Difference between revisions of "Software RAID and LVM"

Revision as of 06:43, 11 December 2013

Introduction

Warning: Be sure to review the RAID article and be aware of all applicable warnings, particularly if you select RAID5.

Although RAID and LVM may seem like analogous technologies they each present unique features. This article uses an example with three similar 1TB SATA hard drives. The article assumes that the drives are accessible as /dev/sda, /dev/sdb, and /dev/sdc. If you are using IDE drives, for maximum performance make sure that each drive is a master on its own separate channel.

Tip: It is good practice to ensure that only the drives involved in the installation are attached while performing the installation.

LVM Logical Volumes / /var /swap /home

LVM Volume Groups /dev/VolGroupArray

RAID Arrays /dev/md0 /dev/md1

Physical Partitions /dev/sda1 /dev/sdb1 /dev/sdc1 /dev/sda2 /dev/sdb2 /dev/sdc2

Hard Drives /dev/sda /dev/sdb /dev/sdc

Swap space

Note: If you want extra performance, just let the kernel use distinct swap partitions as it does striping by default.

Many tutorials treat the swap space differently, either by creating a separate RAID1 array or a LVM logical volume. Creating the swap space on a separate array is not intended to provide additional redundancy, but instead, to prevent a corrupt swap space from rendering the system inoperable, which is more likely to happen when the swap space is located on the same partition as the root directory.

MBR vs. GPT

Template:Wikipedia The widespread Master Boot Record (MBR) partitioning scheme, dating from the early 1980s, imposed limitations which affected the use of modern hardware. GUID Partition Table (GPT) is a new standard for the layout of the partition table based on the UEFI specification derived from Intel. Although GPT provides a significant improvement over a MBR, it does require the additional step of creating an additional partition at the beginning of each disk for GRUB2 (see: GPT specific instructions).

Boot loader

This tutorial will use SYSLINUX instead of GRUB. GRUB when used in conjunction with GPT requires an additional BIOS Boot Partition. Additionally, the 2011.08.19 Arch Linux installer does not support GRUB.

GRUB supports the default style of metadata currently created by mdadm (i.e. 1.2) when combined with an initramfs, which has replaced in Arch Linux with mkinitcpio. SYSLINUX only supports version 1.0, and therefore requires the --metadata=1.0 option.

Some boot loaders (e.g. GRUB Legacy, LILO) will not support any 1.x metadata versions, and instead require the older version, 0.90. If you would like to use one of those boot loaders make sure to add the option --metadata=0.90 to the /boot array during RAID installation.

Installation

Obtain the latest installation media and boot the Arch Linux installer as outlined in the Beginners' Guide, or alternatively, in the Official Arch Linux Install Guide. Follow the directions outlined there until you have configured your network.

Load kernel modules

Enter another TTY terminal by typing Alt+F2. Load the appropriate RAID (e.g. raid0, raid1, raid5, raid6, raid10) and LVM (i.e. dm-mod) modules. The following example makes use of RAID1 and RAID5.

# modprobe raid1
# modprobe raid5
# modprobe dm-mod

Prepare the hard drives

Note: If your hard drives are already prepared and all you want to do is activate RAID and LVM jump to Activate existing RAID devices and LVM volumes. This can be achieved with alternative partitioning software (see: Article).

Each hard drive will have a 100MB /boot partition, 2048MB /swap partition, and a / partition that takes up the remainder of the disk.

The boot partition must be RAID1, because GRUB does not have RAID drivers. Any other level will prevent your system from booting. Additionally, if there is a problem with one boot partition, the boot loader can boot normally from the other two partitions in the /boot array. Finally, the partition you boot from must not be striped (i.e. RAID5, RAID0).

Install gdisk

Since most disk partitioning software (i.e. fdisk and sfdisk) does not support GPT you will need to install gptfdisk to set the partition type of the boot loader partitions.

Update the pacman database:

$ pacman-db-upgrade

Refresh the package list:

$ pacman -Syy

Install gptfdisk.

Partition hard drives

We will use gdisk to create three partitions on each of the three hard drives (i.e. /dev/sda, /dev/sdb, /dev/sdc):

   Name        Flags      Part Type  FS Type          [Label]        Size (MB)
-------------------------------------------------------------------------------
   sda1        Boot        Primary   linux_raid_m                       100.00  # /boot
   sda2                    Primary   linux_raid_m                      2000.00  # /swap
   sda3                    Primary   linux_raid_m                     97900.00  # /

Open gdisk with the first hard drive:

# gdisk /dev/sda

and type the following commands at the prompt:

Add a new partition: n
Select the default partition number: Enter
Use the default for the first sector: Enter
For sda1 and sda2 type the appropriate size in MB (i.e. +100MB and +2048M). For sda3 just hit Enter to select the remainder of the disk.
Select Linux RAID as the partition type: fd00
Write the table to disk and exit: w

Repeat this process for /dev/sdb and /dev/sdc or use the alternate sgdisk method below. You may need to reboot to allow the kernel to recognize the new tables.

