Describes my recent experience with a Kubernetes cluster which the master had low memory and slow disk and how it affected the cluster.
Rename RDMA NIC interface names (kernel names, pci, guid, or fixed).
Setting up NFS over RDMA and benchmarking it.
Flash a 10G Mellanox ConnectX-4 Lx NIC to 25G.
Test the InfiniBand performance between 2 nodes.