EVA4400 - Disk usage alert

[JorisK]

Hi,

I'm about to reinstall an EVA and i heard different stories about what happens when disk failure occurs when all diskspace is used for a 100%.

Here's my question:
What happens when you set Disk drive failure protection to 'Single' and set the alarmlevel to a 100% and dedicate all availible space for raid5 vdisks?

Someone said i would have no disk protection since the disks are for a 100% in use, someone else said only one disk could fail and the EVA would not be able to automatically level the data but no data would be lost.

Anyone?

[Eemans Dany]

Hi,

If you have your EVA set to single or double diskfailure, the EVA will reserve sufficiant diskspace to rebuild a disk in case of a faillure.
This diskspace is hidden from you and is not presented as usable (free) diskspace.
So if you fillup you disks to 100% there is still space free in case of a diskfaillure.

if you use the same settings (single or double diskfaillure) and diskspace to 90 or 95 % you get a warning (treshold) which can be set to 100% to disable.

If you have less the 100% filled up, the EVA is going to use the free space first (in case of a diskfaillure) and then the reserved space.

Hope this helps a bit.

Dany

[JorisK]

Thanks.

So in a case of 100% usage when a disk fails, the diskgroep will be still online right? (but in degraded mode?)

[Víctor Cespón]

The "Disk drive failure protection" takes some space from disks to have space to rebuild the RAIDs when a disk fails.

This is space is reserved and not available for vdisks, so it does not count for the disk group total.

When a disk fails, the EVA tries to rebuild using the free space on the disk group. If you have all occupied by vdisks, then it uses the space reserved through "disk drive failure protection".

But be aware that rebuild and leveling need a minimum of 5 GB free. And also, to be able to ungroup and replace a disk you need enough space free on the disk group.

Having the disk group almost at 100% and "disk drive failure protection" = None can lead to some disks reaching 100% while trying to rebuild, and then you cannot do anything (can't add disks, ungroup, leveling...) until you free some space by deletting vdisks.

[Marcus Schack]

VCESPON is bang on!!! If you set the alarm level at 100% you will not be able to do anything with the EVA should you have a failure. Even trying to delete LUNs can take forever.

The Best Practice is never to go over 90%, ideally leave it at 80% and have sparing set to SINGLE. Don't put yourself into a position where you put your data into jeopardy. Unlike a server, you dont want to have to "reboot/reinitilize" an EVA or have to recover from backup.

[Marcus Schack]

I forgot to add this document to my reply. It is a Best Practice paper that wil give you more insight.