Supportability and Performance Impact of Removing Resource Reservations for NSX Infrastructure VMs in VCF
search cancel

Supportability and Performance Impact of Removing Resource Reservations for NSX Infrastructure VMs in VCF

book

Article ID: 453083

calendar_today

Updated On:

Products

VMware NSX

Issue/Introduction

Broadcom provides guidance on the support status, performance risks, and lifecycle management impact of removing CPU or memory reservations for NSX Manager and NSX Edge nodes in production VMware Cloud Foundation (VCF) environments, including VCF 5.2.

Symptoms

Customers seeking to reduce resource overhead or "Cost to Serve" by removing reservations may observe the following issues if changes are implemented manually in vSphere:

  • Network performance degradation or intermittent connectivity for traffic traversing NSX Edge nodes.
  • Increased CPU Ready time (%RDY) exceeding 8% in esxtop analysis.
  • Management plane latency or database desynchronization in NSX Manager clusters.
  • Failures during VCF lifecycle operations (upgrades, cluster expansions, or password rotations) due to "Configuration Drift" detection by SDDC Manager.

NSX infrastructure appliances are deployed via OVF templates with pre-defined resource reservations to ensure deterministic performance for the management plane and low-latency packet processing for the dataplane (DPDK). VCF SDDC Manager expects these reservations to remain at 100% to ensure environment stability and support compliance. Manual modification of these settings in vCenter results in configuration drift and is not a supported workflow.

Environment

  • VMware Cloud Foundation (VCF) 4.x / 5.x / 5.2
  • VMware NSX 4.x

Resolution

Removing CPU and memory reservations for NSX infrastructure VMs is not supported in production VCF environments. To maintain support and performance, follow the requirements below:

  1. Maintain 100% resource reservations for all NSX Manager and NSX Edge VMs as defined by their deployment form factor (Medium, Large, X-Large).

  2. Scale the underlying physical hardware by adding ESXi hosts to the cluster if resource pressure is the primary driver for reservation removal.

  3. Verify current host capacity using the following vsish command to ensure the physical cores can honor the required reservations: vsish -e get /sched/groups/4/stats/capacity

  4. Review physical core frequency to ensure a single physical core can back the per-vCPU MHz reservation (see KB 443517).

If a solution is not confirmed, this guidance remains the official support stance. For updates on fix status regarding lifecycle automation, subscribe to this article see (KB 275360).

Additional Information

Note: This article was created to address specific supportability queries for VCF 5.2 and NSX 4.2 regarding cost-saving measures through reservation removal.