HP EVA Disk Redundancy Question

[Philip Teale]

Hi

I apologise if this question has been answered before, but I have looked through this forum and can't see an answer.

I am having difficulty understanding the various levels of redundancy that are present in our EVA 4000 2C4D. We have several disk groups, each of which has single disk protection enabled. We run a mixture of VRAID5 and VRAID1. My understanding of the levels of protection are that if a disk fails in one of the disk groups, the space reserved by the single disk protection is used to reconstruct the RAID information from the failed disk, be it VRAID5 or VRAID1. If, for sake of argument, this failed disk goes unnoticed, are we now running in a situation where another failed disk in the same disk group will lead to data loss? Or will the array then (after a second disk failure) be in degraded mode? I also don't fully understand why single disk protection reserves an area of storage twice the size of the largest disk.

I think I understand the concept of the RSS, in that in this scenario (one failed disk goes unnoticed), if another disk in the same RSS fails, no data will be lost, but I am confused as to where the data on the second lost disk will be reconstructed; the reserved space was used to reconstruct the first failed disk was it not?

I hope someone can help.

Cheers

Phil

[IBaltay]

Hi,
a) the single disk protection means 2 virtual spare disks
the double disk protection means 4 virtual spare disks
b) the RSS allows you to survive the 1 disk failure in each RSS at the same time
c) the DG in both RAID1 and RAID5 can survive in a degraded mode only the 1 disk failure in 1 RSS. If 2 disks fail in 1 RSS it is a double disk failure of the whole disk group
d) after the first disk failure, the failed disk starts to rebuild into the spare space. At that time, the disk group is in the warning/"degraded" state. If then any other disk fails there is still another virtaul spare available

[Uwe Zessin]

When a disk drive has failed, the EVA first draws from the space that is available for creating vdisks, clones, snapshots, etc. to rebuild the redundancy - it does not immediately exhaust space that was defined by the so-called 'disk drive failure protection (level)'.

If all free space has been used, yes, the EVA will operate in 'degraded mode', like every other RAID-array.

--
> reserves an area of storage twice the size of the largest disk.

The reason is that the EVA writes VRAID-1 data to pairs of physical disk drives. Those pairs are fixed at a very low level and can only be changed when a new disk group is created or disk drives are added/removed to/from the group.
If one member fails, the VRAID-1 data from the surviving member must be relocated to a new pair of disk drives.

If you have a disk group with an odd number of disk drives, fill it completely with VRAID-1 data, you will see that one disk drive does not contain any user data: because it has no partner disk.

[Philip Teale]

Hi Uwe

Thanks for your reply; I use VRAID5 for all the LUNs presented from a diskpool that has single drive protection. Does that mean that I can lose two disks (as per IBaltay's post above)? Or will the second "disk" only be used in the event of a VRAID1 rebuild being necessary?

Cheers

Phil

[IBaltay]

Uwe forgive me, but I will add my comment too :-).
Phillip I will add some details to my previous comment 2 b clearer:
d)
1. after the first disk failure, the failed disk starts to rebuild into the spare space.

2. At that time, the disk group is in the "warning/"degraded" state.

3. if any other other disk fails (during the rebuilt) in the same RSS, it is a double disk failure of the whole DG and its data is lost

4. After the successful rebuild of the failed disk to the spare space (the disk group will become healthy), if any other disk fails, there is still another spare drive available, though it will cause
that the single disk protection will fail to none.

Therefore it is very important to
5. replace the failed drive ASAP
to always keep the single protection alive

[Uwe Zessin]

> Does that mean that I can lose two disks

Yes, provided that there is no concurrent failure in one RSS. If the EVA has to draw from the 'protection space' in order to recover from the first disk failure, the current protection level will switch to none, but that does not mean that the remaining capacity cannot be used.