabel gpt

Create primary partition, use all available space

(parted)mkpart primary 1 -1

Set partition name to raiddata0

(parted)name 1 raiddata0

Add lvm flag to new partition

(parted)set 1 lvm on

Result should be:

(parted)print
Model: ATA ST6000VN0033-2EE (scsi)
Disk /dev/sdc: 6001GB
Sector size (logical/physical): 512B/4096B
Partition Table: gpt
Disk Flags: 

Number  Start   End     Size    File system  Name       Flags
 1      1049kB  6001GB  6001GB               raiddata0  lvm

Execute same parted commands for /dev/sdY

LVM

Next steps will be to create physical volumes on both disks, add both physical volume to same volume group and create logical volume with raid1 logic

Physical volume

Create physical LVM volumes on first disk on first partition

root #lvm pvcreate /dev/sdX1

Create physical LVM volumes on second disk on first partition

root #lvm pvcreate /dev/sdY1

Volume group

Include both physical volumes to one volume group with name *raid0vg0*

root #vgcreate raid0vg0 /dev/sdX1 /dev/sdY1

Now both disks in same volume group.

Logical Volume

Create logical volume with name *raid0lv0* on volume group *raid0vg0* with RAID1 logic, use all available space. --nosync means skip the initial synchronization for raid1 (because this is new raid without any data on it)

root # lvcreate --mirrors 1 --type raid1 -l 100%FREE --nosync -n raid0lv0 raid0vg0

Now raid1 created on both disks /dev/sdX and /dev/sdY . Last step left - create FS and mount this FS on boot. see section below

EXT4 Filesystem (non encrypted)

Create filesystem on VolumeGroup *raid0vg0* on logical volume r*aid0lv0*

root #mkfs.ext4 /dev/raid0vg0/raid0lv0

Done

Important
Please, do not forget to add lvm2 service at boot: rc-update add lvm2 boot
Important
Your kernel should include LVM modules in initrd on compiled into kernel. See

Mount filesystem on boot

Run blkid to find UUID of ext4 filesystem on our LVM raid1

root #blkid
...
/dev/mapper/raid0vg0-raid0lv0_rimage_0: UUID="10092fa9-43f5-421e-a0a1-ca96323c6388" TYPE="ext4"
/dev/mapper/raid0vg0-raid0lv0_rimage_1: UUID="10092fa9-43f5-421e-a0a1-ca96323c6388" TYPE="ext4"
/dev/mapper/raid0vg0-raid0lv0: UUID="10092fa9-43f5-421e-a0a1-ca96323c6388" TYPE="ext4"
...
  • UUID="10092fa9-43f5-421e-a0a1-ca96323c6388"* is id of our ext4 filesystem on raid1. Last thing is to add fs uuid to fstab

Create mountpoint /mnt/data

root #mkdir /mnt/data

Add to fstab mounting

FILE /etc/fstab
<syntaxhighlight lang="bash">...
UUID=10092fa9-43f5-421e-a0a1-ca96323c6388  /mnt/data        ext4        defaults        0 2
...</syntaxhighlight>

EXT4 Filesystem (encrypted with LUKS)

Create luks AES encrypted partition on top of Volume Group *raid0vg0* in Logical Volume *raid0lv0* (raid1)

Important
Please, see Full Disk Encryption From Scratch Simplified
root #cryptsetup luksFormat -c aes-xts-plain64:sha256 -s 256 /dev/raid0vg0/raid0lv0

Map encrypted luks disk as *raid0lv0encripted*

root #cryptsetup luksOpen /dev/raid0vg0/raid0lv0 raid0lv0encripted

Create EXT4 filesystem in luks disk

root #mkfs.ext4 /dev/mapped/raid0lv0encripted

Mount LUKS encrypted device on boot from LVM raid1

First, create directrory, that will contain keys for encrypting/decryption devices

root@localhost:/#mkdir /etc/keyfiles
root@localhost:/#chmod 0400 /etc/keyfiles

Create 4k keyfile with name main

root@localhost:/#dd if=/dev/urandom of=/etc/keyfiles/main bs=1024 count=4
root #chmod 0400 /etc/keyfiles/main

