additional redundancy on eva4100

[Prokopets]

Hi everyone !
Can anyone say, is there any way to configure eva, so in can continue work with two failed drives on one volume (maybe - on one vraid)? If it can be done - home many disks at least must be in array in this case?

[Patrick Terlisten]

Hello,

if you configure double protection level for your disk group and the second drive fails after the first was rebuilded, you can lose two drives. Two simultaneous drives failures can also be covered, but it depends on the used VRaid level which disks fails and to which RSS the disk belongs.

Hope this helps.

Best regards,
Patrick

[Prokopets]

Patrick, thanks for reply, but can you clarify some thigs?
1) I read in "EVA best practices" that that's not recommended to use double protection - can you comment it ?
2) Afaik, but not sure, if failed disks are in the same RSS, volume will be dead regardless of vraid level - is it true ?

I mean that we can speak _only_ about probability of stability of array in case of two simultaneous drive failed, but not about something definite, am i right ?

[Uwe Zessin · Accepted solution]

There is no way to configure a single EVA so that it can always cover the concurrent failure of two disk drives. With at last 12 disk drives in a disk group, the EVA will start to form 'sub-groups' (=RSSes) in that disk group, which does improve the situation a bit.

Remember that the 'double protection' (level) only defines how much disk space is reserved so that the EVA can rebuild redundancy after a disk drive has failed, but it does not actively protect your data - that is done by a virtual disk's VRAID level.
No amount of disk drives or setting of the 'protection level' can prevent data loss for VRAID-0 when a disk drive has failed.

I have not checked the best practices right now, but I think the recommendation you mention in (1) is just to save some space for user data. In most cases an EVA is under constant monitoring from WEBES/ISEE so a failed disk gets replaced rather quick.

(2): a VRAID-1 vdisk can theoretically survive as long as no mirror-members are lost within the same RSS. An RSS can have between 3 and 5 of such members.

But as you correctly say: it is not something definite...

[Patrick Terlisten]

Hello,

I never used double protection level for a disk group. IMHO it's just a waste if disk space. As long as you have enough space left in the disk group, the EVA will rebuild a failed disk into the free space of a disk group, and leaves the "reserved" space unused.

Best regards,
Patrick

[Prokopets]

Ok, thank you all for the answers !