Disk Group Space Calculation Inconsistency

[mt_1299]

EVA 4400
4 Disk Groups
Disk Drive Failure Protection = Single on all Disk Groups
All disks 146GB raw, 136.5 formatted

For 3 of the 4 disk groups, I am able to calculate the Disk Group total capacity using the following formula:

(Disk Grp spindle count x 136.5) - (2 x 136.5)

where 2 x 136.5 is the space reserved for Single disk drive failure protection

For example: Disk Group 1 has 16 disks

(16 x 136.5) - (2 x 136.5) = 1911 GB

This number matches the Command View / Disk Groups / Group 1 / Total Capacity value.

For the remaining disk group with 52 disks, I would expect:

(52 x 136.5) - (2 x 136.5) = 6825 GB but Command View shows a Total Capacity of 6711.

The disk groups for which the calculation works, have either 8 or 16 disks each. Could the issue be related to RSS and not having a disk count divisible by 8?

Any insight would be much appreciated.

[Víctor Cespón]

Small differences like that can be due to the RSS distribution and the state of the leveling. Checking disk occupancy sometimes not all disks are at the same %. This depends also on whether there is an RSS with an odd number of disks (so less RAID 1 data can be stored on it).

That can change the next time a disk is added or removed from the disk group, and the RSS distribution changes.

[mt_1299]

Thanks for the insight.

Am I able to use SSSU to gain any more detail?

Disk Occupancy alarm level is set to 90% for all Disk Groups and no leveling appears to be taking place.

I have heard conflicting reports -- one that HP Best practices is that the occupancy alarm level should not be below 90% and another that this is an old recommendation and that the new recommendation is that each Disk Group should have a minimum of 5GB of unused space.

Any thoughts on this?

[Víctor Cespón · Accepted solution]

There's another thread about that:

http://forums13.itrc.hp.com/service/forums/questionanswer.do?threadId=1415181

5 GB is the absolute minimum, to be able to do rebuilding and leveling.

"Disk failure protection" is there to make sure that you have at least 2x(biggest disk) free. If free space is needed, can be taken from there, until a new physical disk can be added or replaced.

Of course the alarm level is there for a reasn too, specially if you use demand-allocated snapshots or Continous Access. Those use a variable space, and if you have only a few GB free, you'll run out of space pretty quickly.

In the case of snapshots, if a snapshot needs more space to grow and there's nothing left, all snapshots will be invalidated and deletted. This can ruin your backup strategy.

In Continuous Access, is the log grows until it fills all empty space, the DR Group will be marked for a full copy. This will have a significant perfomance impact. We have had cases of customers with so litlle space and links so slow that this leads to a chain reaction, where the data from other DR groups cannot be replicated quickly enough, so they end up logging and starting full copies too..

[mt_1299]

Thanks for the help.

Are you able to recommend anything on using SSSU to get more insight into RSS?