Add main keyfile to list of keys, that can decrypt disk (technically: add keyfile to LUKS slot)

root #cryptsetup luksAddKey /dev/raid0vg0/raid0lv0 /etc/keyfiles/main


Find id of LUKS device (should be same as LV raid0lv0)

root #blkid
/dev/sdc1: UUID="OxJaqA-yMAP-sOjE-T5BR-H9Lp-rtPN-pl7rFC" TYPE="LVM2_member" PARTLABEL="raiddata1" PARTUUID="9c794e91-22a8-4b58-bedd-c3f656d82bd9"
/dev/sdb1: UUID="gNcHvg-Rocv-pFFc-VzvF-49tX-D1d3-odSe2h" TYPE="LVM2_member" PARTLABEL="raiddata0" PARTUUID="70121885-4a45-4a2b-8d3e-49edd8fffd34"
/dev/mapper/raid0vg0-raid0lv0_rimage_0: UUID="10092fa9-43f5-421e-a0a1-ca96323c6388" TYPE="ext4"
/dev/mapper/raid0vg0-raid0lv0_rimage_1: UUID="10092fa9-43f5-421e-a0a1-ca96323c6388" TYPE="ext4"
/dev/mapper/raid0vg0-raid0lv0: UUID="cd5740a1-b642-4359-a0b9-af84a8f01092" TYPE="crypto_LUKS"
/dev/mapper/raid0lv0encripted: UUID="fc7ec587-35e4-4726-815d-e1693cd89b70" TYPE="ext4"

In our case it is UUID="cd5740a1-b642-4359-a0b9-af84a8f01092"

Add to file /etc/conf.d/dmcrypt

FILE /etc/conf.d/dmcrypt
target='raid0lv0encripted'
source=UUID='cd5740a1-b642-4359-a0b9-af84a8f01092'
key='/etc/keyfiles/main'

Add dmcrypt to be started at boot

root #rc-update add dmcrypt boot

Create mountpoint /mnt/data

root #mkdir /mnt/data


Find EXT4 filesystem UUID

root #blkid
/dev/sdc1: UUID="OxJaqA-yMAP-sOjE-T5BR-H9Lp-rtPN-pl7rFC" TYPE="LVM2_member" PARTLABEL="raiddata1" PARTUUID="9c794e91-22a8-4b58-bedd-c3f656d82bd9"
/dev/sdb1: UUID="gNcHvg-Rocv-pFFc-VzvF-49tX-D1d3-odSe2h" TYPE="LVM2_member" PARTLABEL="raiddata0" PARTUUID="70121885-4a45-4a2b-8d3e-49edd8fffd34"
/dev/mapper/raid0vg0-raid0lv0_rimage_0: UUID="10092fa9-43f5-421e-a0a1-ca96323c6388" TYPE="ext4"
/dev/mapper/raid0vg0-raid0lv0_rimage_1: UUID="10092fa9-43f5-421e-a0a1-ca96323c6388" TYPE="ext4"
/dev/mapper/raid0vg0-raid0lv0: UUID="cd5740a1-b642-4359-a0b9-af84a8f01092" TYPE="crypto_LUKS"
/dev/mapper/raid0lv0encripted: UUID="fc7ec587-35e4-4726-815d-e1693cd89b70" TYPE="ext4"

In our case it is UUID="fc7ec587-35e4-4726-815d-e1693cd89b70"

Add to fstab mounting

FILE /etc/fstab
<syntaxhighlight lang="bash">...
UUID=fc7ec587-35e4-4726-815d-e1693cd89b70  /mnt/data        ext4        defaults        0 2
...</syntaxhighlight>

Check LVM RAID1 status

To check lvm raid status for volume group raid0vg0

root #lvs -a -o name,copy_percent,devices raid0vg0
  LV                  Cpy%Sync Devices                                  
  raid0lv0            100.00   raid0lv0_rimage_0(0),raid0lv0_rimage_1(0)
  [raid0lv0_rimage_0]          /dev/sdc1(1)                             
  [raid0lv0_rimage_1]          /dev/sdb1(1)                             
  [raid0lv0_rmeta_0]           /dev/sdc1(0)                             
  [raid0lv0_rmeta_1]           /dev/sdb1(0)

Performance tunnig

It is possible to tunne performance for raid1. By default, in raid1 all disk have same rangs for read/write. If one of disks are much slower then another, it possible to improve WRITE performance (with small penalty of read performance) for such scenario with disabling read from slowest drive.

In such scenario, slowest drive will only write data (without reading), and faster drive will read/write data.

root #lvchange --raidwritemostly /dev/sdb1 raid0vg0
Logical volume raid0vg0/raid0lv0 changed.

Where /dev/sdb1 - physical drive in vg0 (slowest drive), raid0vg0 - volume group

See also

External resources

This article is issued from Gentoo. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.