Before we begin, let me point out some benefits of using raid 1 mirroring. First, your mp3 files won't have their meta tags jumbled up and mangled by typical hard drive corruption. Also your movie files won't suddnely stop playing past certain points due to corruption. Umm... your exe files won't randomly give you errors about not being winows 32bit or 64bit executables... So all that stuff is good. So if you need to store files intact for long periods of time, you should spreed them across two HDDs which will be able to check for corruption and auto-heal themselves on the fly.

Now then, let us begin.

In this entry we will setup a RAID 1 (mirror) on our debian system. This is simply a bare procedure. For help maintaining a RAID, see the references. Even though I made sure to buy a motherboard that supported RAID 1, it would turn out that this 'support' relies entirely on the CPU to deal with the logic of the RAID setup (these types of raid controllers are called fakeraid). Windows machines will be completely faked out by these raids, even though they consume CPU resources and 'could' show up as two separate devices, the drivers for fakeraids will hide this information from the administrator. I like that kind of stuff in Windows, but in linux I like to know more about what's going on. For this reason, I'm not going to call our setup hardware raid, nor software raid, it just confuses google matters.

Tools:
mdadm - The program in charge of making our RAID work is called mdadm (Multiple Device ADMin).
cfdisk - this utility formats a hard disk.
mkfs.ext4 - this utility creates ext4 partitions.



1) Prep the drives

We need to format one of the drives and give them partitions so mdadm can mirror them for us. For this tutorial, the drives we will be committing to a RAID are /dev/sda and /dev/sdb. You may use fdisk -l to see what drives you have ready to raid.
ref: http://linux.justinh...cond_hard_drive


(partition)

$  cfdisk /dev/sda
$  cfdisk /dev/sdb
1. New >> Primary >> Size in MB (skim 100mb less or so to make it possible to replace the drive with an HDD with a couple bad sectors)
2. Set to BOOTABLE if desired.
3. Once done select Write
4. Select Quit

2) Installation of mdadm

$  apt-get install mdadm
$  modprobe raid456
$  modprobe raid1

ref: https://raid.wiki.ke...setup_cbb2.html

After those commands are run, you should see that the virtual file mdstat will yield this output.

$  cat /proc/mdstat
Personalities : [raid6] [raid5] [raid4] [raid1]
unused devices: <none>

3) Configuration of mdadm

This command will setup the devices to be configured in the RAID. After entering the command, your HDD will light up. This indicates that the drives are being synchronized.

mdadm --create --verbose /dev/md0 --level=mirror --raid-devices=2 /dev/sda1 /dev/sdb1

To check the progress of the syncronization, do

$  cat /proc/mdstat

Don't worry, this process is completely unobtrusive and will run in the background until it has completed the mirroring. I think it will even survive reboots, although I'm not going to test that now.

Finally save your mdadm configuration,

$  mdadm --detail --scan >> /etc/mdadm/mdadm.conf

4) Create your filesystem

From this point on, your raid device at /dev/md0 can now be manipulated as though it were any other disk like /dev/sdc.

Google has recently adopted ext 4 as their file system of choice (2010). Let's setup an ext 4 partition:

mkfs.ext4 /dev/md0 -b 4096 -E stride=16,stripe-width=80

That -b 4096 pattern makes our drive work well with windows. I might evaluate windows server and write up some tuts, even though it's a proprietary and likely won't even work with ext4...








5) Starting and Stopping the RAID:

If you followed my instructions thus far, you now have a "Persistent Superblock" written to all of the drives participating in the array. This lets the kernel know that they are a RAID for sure, and to treat them as such.

Every time your computer starts, it needs to reassemble the RAID from it's constituent disks and will do this automatically. This process can also be invoked manually by the command,

mdadm --assemble /dev/md0 /dev/sda1 /dev/sdb1

To close down the raid, /dev/md0, use the command,

$  mdadm --stop /dev/md0

And of course, she can be brought right back online again using the former command mentioned.



6) FSTAB, for auto mounting the RAID md0

Well, now that you have this new /dev/md0 device, you can treat it just like a normal HDD. We've already put an ext4 filesystem on it, and it's ready to be mounted.

$  mkdir /storage
$  mount -t ext4 /dev/md0 /storage
$  echo 'Hello world of RAID!' >> /storage/h.txt

So let's make that mount point a permanent mount by editing the /etc/fstab file.

Add this line to your fstab file:
(/etc/fstab)

.
.
.
/dev/md0 /storage ext4 relatime,errors=remount-ro 0 1 

That pretty much sums up the procedure. The next thing you should do is probably follow a samba procedure and get some file sharing going on those hard drives.

End Of Tut
==============================================================================



FAQi:

Systems Green
How do you know that your drives are in working condition?

You'll see that mdstat presents your situation something like this:

$  cat /proc/mdstat
Personalities : [raid1]
md0 : active raid1 sda1[0] sdb1[1]
      1464845671 blocks super 1.2 [2/2] [UU]    # [UU] IS GREAT!

unused devices: <none>

Also, you'll be able to mount /dev/md0 as though it were any other partition.


System's Red
How do you know when one of your drives had gone down?

$  cat /proc/mdstat
Personalities : [raid1]
md0 : active raid1 sda1[0]
      1464845671 blocks super 1.2 [2/1] [U_]     # [U_] IS BAD!

Confirm Mirroring
When I unplug one of my hard drives while the server is running (for testing purposes), my array breaks. How do it put things back together?


When you unplugged the hard drive, mdadm may have lost it forever due to random corruption or perhaps the original drive "got ahead" of the unplugged one and therefore the drives became out of sync I guess... (this bug happened to me on my first test) So you may actually need to explicitly tell mdadm to fail that device. Then you can add the device back into the array.

ref: http://www.ducea.com...dm-cheat-sheet/
(Let's assume that it was sdb1 that we pulled...)

#  Clone partitions on /dev/sda to /dev/sdb to ensure proper mirroring
$  sfdisk -d /dev/sda | sfdisk /dev/sdb

$  mdadm --fail /dev/md0 /dev/sdb1
$  mdadm --remove /dev/md0 /dev/sda1
$  mdadm --add /dev/md0 /dev/sdb1

What are some other random commands involving mdadm?

#  Rebuild your mdadm.conf file by scanning hdd metadata I believe...
$  mdadm --examine --scan --config=mdadm.conf >> /etc/mdadm/mdadm.conf

#  Fail and remove a hard drive on one line
$  mdadm /dev/md0 --fail /dev/sda1 --remove /dev/sda1

#  Build a fresh array in a very concise manor
$  mdadm -Cv /dev/md0 -l1 -n2 /dev/sd[ab]1 

#  Delete an array
$  mdadm --remove /dev/md0