Physical disk and data migration.

[markbridgettuk]

Hi all,

I have a requirement to physically move disks plus their data from one 3PAR to another (please don't question why!).  Has anyone actually done this and if so what was the procedure used (obviously to ensure data integrity).  I am guessing this must be possible somehow.  I am happy enough if this is possible by moving (for the sake of argument) the first 8 disks in the array assuming they are in the same CPG of course and putting them into the new array in exactly the same slots.  Ideally I would like to be abe to be more flexible, but either option would be acceptable.

thanks

Mark

[Sheldon Smith]

"Plus their data"? No. Not aware of any way to do that.

Outside of possible questions of system drive licensing, Service Agreement contracts, and incompatibility between various models of 3PAR system, the physical disks get added to the target system using either admithw or admitpd. Either of those puts all the newly added disks' chunklets into the initialize queue where they are wiped before being turned over to the system.

Could you move data to small volumes on the source system then replicate using Remote Copy or Peer Motion to the target system? Once moved, delete source system small volume, move physical disks (which is non-trivial), and repeat?
It still assumes the drives are physically compatible between the systems, etc.

[markbridgettuk]

Thanks Sheldon - your alternate migration suggestions are certainly on the cards (and indeed maybe the only mechanism available).

In this specific instance the arrays could be made identical (e.g. configuration, array model, O/S version and so on so incompatibilities are not an issue to us).

One of the reasons for wanting to do this is speed (I am not at liberty to go into specifics on an open forum). and "simply" moving disks would have been ideal!

I was kind of assuming HPE must have some mechanism of doing this from a support perspective - e.g. if an array suffered some kind of fundamental failure requiring the replacement of the controller nodes.  Our HPE assigned storage guy is looking into this for us as well.

I may be on a hiding to nothing though in which case RC or Peer Motion will need to be considered and may be the only viaeble opton!

[Sheldon Smith]

Yeah, there's just no way to move the metadata which keeps track of which chunklet goes to which logical disk and from there to the admin and data regions of the virtual volumes.