2

I'm setting up a small, but hopefully durable server for a student community, and the hardware is an IBM x3620 M3 with a builtin M1015 RAID card.

To make the setup future proof and avoid vendor lock-in as much as possible, I want to use software RAID via the generic Linux md raid-option, or maybe even directly via btrfs.

This is where it gets tricky. The IBM Redbook on M1015 does not explicitly state that a complete "pass-through" option (Is that what is called JBOD?) is available. I've booted the server and tried installing Debian on just a single harddisk without setting up any RAID in the controller's setup, so clearly some kind of "pass-through" is going on.

My questions are thus:

  • Does the controller leave some kind of metadata on the disks that forces my md raid-array to always be run off this particular controller, or will it be a completely generic array that I could plug into "raw" SATA-ports and run just as well?
  • Will I be able to hot-swap and rebuild my array, even though I am "bypassing" the hardware controller as much as possible?
  • Am I doing it all wrong? Should I flash a different firmware to the controller, the possibility of which is mentioned on Serverfault itself? Why?
  • etc. (Howto: M1015 software raid on Linux)

Thanks!

Community
  • 1
Sarah
  • 155
  • 1
  • 6

2 Answers2

1

Does the controller leave some kind of metadata on the disks that forces my md raid-array to always be run off this particular controller, or will it be a completely generic array that I could plug into "raw" SATA-ports and run just as well?

If you mean will you be able to take the disks and plug them into any other controller then probably not no - even with passthrough I wouldn't 'bet the farm' on this happening.

Will I be able to hot-swap and rebuild my array, even though I am "bypassing" the hardware controller as much as possible?

Yes, I don't think that would be a problem at all.

Am I doing it all wrong? Should I flash a different firmware to the controller, the possibility of which is mentioned on Serverfault itself?

I'd personally just leave the controller to do the job - I'm no fan of software RAID, for a start you end up carrying all writes over the buses multiple times and whatever 'vendor independence' you think you might have it's always easier to backup a server and restore it to another than risk moving the disks to another server and having no way back - some new servers might just wipe the disks anyway. Trust in the hardware, make sure you backup and test those backups periodically.

Chopper3
  • 100,240
  • 9
  • 106
  • 238
0

I dont think the controller leaves some vendor specific stuff:

I purchased an intertec 4342L (24 bay) with a M1015 controller in order to expand an existing md RAID5. I simply moved the disks to the enclosure. The previously existing IDE boot HD was cloned with dd to a new SATA drive, which then was also plugged into the enclosure. Everything went smooth, the system booted without complaints and the md array started as before. Even smarttool can be used to read HDD information directly.

Roman
  • 3,825
  • 3
  • 20
  • 33
error
  • 11
  • This isn't really an answer to the question. It's also a wall of text. – Tom O'Connor Jan 24 '14 at 13:40
  • Hold your horses, I just edited him. – Roman Jan 24 '14 at 13:45
  • well i just wanted to show how i changed my system from mainboard sata controller to m1015 sas controller (even the boot device from ide to cloned sata) and everything went fine without reinstall. i am not native english so i guess that is how my wall of text came up ;) – error Jan 24 '14 at 13:54