rmc_dell

updated

9 years ago

R

rmc_dell

1 Rookie

47 Posts

0

4593

August 1st, 2017 11:00

Pool Drive Failure

Unity 500F Storage Array.

Pool1 12+1 (14 SSD disks drives).

Can I lose more than one disk drive at the same time in the Pool without losing any data?

  • rmc_dell

    1 Rookie

    47 Posts

    3778

    0

    Posted August 1st, 2017 13:00

    The Pool is a Dynamic Pool.

  • 3778

    0

    Posted August 2nd, 2017 00:00

    Hi,

    That would also be true for traditionnal pools.

    For both pool types, probability is higher with 12+1 than 4+1. But two drives of the same raid set means DU, or worst DL if disks wont spin up again.

    For dynamic pools, if the a RAID5 4+1 choice means a DU when 2 disks amongst 200 (for example) fail at the same time, that is a risky option.

  • 3778

    0

    Posted August 29th, 2017 01:00

    I have experience with VNX, and always prefer to use raid 4+1 in traditional pools. A pool with 20 disks would have 4 internal raids, so it could be possible to loose more than one drive as loong as they belong to different raid group.

    But now I am planning to by a 450F with 20 1,6TB ssd, and I am not sure what configuration select:

    - 1 dynamic pool with 20 disks (23,5 TB)

    - 2 dynamic pools with 10 disks (21,7TB)

    I am worried about the fact of loosing more than one drive at the same time.

  • Rainer_EMC

    6 Operator

    8645 Posts

    3778

    0

    Posted September 1st, 2017 06:00

    a LUN isnt restricted to one raidgroup

    if you want to be protected against dual drive failue than you should use RAID6

  • 3778

    0

    Posted September 1st, 2017 08:00

    The problem for raid6 will be the performance. It is not recommended.

    That is the reason for me to consider 2 different pools.

    The question is that if I create a traditional pool with 20 disks, I can choose raid5 4+1 and could (potentially) tolerate 4 failed disks. With dynamic pool, it internally create one isolation domain every 64 disks, so for me it seems more probable 2 failures in 64 disks thant 2 failures in 5 disks.

    I suppose that EMC should have a good explanation for doing this, but I cannot find it. Even xtreamio uses somethink like raid 6.