NVMe load balancing between nodes

[FrankWest]

Hi,

we have a 2-node AFF-A250 running ontap 9.14.1. I'm experimenting with NVMe over TCP and have a question:

how can you load balance namespaces between the 2 nodes? When creating a namespace it's creates the volume automatically and chooses the aggregate by itself. How can you ensure that the NVMe namespace load is evenly distributed between the 2 nodes in the cluster?

[elementx]

> NVMe LIFs and namespaces must be hosted on the same node.

https://docs.netapp.com/us-en/ontap/nvme/support-limitations.html#additional-limitations

[FrankWest]

Hi,

I have 4 lifs. 2 lifs per node for NVMe. Those were created after entering the 4 IP addresses when creating the SVM.

[elementx]

That still fits within limitations described in that limitation, IMO.

*Within* a node load balancing to the same volume may be possible and happening (I haven't used NVMe/TCP, so I don't know for sure), but that wouldn't be your ask which is load-balancing to the same namespace *across* different nodes.

[jaikumar]

Only way is to loadbalance between NVMe TCP IO workloads between 2 Nodes is use of Volume Move, Volume is NonDisruptive ( we set ana transition ).

--
Note:

• NVMe is still Active/Passive on both ASA and AFF.
• Minimum 1 NVMe TCP LIF on Namespace owning Node and its HA Partner
• We can use NVMe Host Priority If wanted throttle the IO or QoS 
• We dont Support NVMe Namespace Move 
• And may be use namespace statistics to IO workload between the Node.