4

When using RAID1, is it best practice to buy spare disks at the same time as the original disks to ensure you have the identical type in case they become unavailable in future?

4
  • 1
    If you do, be sure to buy them from different vendors, or buy them delayed in a different batch. That way if they make a dodgy batch that wont work or will break easily then you can quickly put in your backups. Reduces risk of getting an entire set all with the same problem
    – user176486
    Jun 4, 2013 at 15:36
  • Simple answer - YES. Buy spare drives along with any mounting needed for the chassis (so you wont waste time unmounting and remounting). If the RAID controller and chassis support a hot spare keep one in place. Also, make sure you setup event monitoring and REPORTING for a failed drive. You don't want the hot swap to go auto online without being notified. Mark H
    – Mark H
    Jun 12, 2013 at 14:42
  • @MarkH shouldn't there be some sort of cost-benefit analysis for this? Blindly suggesting that everyone should buy extra hardware that is likely (on the whole) to never be used seems wasteful at best.
    – Chris S
    Jun 12, 2013 at 15:56
  • Is what unavailability? mean, keep finding ST380011A in ebay when am on a need for one. What is the lifetime for such a disk?
    – 178024
    Jun 16, 2013 at 18:43

3 Answers 3

3

You won't need an identical type.

The only "hard" requirement with RAID controllers is that the replacement disk is the same bus type (do not replace SAS disks with SATA or vice-versa) and at least the size of the original disk. Software RAID solutions would not even enforce the same bus type.

It makes sense to have similar performance characteristics (rotational speed, peak transfer rate, access times) on all members of an array so you would not give away performance, but if you replace a slower disk with a faster one, this is of no concern.

4
  • Sorry, but many raid controllers have no problem replacing SAS with SATA - just make sure there is enough space. The rest is done by the controller. Not saying it is wise...
    – TomTom
    Jun 4, 2013 at 11:19
  • 1
    @TomTom which ones? The ones I know (LSI, Adaptec, HP) do not support this.
    – the-wabbit
    Jun 4, 2013 at 11:48
  • Adaptec does, or? Where does it say no? Never had that case, though - but is intersting as I am planning a SAS hot spare also for SATA ssd.... if that does not work... better get more SSD.
    – TomTom
    Jun 4, 2013 at 11:54
  • 4
    Some controllers support SAS or SATA but very rarely mixed in the same array.
    – JamesRyan
    Jun 4, 2013 at 13:13
12

There's no real best practice about this. It depends on a few things.

  1. How critical is the system? If its very important and not clustered in any way, you may want a hot spare, which would be purchased with the server.

  2. What is your support contract like? Do you have 24x7x365 4 hour response time? Is 4 hours an acceptable window to run degraded in? If so, you may not need an extra disk.

  3. Is the server clustered or do you have any sort of system-level redundancy? If so, a node being offline is less important. In this case you may not need another disk at the time of purchase.

  4. Do you have a lot of systems that use the same model of disk? If so, maybe you should keep 1-2 cold spare on hand in case you don't feel like waiting for that 4 hour support window to roll around.


It's not necessary to always use the exact same make/model of disk as long as you get them directly from your vendor of choice. If you buy Dell or HP, always get your replacement disks from Dell or HP. This will ensure compatibility. Dell routinely shipped me 300GB disks to replace dead 146GB disks when they ran out of 146s in their warehouses. There's no problem using a different size as long as the other characteristics are the same. The important thing is keeping your server in a supported state.

6
  • 1
    On that note, if it's a hot spare ensuring the drive comes from a different batch (not sure how this could be done, though) to reduce the chance of simultaneous failure is a good idea. Usually this is when going with a 3rd party rather than Dell/HP though.
    – Nathan C
    Jun 4, 2013 at 11:03
  • 2
    I hear this from time to time, but with 4 hour response (which you should have for critical servers) the odds of two disks failing in that window after they've been burnt in is very low. Also, even if you lose the hotspare too, you're still not down. I don't subscribe to "buy from different batches" as long as your server is under support from a major vendor, it's too much effort when you hit a certain size. But, of course, to each their own.
    – MDMarra
    Jun 4, 2013 at 11:06
  • I see. I hear different recommendations from different people. Just other things to consider.
    – Nathan C
    Jun 4, 2013 at 11:08
  • Yeah, it's definitely worth discussing, and can be done when you only have a handful of servers, but when you start getting into triple digits, using other channels the purchase disks to ensure a different batch becomes unwieldy. Also, you have to keep track of "is this disk under Dell warranty or vendor X Warranty" when it fails. It's usually much easier to just call Dell and yell "DEAD DISK" into the phone until they send a replacement :)
    – MDMarra
    Jun 4, 2013 at 11:11
  • 1
    This is all true. When you get to a certain level of hardware, rely on manufacturer spares or just buy the same OEM model/line of drive. Don't over-think this. The only case of replacement drive incompatibility I can recall is an extremely-rare condition where HP was using 73.5GB disks instead of 72GB disks; even though they had the same part number.
    – ewwhite
    Jun 4, 2013 at 12:13
0

Plus one other thing to remember:

Spare disks do have a way to get "used" - and are not available when in urgent need. So you must make very sure you really have them as spares and not plug them into any server that just needs some extra gigs...

tsg

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .