Drive Failure Protection; mirrorclone

[Ken O'Neil]

On an EVA6100, each disk group has a setting for "Drive Failure Protection", with options of None, Single, Double. Setting this to Single seems to claim additional space 2x the largest physical disk in the group. But if you're building LUNs with RAID 1 or RAID 5, what does this really buy you? Is DFP really needed with RAID?

Also, we're adding an EVA6100 into our storage environment with our existing EVA3000, and there is a new clone feature on the 6100 that we didn't have on the 3000 called mirrorclone. This is in addition to the snapshot and snapclone features, and am wondering how mirrorclone is different.

Thanks in advance.

[Ivan Ferreira]

DFP could be though as the old "spare drives", that is, space that cannot be used, and the data will be reconstructed there if a disk fails.

This gives you more time to replace a failed disk, as your VDISKS won't be in "reduced" state for long time.

MirrorClone with fast resync : MirrorClone now allows a user to pre-normalize a mirror and then fracture the mirror to create an instantaneous byte-for-byte clone. MirrorClone also maintains its relationship with the source volume accelerating the re-sync time.

[Ken O'Neil]

Thanks for the reply and definitions Ivan. To my question though, what's your take on how useful that is in a disk group where all the LUNs are configured as R5? More than a definition, I'm looking for input from you and others to get a sense of perspective on the pros and cons of employing this. I'm not inclined to claim raw space if there isn't any specific benefit and a consultant has recommended a "none" setting. At the same time, I know that features like this are built into a system in response to some type of requirement or need and it must serve some useful purpose, so I'm somewhat suspicious of not using it at all.

In case you need it, here's some basic info on this SAN:
56 x 146GB 15K FC drives in DG1, 14 x 500GB FATA drives in DG2, 5 shelves, all full. SMS w/ CV7.

TIA

[Uwe Zessin]

The so-called "Drive Failure Protection" is a hard limit to reserve disk space for rebuilding redundancy after a disk drive has failed.

It:
- prevents you from accidently filling a disk group with too much data

- prevents vSnaps or the CA Write Log History to eat up space that might be needed for recovery

The EVA does work without any reservation - in fact, if a disk drive has failed, it first draws capacity from the space that is available for other storage (vdisks, snaps, WHL).

Using a level > "none" relieves you from manually monitoring your disk groups - I strongly recommend to use a value greater than "none", "single" is usually sufficient. It is also suggested in HP's best practice document.