PNIC Queues Get Exhausted on a High NUMA Node System in Enhanced Datapath Standard Mode

On a high NUMA node system in the Enhanced Datapath Standard mode, PNIC queues can get exhausted.

Problem

In a high NUMA node system, individual NUMA nodes are typically connected through a high-speed interconnect, such as a shared memorey bus or a network. Each node in the system has its own memory resources and processor cores, enabling localized access to data and reducing memory access latency. To manage workload with reduced latency, NSX assigns VMs to different NUMA nodes on the high NUMA node system. However, as the number of VMs are distributed across different numa nodes, it is likely that as new VMs are added to the NUMA node system, NSX might not be able to assign a dedicated PNIC queue or MAC filter to these new VMs.

Solution

• ♦ Use the driver parameters to configure more PNIC queues than the default number of queues.