HPE Storage Users Group

Does anyone have a guide, or best practices on adding disks to a 7200 or 7400?
I have a 7400 with 288 drives, running Raid 6 (6+2) HA availability.
14 M6710 enclosures, so a total of 18 disk shelves.

I was wondering what would be the be recommended way to add disks, but keep the HA availability and raid 6 (6+2).
As i have 18 shelves in total, is it recommended to add 2 disks per shelf?, for a minimum of 32 additional disks?

Is this a 2 node or 4 node?


14 enclosures plus 4 nodes is only 16 enclosures, where is your 18 disk shelves number coming from?

As far as I know you must add drives in 2's with 2 nodes and 4's with 4 nodes. To keep shelf/cage level HA you'd need to add the same number of drives, so assuming you had 16 shelves you should be able to just add 16 drives (as 16 is a multiple of both 2 and 4)

Hi, it's a 4node 7400, you right 16 shelves in total, not 18.

As spencer said you need to add in pairs across all cages in order to maintain cage availability. I thought there wasn't a requirement for 4 drives for a 4-node system (rather 2 drives), but I can go back and look at some documentation on this. At any rate you need to add across all cages of the same type of class of disk (SDD, FC, or NL) at the same time to maintain a balanced system with cage level availability. I am assuming that since you are running RAID 6+2 that everything is in NL?

_________________
Nathan Bell
Solution Architect Principal
Dyncorp International

Nice, just what I was looking for. I've got 8 more disks for my 3PAR 7200 2node on the way right now. I'm currently only running 16 disks (8 in each cage). I'm also looking for a best practices guide or something similar for this. The HP engineer is asking if we want to schedule it or if I can do it myself... anyone have an overview or guide? Here's my best guess:

Apply the license through the HP IMC.
Pop the disks in physically (4 in each cage).
Uh... now it gets hazy... since these are FC drives (I'm all FC) and I'm planning on adding them to my production CPG, I'll have to run tunesys right?

For HA cage you just need to think how the stripes get built. Disks must be added in pairs, this is because it's a symmetric active/active system and so each node will grab half the disk and create its own raid 6 (6+2) LD's. Therefore for 6+2 with HA cage, minimum would be 8 disks per node (16 total).

If you wan't to keep a 4 node system perfectly balanced (which isn't essential) then you'd repeat the same on the other node pair for a total of 32 disks. Alternatively you could add the remaining 16 disks to the second node pair as a later upgrade further down the line. Of course all of this assumes you want to guarantee HA cage across 100% of your capacity which isn't always the case.

The other bit is the number of cages for raid 1 you need min 2 cages per node pair. For raid 5 the cage number per node pair must match or exceed the stripe size. For raid 6 the cage number per node pair must be minimum half the stripe size or greater.

Admit HW should pick them up automatically after a few minutes, yes you'll want to use Tunesys or DO to a new CPG

Hey thanks for the reply Cleanur, it actually ended up being really easy. After I applied my new licenses for the disks plus their vcopy and DO, I slid the new drives in (4 in each cage) the auto_admithw picked them up instantly. Did it late Tuesday and had to run to San Fran so I'm just getting back and checking out how they've been getting on. Pretty nice to see the 3PAR intelligently creating all new LDs on the new drives. I was 86% allocation on all physical drives, today im at 78-79% on each disk, while allocation on new drives is of course growing. The 3PAR makes life too easy for 1 CPG chumps like myself.

Hi fellow chump, Gakecs. Similar situation here; 2 node 7200 with 36 disks. I'll be adding 12 more, 6 in each cage. Did you wind up creating new CPG's and migrating to them? I only have a RAID 5 and 1 CPG.

From this thread I'm assuming you physically insert the drives, wait for them to be integrated into the system, create new CPG(s)... then it becomes fuzzy. Migrating volumes from one CPG to another; is this dangerous? Will a 5TB database require downtime or become unresponsive while DO happens? Why is a CPG migration even needed? Our HP reps made it out to be as simple as, "If you need more space, add the drives and you're done.". No mention of migrating vv or anything.

You don't HAVE to do anything, assuming you didn't select individual disks when creating your CPG (bad idea!) all the vvols will grow into the additional space, however for best performance a tunesys is advised as it will respread the existing chunklets across all the spindles (new + old) providing maximum and consistent performance ( if you don't tunesys then the existing vdisks will only use your existing spindles and future vdisks may use the remaining free space + new spindles until your free space on the existing spindles and then they will only be able to use the new spindles). Tunesys is done as a background priority process so you should not notice a performance impact (assuming you aren't running at near 100% IOPS all week long) and you certainly never have to take anything offline to allow the system to do its thing.