MSA2040 changing disk position

[cgiesers]

Hello!

We're running an MSA2040 SAN/FC with tiering license and SSDs as cache.

We will add more SSDs to the system and since only local storage can run on SAS12G, I want to move some of the HDDs to the, already attached and partly equiped, D2600 case.

Is it possible to shut the system fully down (power off) and just relocate the HDDs, plug the new SSDs in, power it up and everything is fine or will it break my RAID?

(2nd step would be to remove the SSD tier/group, remove the vdisk, create a new one including new SSDs and readd it as SSD tier the pool again)

Regards, Christof

[HPSDMike · Accepted solution]

I have always heard that you can reposition drives (with the system powered off)  in an MSA without affecting data. This is essentially what happens during a data-in-place upgrade when moving between MSA platforms. The second bullet point on page 11 of https://h20195.www2.hpe.com/V2/getpdf.aspx/4AA4-6830ENW.pdf? would back that belief of where it says "The same drive order is not compulsory as the metadata with be identified by the MSA
2040/2042 controllers. Moving drives should only be done while the system is offline."

Of course, make sure you have a backup of your data. 

On your second point you mention having to remove your current SSD vdisk and recreate. You also mention using the SSD's as "cache" now. So, not sure exactly how you are working now. If you are using the SSD's as read cache then you can simply mark additional SSDs as read cache for each pool (up to two SSDs per pool). If you are using the SSDs as a "tier" within the pool then you have two options. Easiest is to create new SSD vdisks and add them to each pool (ideally the vdisk would match the RAID type, and set size, of the previous SSD vdisk). Of course, you can also remove the current SSD vdisks, and recreate them with new drives included, but there will be some implications to this. You'll need to wait for the SSD vdisk(s) to be drained (VDRAIN). There will be some performance implications of the VDRAIN itself along with the fact that you temporarily don't have SSD in the pool. 

[cgiesers]

Hey Mike!

Sorry, that way my fault: It's used as SSD tier, not as cache.

SSDs - especially the ones, you need for MSAs, means SAS/DP - became freakin' expensive, compared to usual SSDs, which are cheap to buy a bunch of, so we just have 5 currently and I will add another 3. Since for, data integrity, they're running as RAID6 (I know, RAID10 would be faster, but for costs reasons and since they area still way faster than the HDD RAIDs...), so there's no way around rebuilding that RAID out of the 8 disks and keep 6 capacity-usuable ones.

I know that I will need to wait a bit and the performance will go down until the new tier is back in and MSA has done some work, but that's okay, as it's just temporary and could be done on a weekend.

Thanks for your help and comments!

[HPSDMike]

No problem and good luck. BTW, it is recommended that you make sure you are running the latest firmware prior to this change (Currently GL220P010). There have been some improvements made to the VDRAIN process over certain firmware levels so best making sure you have the latest. 

[cgiesers]

Hey Mike!

Just to give a final feedback:

Firmwareupgrade went successful (though it needed about 50 instead of anounced "about 30 minutes", patience is mandatory...).

VDRAIN worked flawlessly and SSDs were taken out of use (lights went off, smart!).

I shut the whole system down, but one needs to get used to MSA still running, just controllers being shut down. I expected the MSA to be powered off. So I plugged power cables off of MSA and D2700.

7 HDDs were relocated to D2700 chassis, 3 SSDs plugged into MSA (so I don't need to shutdown for further (4) SSD enhancements, something one should consider when planing an MSA - I didn't).

Cables plugged back and MSA booted up: After a while everything was shown as running (vDISKs as FTOL). Added the new, bigger, SSD-tier and now it's slowly filling SSDs with 'hot' data.

...of course I fully shut down our XenPool before shutting down the MSA, while I kept most (not mandatory) VMs up, when I upgraded the firmware from GL220P009 to GL220P010 (which are both certified for XenServer 7.x, regarding HPE SPOCK). ;-)

Full backup included, the process now ate my weekend, everything is running and noone will (negatively) notice on monday morning.

Thanks!

[HPSDMike]

Excellent followup, thank you!