which risk to set level of disk failure protection to "none" ?

[patbourdon]

Hi, due to low disk space on my 2 EVA 4400, i study several ways to save space.
One of them is to set "level of disk failure protection" from "single" to "none"... (i already setted it from "double" to "single" some months ago...).
I wonder what kinf of problem can i face if i do it ?
i think :
- no more log space for CA Replication
- no security for raid0 vdisk if i lose a physical disk
- what else ?
Knowing that i have 2 EVA 4400 with CA replication on critical vdisk, no critical raid0 vdisk (only some snapshots), is it dangerous ? what dou you think about this ?

TIA

Pat

[IBaltay]

Hi,
in general it is always better to have virtual spares at EVA. With them the controllers automaticaly ungroup the failed disks to recreate the full health of the whole disk group. Without it each HDD failure means the degradation of the whole Disk group and if another disk fails (in the same RSS) thenafter, it is the data loss of the whole disk group. The disk protection (single/double) (the space of 2 HDD) is therefore highly recommended.

[Patrick Terlisten]

Hello Pat,

the "protection level" has NO effects on VRAID 0 vdisks. If you lose a disk in a disk-group, all VRAID 0 vdisks will lose data, regardless of the protection level.

The CA log is not included in space that is reserved by the protection level. This is only a reservation of space for the rebuild of failed disks. If free space is available in a disk group, the EVA will use first this space for a rebuild. This space is also used to the write history logging of CA. You can put the CA log into another disk group, if there is more free space available.

You can set the protection level to none, if it is for a manageable timeframe. I would not recommend this as a general setting.

Best regards,
Patrick

[patbourdon]

Hi Patrick & IBaltay, i only have 1 diskgroup and no spare disk (1 disk cabinet filled with 12 disk on each EVA).
I plan to buy another disk cabinet but it may take 2 or 3 months to get it...
So i understand that i my case, the risk for setting level to "none" is the loss of the whole group in case of disk failure ? But if the EVA contains the replicated vdisk (destination), may be it'snt a big risk ? in case of failure, i replace disk and the full replication process will start, isn't it ?

[Patrick Terlisten]

Hello,

you only lose data of VRAID 0 vdisks if a disk fails. I hope you're using VRAID 0 not for business critical data. ;) VRAID 1 and VRAID 5 are protected by VRAID redundancy. If free space is available in the disk group, the EVA will start the recovery of the failed disk into this free space. Be sure that there is enough space (at least two times of the greatest disk in the disk group). On top you should have free space for the write history log.

If the Vdisk is replicated by CA, the data is on the other side available, if a disk on the source side fails.

Best regards,
Patrick

[IBaltay]

Vraid0 best practice to optimize availability: Use of Vraid0 is not advised as a best practice for availability.

[Patrick Terlisten]

Sounds like "Best practise for good health: Don't hurt yourself." :D

[Uwe Zessin]

My VRAID-0 best practice is to not use VRAID-0 ;-)

- the failure of a single disk will cause data loss, because NO redundacy is provided. Yes, I've been told/asked that 'the virtualization' provides it or the so-called 'protection level', but it is called VRAID-zero for a reason.

- you can't do an online disk drive firmware upgrade on a disk that stores VRAID-0 data. An error can make the upgrade fail and render you disk inoperable - good bye data!


I am afraid that the (VRAID-0) recovery from a disk failure is more complex than simply plugging in the replacement disk and hoping CA will fix it. I've never tried it myself, but think you will have to delete and recreate the virtual disk(s) and data replication group(s).

[Rob Leadbeater]

Hi,

All the talk of VRAID-0 reminded me of an issue I faced a while ago.

http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=968041

Now things may have changed on the EVA4400 compared to the EVA5000, but I'll quote my final line...

"I guess the moral of the story is don't use a disk group with no sparing, even if you don't particularly care about the data on the disks..."

Hope this helps,

Regards,

Rob

[patbourdon]

well... thnaks all for your answers but it's a little bit confusing for me : i have to translate in french (wine and smoking cheese country...) and make a synthesis of all replies to start to understand ;)))

- i only use raid-0 for temporary snapshot (for backup usage), so no risk i guess ?

- i have 12 disk (FC 300 GB), no spare disk, and only 100 GB remaining in my single diskgroup.

- of course i know that it will be better to have more disks, with spare and separate diskgroup but i work in an hospital and even with special prices it's very hard to buy something, so i try to do my best with what i have ;)))

Knowing that, what dou you think of "the EVA contains the replicated vdisk (destination), may be it'snt a big risk to "noning" protection on this second EVA ? in case of failure, i replace disk and the full replication process will start, isn't it ?"