Hello,
we are deploying SQLSERVER and NAS servers on HYPER-V 3 which have virtual HBAs to access directly LUNs in the SAN.
We have in the hosts 4 physical HBAs ports. We are using 2 for accessing the CSVs and 2 for fiber vSwitch (normally we put the operating system in CSV and dedicated LUNS: logs, databases... in direct assign LUNs). We have two physical fiber
switches in our SAN and each one of the HBAs (of each group:CSV or for vSwitch) goes to one of them to assure fault tolerance.
I haven't found any best Practice for how to establish fault-tolerance/multipath aggregation. As I see , there are two options:
1. Create a fiber vSwitch with two physical HBAs and assign one or two virtual HBAs to each machine. This way I do the same zonning on both switches, but I don´t know which path they will go.
2. Create two fiber vSwitchs (one with each HBA) and two virtual HBAs on each guest (one over each vSwitch) and do a different zonning in each switch.
We are using for the firsts guests option one, with only one virtual HBA for each guest (it is more a prove that a final design). But I think the second one is more like the physical way and you assure two paths (also you have less innitiators/masking in your Storage). I don´t really know (and haven't found a clue) how HYPER-V assigns the virtual HBAs to the physicals ones (round robin?, load...)...
Any best pratice or advice?
thanks