1

My lab is considering setting up a server with an attached RAID device for storing our large imaging datasets. (probably a Mac Mini server with attached Promise SmartStor DAS w/ 4 1TB hard drives)

Since there will be 4 1 TB hard drives, I would like to set up a RAID 1 configuration where 2 of the drives (say, A and b), will be mirrored onto the other two (say, C and D).

Here are the questions:

1) Is my understanding of RAID 1 correct to do the above?

2) If, for instance, drive A fails, the mirrored drives will immediately take over, right?

3) When such a failure occurs (such as when A fails), will swapping it for a new empty hard drive (a new, drive E) be all that I need to do? That is, will the still functioning mirrors (C and D) automatically be used to rebuild the new drive (E)? Is this something I should set up in the server software?

4) Hypothetically speaking: suppose I disconnect one of the drives, say B, and connect it to a separate computer, add/remove a few files, plug it back into the RAID 1 group, will this screw things up?

Thank you very much!

hpy
  • 835
  • 3
  • 18
  • 28

5 Answers5

2

1) Correct.

2) There's not really a "taking over", they're always active.

3) It depends. Some systems automatically start a rebuild of replaced drives, others require manual intervention to force a rebuild.

4) In all the systems I've seen, that's an unsupported operation. You might be able to get away with it, but you're not using it the way it's meant to be. I suspect that in most cases you'll lose the changes made. You could possible corrupt the whole array too I suppose.

Brian Knoblauch
  • 2,188
  • 2
  • 32
  • 45
2

That storage box you mention only support RAID levels 0, 1 & 5, not RAID 10 (i.e. the mirroring of multiple striped disks you mention in your second paragraph).

That said the basic idea you mention is right, in that in the event of a single disk failing (being pulled out manually is effectively a failure) then the other side of the mirror would continue to work. Inserting a new working disk does indeed initiate an array rebuild, which when complete would be capable of allowing the array to survive another disk failure.

Oh and your "4)" would definitely screw it up.

I'd be tempted to buy a RAID 10-capable box by the way, oh and make sure it supports Firewire 800 as that's the fastest external interface you have available on a MacMini.

Chopper3
  • 100,240
  • 9
  • 106
  • 238
  • OK. What about setting up "two" RAID 1 configurations in the same box? That is, A and B would be mirrors of each other, while C and D would be mirrors of each other, too? – hpy Sep 27 '10 at 15:45
  • Some boxes will support that yes, but there's some wording in the tech specs of that box about 'one array per enclosure' or something that made it unclear whether you could do that - best ask them directly. – Chopper3 Sep 27 '10 at 16:32
1

Why not use RAID 5 instead of mirroring? With more than 2 drives, it usually makes more sense -- it'll keep running if any single drive fails (like a mirror), but it only uses one drive for redundant data (unlike a mirror, which uses half of the drives for redundancy). You could either config it as a single 4-drive RAID (with 3TB capacity), or a 3-drive RAID (2TB capacity) with a hot spare it'll automatically rebuild onto in case of a failure.

Gordon Davisson
  • 11,036
  • 3
  • 27
  • 33
  • 1
    Hi, there's no reason for you to know but R5 is a bit of a SF bug-bear, with the vast majority of posters, myself included, utterly hating it for dull technical and reliability reasons. Just wanted you to be aware you may, in a polite way, get jumped on - but thanks for your response. – Chopper3 Sep 27 '10 at 19:54
1

Keep in mind that RAID-1 is a way to make 2 drives act like one. So when you write to "storage," you will be writing to both drives. When you read a file, it will read from whichever disk is available first. Theoretically, this makes a penalty on writes and a boost on reads. In reality, you might not notice a speed difference.

The real reason that this is important is that it means that there is no such thing as a drive "taking over" when another drive fails. They're both in constant use. Case in point: I had a Mac Pro with two 500GB drives in a RAID-1 array at work. One day I randomly checked that server and discovered that one of the drives was actually dead for the last few months. No one noticed a difference. This computer was the file server for an entire elementary school.

I'll answer question #3 by telling you what happened after finding the dead HD in the Mac Pro. I replaced the drive with only a few minutes of down-time. I replaced it AND added a third drive as a hot spare. So I now have 3 HDs acting like one storage device, i.e. a single "drive," of 500GB. Again, no one can tell the difference.

As a side note: Please note that RAID is NOT a backup solution. I have had situations [plural] where I've lost data in a RAID array. My backups were well worth their expense those days.

Data Scavenger
  • 477
  • 3
  • 9
1

1) Yes, assuming the drives have compatible sizes (see 3) )

2) none of the drives is the "main" drive, they mirror each other

3) Even if your system automatically rebuilds an array if possible, beware that "1 TB" disk is actually a vendor-dependent term. Depending on your setup, the initial process may have claimed all blocks in those disks. If you get a replacement "1 TB" drive later, even from the same vendor, it may have less blocks. In that case, you won't be able to use it as a replacement. You'll have to get a disk with the same or larger number of blocks being used in the array. Therefore, it is often advisable to limit the array to a number of blocks which most "1 TB" disks should fulfill. Usually 1 TB - a few blocks ~ (10^12 - 100 )B is quite safe. And of course it also depends on the drive partitioning alignment.

4) Depends on the SW used.

Ondřej Grover
  • 131
  • 3