Note: Make sure to create the same exact partitions on each disk. If a group of partitions of different sizes are assembled to create a RAID partition, it will work, but the redundant partition will be in multiples of the size of the smallest partition, leaving the unallocated space to waste.

Clone partitions with sgdisk

If you are using GPT, then you can use sgdisk to clone the partition table from /dev/sda to the other two hard drives:

$ sgdisk --backup=table /dev/sda
$ sgdisk --load-backup=table /dev/sdb
$ sgdisk --load-backup=table /dev/sdc

Note: When using this method to clone the partition table of an active drive onto a replacement drive for the same system (e.g. RAID drive replacement), use sgdisk -G /dev/<newDrive> to re-randomise the GUID of the disk and partitions to ensure they are unique.

RAID installation

After creating the physical partitions, you are ready to setup the /boot, '/swap, and / arrays with mdadm. It is an advanced tool for RAID management that will be used to create a /etc/mdadm.conf within the installation environment.

Create the / array at /dev/md0:

# mdadm --create /dev/md0 --level=5 --raid-devices=3 /dev/sd[abc]3

Create the /swap array at /dev/md1:

# mdadm --create /dev/md1 --level=1 --raid-devices=3 /dev/sd[abc]2

Note: If you plan on installing a boot loader that does not support the 1.x version of RAID metadata make sure to add the --metadata=0.90 option to the following command.

Create the /boot array at /dev/md2:

# mdadm --create /dev/md2 --level=1 --raid-devices=3 --metadata=1.0 /dev/sd[abc]1

Synchronization

Tip: If you want to avoid the initial resync with new hard drives add the --assume-clean flag.

After you create a RAID volume, it will synchronize the contents of the physical partitions within the array. You can monitor the progress by refreshing the output of /proc/mdstat ten times per second with:

# watch -n .1 cat /proc/mdstat

Tip: Follow the synchronization in another TTY terminal by typing Alt+F3 and then execute the above command.

Further information about the arrays is accessible with:

# mdadm --misc --detail /dev/md[012] | less

Once synchronization is complete the State line should read clean. Each device in the table at the bottom of the output should read spare or active sync in the State column. active sync means each device is actively in the array.

Note: Since the RAID synchronization is transparent to the file-system you can proceed with the installation and reboot your computer when necessary.

Scrubbing

It is good practice to regularly run data scrubbing to check for and fix errors.

Note: Depending on the size/configuration of the array, a scrub may take multiple hours to complete.

To initiate a data scrub:

# echo check > /sys/block/md0/md/sync_action

As with many tasks/items relating to mdadm, the status of the scrub can be queried:

# cat /proc/mdstat

Example:

$ cat /proc/mdstat

Personalities : [raid6] [raid5] [raid4] [raid1] 
md0 : active raid1 sdb1[0] sdc1[1]
      3906778112 blocks super 1.2 [2/2] [UU]
      [>....................]  check =  4.0% (158288320/3906778112) finish=386.5min speed=161604K/sec
      bitmap: 0/30 pages [0KB], 65536KB chunk

To stop a currently running data scrub safely:

# echo idle > /sys/block/md0/md/sync_action

When the scrub is complete, admins may check how many blocks (if any) have been flagged as bad:

# cat /sys/block/md0/md/mismatch_cnt

The check operation scans the drives for bad sectors and automatically repairs them. If it finds good sectors that contain bad data (the data in a sector does not agree with what the data from another disk indicates that it should be, for example the parity block + the other data blocks would cause us to think that this data block is incorrect), then no action is taken, but the event is logged (see below). This "do nothing" allows admins to inspect the data in the sector and the data that would be produced by rebuilding the sectors from redundant information and pick the correct data to keep.

General Notes on Scrubbing

Note: Users may alternatively echo repair to /sys/block/md0/md/sync_action but this is ill-advised since if a mismatch in the data is encountered, it would be automatically updated to be consistent. The danger is that we really don't know whether it's the parity or the data block that's correct (or which data block in case of RAID1). It's luck-of-the-draw whether or not the operation gets the right data instead of the bad data.

It is a good idea to set up a cron job as root to schedule a periodic scrub. See raid-check^AUR which can assist with this.

