WebRTC Softphone STUN Binding Fails on APAC BYOC Trunk Edge

Voice & Telephony
1

Dealing with a very strange bug here with WebRTC softphone registration failures specifically for agents routed through our Singapore BYOC trunks. The standard SIP flows work perfectly for inbound PSTN calls, but when agents attempt to register the softphone via the v2.0 SDK, the STUN binding times out with SIP 408 Request Timeout after exactly 3 seconds. This only happens when the agent’s IP is outside our corporate VLAN, suggesting a NAT traversal issue unique to the APAC edge configuration.

We have 15 BYOC trunks configured, and this issue is isolated to the trunks handling high-volume analytics reporting traffic. The ice_candidate gathering fails silently in the browser console, and no 200 OK is received from the Genesys Cloud signaling server. Direct SIP trunk health checks show all 15 trunks as ACTIVE with sub-50ms latency.

Has anyone seen STUN binding failures correlate with analytics query loads on the same edge node? We suspect the analytics pipeline might be saturating the UDP ports required for WebRTC signaling. We are using Chrome 120 and Genesys Cloud Web SDK 2.15.0. Any insights on isolating WebRTC signaling traffic from bulk analytics exports would be appreciated.

2

The root of the issue is likely a mismatch in the WebRTC configuration for the specific edge location.

  • Verify the webrtc_enabled flag is explicitly set to true in the edge configuration payload.
  • Ensure the STUN server URLs in the SDK align with the APAC region endpoints, as BYOC trunks often require explicit relay settings.