RAID - Intel Matrix Storage Manager - what is this really?

[wildwolfe]

I recently bought a new system that has an Intel Matrix Storage Manager "RAID controller" (ICH10R/DO) on it. I'm a bit baffled over what this really is. I see at http://www.intel.com/support/chipsets/imsm/sb/cs-020663.htm that this controller is supported by the Linux dmraid and mdadm commands and supported in the 2.6 kernel version for quite a while. This looks as though it is some sort of convergence of a brain-dead hardware chipset that requires software installed in the OS to manage it. Kind of reminds me of the wimpy Windows modems of the past.

How I deployed. . . I set up two Seagate ST31500541AS disks as a mirrored pair in the hardware controller interface (CTL-I setup after POST). I installed Fedora 12 in the usual fashion, though was confusing considering I expected a single "RAID device" to be seen by Anaconda. I went ahead and set up the two native /dev/sda and /dev/sdb as a mirrored RAID device on installation (mdadm under the cover). Recently, Palimpsest is insisting that I have a disk problem with "Disk has many bad sectors" error on /dev/sdb. When I run "dmraid -s" it tells me that the meta device is OK and I see no hardware errors in the messages log. I'm not having kernel panics as others seem to have had on RHEL 5.x.

I haven't seen much on the net over the subject, though what little, the general attitude seems to be that this system/chipset is prone for problems. Has anyone else had experience with the Intel Rapid Storage Technology and/or Palimpsest? Is Palimpsest really trustworthy or would it be better to just forget about using the Intel controller and just do the software-only RAID?

[steelaworkn]

I have this on my computer as I used to run it with Windows XP and Windows 7. To get Fedora on the system, I released the disk from the Intel Matrix F6 Raid. Then I used Fedora to set up Raid and LVM to partion the Raid Array. I run, like my signature says, Raid 5 and it works nicely. Didn't bother with dmraid.

[forkbomb]

Pretty sure that Matrix is Intel's flagship FakeRAID that is shipped onboard with certain boards. Theoretically, Matrix-RAID is supposed to be supported by dmraid. That hypothetically makes it possible to dual-boot off of a single fake RAID volume.
http://en.wikipedia.org/wiki/Intel_Matrix_RAID
http://en.wikipedia.org/wiki/RAID#Fi...FakeRAID.22.29

This came up earlier. Apparently Fedora's more recent Anaconda won't touch certain types of non-hardware RAID out of fear of FUBARing the data on the array.
http://forums.fedoraforum.org/showpo...46&postcount=4
...which is supported by the fact that you say:

I went ahead and set up the two native /dev/sda and /dev/sdb as a mirrored RAID device on installation (mdadm under the cover).

The way I read this is that what happened was that the board's RAID implementation was ignored by Anaconda and you ended up with a full software RAID that Anaconda created.

I'd personally vote for a software-only RAID if you're single booting, unless you can afford Hardware RAID. I don't quite trust any FakeRAID implementation. My main problem with FakeRAID is that if your motherboard croaks, your array is lost... unless you had the forethought to buy a second board from the same production run (i.e. with the same hardware revision of the fakeRAID chip) as a cold standby. With Software RAID, if your motherboard croaks, you can even go so far as take the drives and put them into another working machine, boot up with (e.g.) live media, and recover that way.

Of course, true software RAID entails some (possibly heavy) processor overhead.

[wildwolfe]

forkbomb --> good points. The more I ponder over this, the less I care for this non-feature. Since I don't seem to have any real issue at the moment and that there is nothing sacred on this box, I'll de-fang this serpent when I move on the F13.

Thanks for your comments.

[JEO]

Palimpsest likes to error on the side of caution when it says one of your hard disks is failing. Look at the output of the command:

smartctl -A <drive device>

and see how many sectors have been reallocated. If the number is increasing over time then the drive is indeed failing.