topic Split site failover? in HPE Nimble Storage Solution Specialists

Split site failover?

k_erik — Mon, 10 Aug 2020 16:36:20 GMT

Hello,

Does anyone know of a split site failover setup? As in site A fails and 50% of the workload goes to site B and 50% to site C. 100% virtualized on VMware. Since the Nimble arrays support multiple replications partners it should be theoretically possible, but it gives me the shivers in regards of complexity.
All sites active, so a failure on site B should fail over 50% of the workload to site A and 50% to site C. Same for site C.
The VM network is virtualized so all sites have access to the same VLANs to run VMs.

I've got an executive who pushes for that, even with 4 sites so only 33% workload goes to each of the remaining site.

I can't find any documentation that it isn't possible, but I can't find anyone who has done it either. Any thoughts?

The end goal is to run the compute workload higher than 50% on each site and still be able to fail over to the remaining sites without any resource contention in a DR scenario. Which means redistribution of the workload if there are changes is also important.

Re: Split site failover?

Candhade — Wed, 02 Sep 2020 17:10:59 GMT

With new NOS code 5.2.1.x and above, you can replicate same volume to two different arrays.
you can make replicated volume online when array A is not available. volume can be online on array B but it can't be online on array C at the same time.
As i know, Nimble Arrays do not support split site failover. However, this concept may be implemented in future.

Re: Split site failover?

k_erik — Wed, 02 Sep 2020 18:11:19 GMT

Thanks, luckily I managed to get the exec away from the idea.

His plan wasn't to replicate each volume to two arrays, but 50% of the load (by capacity and performance usage) to one downstream array, and the other 50% to a different downstream array. Then rebalance this automatically when the load changed.
I told him I understand the concept, but that I can't create a such an advanced software.