I am a bit new to hardware configuration in general, much less with linux, so please give me a little slack :)

I have the following setup: 1 320GB Drive that contains the OS (centOS) and 2 250GB drives that have a hardware RAID1 configuration (accomplished via my ASUS motherboard raid utility, they are mirrored as one device)

The problem is that everywhere I am reading online, the centOS is supposed to see the 2 disks that have RAID1 as a single device. However, when I run fdisk -l I am shown the three disks individually, the 320GB drive (/dev/sdc) and the individual 250GB drives (/dev/sda) and (/dev/sdb)

I was under the impression that I was supposed to see two, the /dev/sda (OS drive) and some kind of 'raid' device (/dev/sdb) that I can select for partitioning and then mounting.

What am I doing wrong? Because I will need to repeat this process when I install another 2 drives for RAID1 for another purpose - I need to get this down right.

  • Magically changing original questions ? Wow, what a stupid concept. – adaptr Nov 18 '11 at 11:29
  • From what I understand when you have Hardware raid you dont really care about what OS sees as long as you don't try to mount the mirrored one in read write mode. Hardware should take care of the Raid set. But then I am not 100% sure. Definitely you do not have a true hardware raid but to the OS it should not matter. t is ok to see 3 devices as thats what you have. See in fstab how it is mounted. I used an IBM 7xxx series server before that had similar config and I use centos as well. – Abhishek Dujari Nov 24 '11 at 16:21

What you have is evidently not hardware RAID but software RAID with a BIOS interface, often called fakeRAID. The main job of putting the disks in an array is done by the Windows driver. Related reading: How do I differentiate “fake RAID” from real RAID?

There are two advantages to hardware RAID over software RAID: it's independent of the operating system (which is only useful if you dual-boot), and it can perform better in some configurations (essentially those that require a cross-drive checksum, which is not the case of RAID-1). Software RAID has the advantage of being independent of the hardware (you can take your disks out and plug them into another machine running the same OS) and tends to come with better tools.

So, forget about your motherboard's fake RAID and use Linux's software RAID. The main tool you'll need to use is mdadm. Create a RAID-1 volume encompassing the two disks, then create partitions on that volume. Something like:

mdadm --create /dev/md0 -l 1 -n 2 /dev/sda /dev/sdb
fdisk /dev/md0
  • 1
    @Matt1776 What? In this situation, there's no difference between software RAID and hardware RAID. Either all your disks burn and you restore your data from backups, or enough of your disks survive and you quickly buy replacements. And, as I mentioned, there's an advantage if the RAID controller dies: software RAID works the same in any hardware, so you don't have the risk of being stuck with a particular model that has an unusual disk layout (those are uncommon for RAID-1). – Gilles Nov 4 '11 at 21:13
  • 2
    You have it backwards. All the data that's necessary to recover the RAID array is stored on the array itself. All that's stored on the OS disk is the instructions to mount the array at a certain location at boot time. For a RAID-1 array, any component disk is sufficient to recover the data (by plugging the disk into any computer running Linux). On the contrary, it's with hardware RAID that you're sometimes locked into a particular controller model (never a particular controller!). – Gilles Nov 4 '11 at 21:24

If you did have a good controller you would only see see the volume that combines both devices.

But a controller built into to a sata chipset of a motherboard is almost certainly just a fakeraid controller. Linux does weird things here depending on the controller. You will frequently see the individual drives, and Linux may not actually setup the RAID.

You are probably far better off disabling that, and using the Linux software RAID to set things up.

If you really want to get that fakeraid working then you probably have to mess around with dmraid.

  • I'd add, that the CentOS kernel is very much unlikely to have new patches to support FakeRAID the way Windows (or Ubuntu for that matter) does... – Hubert Kario Nov 24 '11 at 0:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy