Could someone clarify why the Architect flow is dropping connections when we ramp up concurrent IVR legs? We are running load tests using JMeter 5.4.1 against the Genesys Cloud platform to simulate peak call volumes. The setup involves a simple IVR flow with a ‘Play Prompt’ task followed by a ‘Collect Input’ task.
At around 150 concurrent sessions, the Play Prompt task starts failing with a 504 Gateway Timeout error. The logs show the WebSocket connection remains open, but the audio stream never initiates. We have verified that the underlying SIP trunks (BYOC) have sufficient capacity and are not hitting rate limits. The issue seems isolated to the Architect flow processing capacity during high concurrency bursts.
Environment details:
- Genesys Cloud Org:
US-East - Architect Flow Version:
v1 - Load Tool:
JMeter 5.4.1 - Concurrent Sessions:
150-200
Has anyone seen similar latency spikes in the IVR flow engine under load? We are wondering if there is a hidden throttle on the ‘Play Prompt’ task or if we need to adjust our WebSocket keep-alive settings in the test script. Any insights on scaling Architect flows for high-volume IVR scenarios would be appreciated.