HPE EVA Storage

SVSP questions on how to break mirrored front-end LUN

Occasional Advisor

SVSP questions on how to break mirrored front-end LUN

We have presented a few MIRRORED LUNs from SVSP to hosts. Back-end LUNs are carved out of two EVA8400 frames (Names: EVA1/EVA2).

Storage Configuration:
SVSP (SAN Virtualization) is configured as DPM/VSM cluster, going thru two brocade switches, hooked-up to two back-end EVA8400 storage-frames.

Reading from SVSP best practice document we understand that - on a mirrored front-end LUN, READs occur on the very-first EVA configured when the mirror was initially created, in our case EVA2. Ofcourse, Mirror-writes have to go to both EVA8400 boxes. It so happened that when we originally created mirror LUNs, we had listed EVA2 as the first copy of the mirror. Thus, it appears that most of our READ I/Os on our mirrored LUNs are hitting EVA2.

We be happy to discuss to see/show any way to confirm our understanding on this. Given this scenario - our EVA2 seems overloaded with lots of READ I/Os, and we would like to re-distribute some of those read I/Os to go thru EVA1 box., which appears lightly loaded.

Is there a procedure or best-practice to spread the load in our current configuration? Or any of your suggestions on this? (we can't possibly re-build SVSP front-end LUNs from scratch at this time - we have been in production for 8+ months).

We are reading - there may be a process do perform this mirror split - "pause the mirror, then split it" etc. We would like to perform this step without incurring downtime, if we could.

Any thoughts ? Let us know,