2016-09-20 08:57 AM
I'm building out a new 2 node (a single HA-Pair) V-Series cluster. I'm trying to decide on the best way to spread the load such that both controllers are contributing cpu resources. My original assumption was that whichever node the LIF is currently running on is the node which is handling the NFS traffic and cpu resources. Then I started wondering if instead, the disk ownership plays a role in this as well. For this cluster, I have no reason to separate out vservers or LIFs from an organizational perspective so my goal is just to spread the load so that we utilize both controllers and not have one that's idle. Any advice would be much appreciated!
2016-10-04 12:08 PM
Some resources are consumed on each end. The protocol stack consumes resources from the node hosting the lif, and WAFL and the raid engine consume resources from the node hosting the aggregate. Generally you want the lif to be hosted on the same node that owns the data aggregate being accessed, but thats not a requirement and the architecture is pretty flexible. Block protocols will find the active/optimized path via ALUA, and NAS protocols will use indirect paths if necessary.