Sometimes you got to answer specific questions with solutions that do not exist (in your portfolio). This is one I got this week:
Prerequisites:
* streched NAS solution (homefolders for VDI)
* active/active on failover
Background information:
* 2 sites close to each other with dark fiber in between
The fist challenge is finding Active/Active stretched NAS clusters. They just do not exist. There are single namespace solutions that will reopen your session but the files will be closed and reopened. So this is not an option.
Now think out-of-the-box … there is a dark fiber and the customer tells you there is ‘no’ latency between the sites. This opens doors! Why don’t we see it as one dataroom with 2 racks? This makes it easy: take any HP x9000 cluster and your set. If one controller fails, no sweat. But what happens if half of the enclosures fail? Ouch again. So now we need to find a solution to lose half of your disk enclosures. Mmm … wait a minute, we got P4000!
So here’s the solution we came up with:
DISCLAIMER: this is not (yet) approved by HP technicians so is not really supported at this time. We still need to look at the maximum latency these 2 solutions can handle. Will follow up soon.
UPDATE: the design has been disapproved by an HP technician at HPdiscover. It appears that the x9300 Active/Active gateway relies on the iLO to stay online which would not result in a failover on a site-to-site link failure. My point of critique: what will happen on a MoBo crash? iLO dies too in that occasion.