RAID1 and RAID10 Notes on Scrubbing

Due to the fact that RAID1 and RAID10 writes in the kernel are unbuffered, an array can have non-0 mismatch counts even when the array is healthy. These non-0 counts will only exist in transient data areas where they don't pose a problem. However, since we can't tell the difference between a non-0 count that is just in transient data or a non-0 count that signifies a real problem. This fact is a source of false positives for RAID1 and RAID10 arrays. It is however recommended to still scrub to catch and correct any bad sectors there might be in the devices.

LVM installation

This section will convert the two RAIDs into physical volumes (PVs). Then combine those PVs into a volume group (VG). The VG will then be divided into logical volumes (LVs) that will act like physical partitions (e.g. /, /var, /home). If you did not understand that make sure you read the LVM Introduction section.

Create physical volumes

Make the RAIDs accessible to LVM by converting them into physical volumes (PVs) using the following command. Repeat this action for each of the RAID arrays created above.

# pvcreate /dev/md0

Note: This might fail if you are creating PVs on an existing Volume Group. If so you might want to add -ff option.

Confirm that LVM has added the PVs with:

# pvdisplay

Create the volume group

Next step is to create a volume group (VG) on the PVs.

Create a volume group (VG) with the first PV:

# vgcreate VolGroupArray /dev/md0

Confirm that LVM has added the VG with:

# vgdisplay

Create logical volumes

Now we need to create logical volumes (LVs) on the VG, much like we would normally prepare a hard drive. In this example we will create separate /, /var, /swap, /home LVs. The LVs will be accessible as /dev/mapper/VolGroupArray-<lvname> or /dev/VolGroupArray/<lvname>.

Create a / LV:

# lvcreate -L 20G VolGroupArray -n lvroot

Create a /var LV:

# lvcreate -L 15G VolGroupArray -n lvvar

Note: If you would like to add the swap space to the LVM create a /swap LV with the -C y option, which creates a contiguous partition, so that your swap space does not get partitioned over one or more disks nor over non-contiguous physical extents:

# lvcreate -C y -L 2G VolGroupArray -n lvswap

Create a /home LV that takes up the remainder of space in the VG:

# lvcreate -l +100%FREE VolGroupArray -n lvhome

Confirm that LVM has created the LVs with:

# lvdisplay

Tip: You can start out with relatively small logical volumes and expand them later if needed. For simplicity, leave some free space in the volume group so there is room for expansion.

Update RAID configuration

Since the installer builds the initrd using /etc/mdadm.conf in the target system, you should update that file with your RAID configuration. The original file can simply be deleted because it contains comments on how to fill it correctly, and that is something mdadm can do automatically for you. So let us delete the original and have mdadm create you a new one with the current setup:

# mdadm --examine --scan > /etc/mdadm.conf

Note: Read the note in the Update configuration file section about ensuring that you write to the correct mdadm.conf file from within the installer.

Prepare hard drive

Follow the directions outlined the Installation section until you reach the Prepare Hard Drive section. Skip the first two steps and navigate to the Manually Configure block devices, filesystems and mountpoints page. Remember to only configure the PVs (e.g. /dev/mapper/VolGroupArray-lvhome) and not the actual disks (e.g. /dev/sda1).

Warning: mkfs.xfs will not align the chunk size and stripe size for optimum performance (see: Optimum RAID).

Configure system

Warning: Follow the steps in the LVM Important section before proceeding with the installation.

mkinitcpio.conf

mkinitcpio can use a hook to assemble the arrays on boot. For more information see mkinitpcio Using RAID.

Add the dm_mod module to the MODULES list in /etc/mkinitcpio.conf.
Add the mdadm_udev and lvm2 hooks to the HOOKS list in /etc/mkinitcpio.conf after udev.

Conclusion

Once it is complete you can safely reboot your machine:

# reboot

Install the bootloader on the Alternate Boot Drives

Once you have successfully booted your new system for the first time, you will want to install the bootloader onto the other two disks (or on the other disk if you have only 2 HDDs) so that, in the event of disk failure, the system can be booted from any of the remaining drives (e.g. by switching the boot order in the BIOS). The method depends on the bootloader system you're using:

Syslinux

Log in to your new system as root and do:

# /usr/sbin/syslinux-install_update -iam

Note: For this to work with GPT, the gptfdisk package is needed as the backend for setting the boot flag.

Syslinux will deal with installing the bootloader to the MBR on each of the members of the RAID array:

Detected RAID on /boot - installing Syslinux with --raid
Syslinux install successful

Attribute Legacy Bios Bootable Set - /dev/sda1
Attribute Legacy Bios Bootable Set - /dev/sdb1
Installed MBR (/usr/lib/syslinux/gptmbr.bin) to /dev/sda
Installed MBR (/usr/lib/syslinux/gptmbr.bin) to /dev/sdb

GRUB legacy