Intermittent BGP Flapping due to Aggressive Keepalive/Hold Timers

Products

• VMware NSX

Issue/Introduction

Intermittent BGP flapping between the NSX Edge gateways and their BGP neighbors. The BGP session drops and then re-establishes automatically within a short period (approximately within a minute)

Environment

VMware NSX
VMware NSX-T Data Center 

Cause

The issue is caused by excessively aggressive BGP timer settings.

Upon investigation of the Edge logs, the BGP neighborship is going down due to a Hold Timer Expired event.

2025-12-18T07:03:42.764Z NSX 11636 - [nsx@6876 comp="nsx-edge" s2comp="nsx-monitoring" tid="11673" level="INFO"] Error message: In Router 310#####-####-####-b53b-########4b58, BGP neighbor e60#####-####-4af7-####-########9434(10.###.##.139) is down. Reason: BGP Notification send {Hold Timer Expired}.

Analysis of the running configuration reveals the following timer values:

Example:

• Keepalive Timer: 1 second

• Hold Timer: 3 seconds

While BGP allows for custom timers, a 3-second hold time is extremely aggressive. With this configuration, if a single Keepalive packet is lost or delayed by just 3 seconds (due to transient network congestion, high CPU on the peer, or minor path latency), the BGP session will immediately tear down.

Resolution

To resolve the intermittent flapping, increase the BGP timers to allow for standard network variance.

Recommended Action: Modify the BGP Neighbor configuration to use the NSX default values or less aggressive custom values.

• Recommended (Default) Values:

  • Keepalive Timer: 60 seconds

  • Hold Down Timer: 180 seconds

By increasing these timers, the BGP session will be able to tolerate minor packet drops or latency without resetting the connection.

Additional Information

Reference: For detailed steps on configuring BGP and timers in NSX, please refer to the administration guide: VMware NSX 4.0 - Configure BGP