Expanding a storage unit on hsg80

[Denzo Waruna]

I've got a storage unit D1 with 8*18.2GB drives in raid 5. I recently bought an additional 5*72GB drives and I"d like to expand D1 - How do I go about it?

[Uwe Zessin · Accepted solution]

You create another RAIDset from your new disks. Then you put a 'concatset' between the first RAIDset and the unit. Finally you add your new RAIDset to the concatset.

> ADD CONCATSET C1 R1
> SET C1 ADD=R2

Note that it can be done only once - you cannot chain more than two storagesets with a concatset.

[SAKET_5]

Hi Denzo,

I recently went through the same exercie on one of our HSG80 controllers.

Please follow Uwe's procedure. It works like a treat. Be aware of the following tho:
1. A CONCATSET can only contain two members, thats it.
2. You can not delete any members from the CONCATSET, so once D1 is made up of two raidset sets e.g. r_raidset_1 and r_raidset_2, you cannot delete any of the raidsets.

Depending on the host that is using D1, you will need OS specific procedures/tools to extend your OS level logical volumes. For Windows you may use Diskpart to extend basic disk partitions or if using Dynamic disks, just right click on the volume and extend it.

Hope it was of help and don't forget to assign points:)

Regards,

[Hein van den Heuvel]

Which OS? LVM?
How do you plan to allow the OS to make uses of the grown space?

It just does not feel right to add those 72GBs to the 18GBs.
The most likely thing that will happen is that the IO distribution will be based on size, not spindles.
So the 72 GB drives, having 4x more block will over time get 4x more IO/sec thrown at them. Is that realy what you want?

What is the problem will just putting them in a new group and expose that to the world (OS). Then, at the OS level you can still concatinate the units (LSM) or Tru64 AdvFS 'addvol'. You may then still be able to control the file placement somewhat, whereas you have no control when you expose all drives as a single unit to the OS.

Also... Perhaps you should consider to put the 5 x 72 GB in a RAID-1 group, not Raid-5 in order to give them a little performance advantage over the 8 smaller spindles. This is assuming that you do not critically need every last block of the new space but that you are sensitive to performance.
I suspect (never tested) that 8 drives in raid-5 will perform much similar to 5 drives in raid-1, or at least much more similar that those 5 in raid-5.

hth,
Hein.