TimeoutException: Connect timed out
Running a stress test against our main IVR flow to determine the breaking point for concurrent voice sessions. The setup involves JMeter 5.6.2 hitting a custom REST API endpoint that triggers the api/v2/architect/flows/execute method. We are simulating 500 concurrent users ramping up over 2 minutes from our Singapore-based load generator.
The first 200 calls complete successfully. However, as the concurrency hits 300+, the REST API responses start failing with 504 Gateway Timeout. The Genesys Cloud logs show the Architect flow actually starts processing, but the HTTP response never returns to JMeter within the 30-second timeout window. We have increased the JMeter HTTP Request sampler timeout to 120 seconds, but the 504s persist.
The flow itself is simple: Get JSON data from a CSV file, play a greeting, and transfer to a queue. No complex scripting or long-running tasks. The issue seems to be related to the platform’s ability to manage the state of these concurrent Architect executions via the API rather than the media path itself.
Environment details:
- Genesys Cloud Org ID: [Redacted]
- JMeter Version: 5.6.2
- Load Generator: AWS EC2 t3.xlarge in ap-southeast-1 (Singapore)
- Architect Flow: Simple IVR with CSV data lookup
- API Endpoint: POST /api/v2/architect/flows/execute
- Concurrency: 500 users
- RPS: ~25 requests per second
We suspect this might be hitting a hidden rate limit or connection pool exhaustion on the Architect execution service. Is there a known limit for concurrent API-triggered Architect flows? Or should we be using a different method for high-volume IVR testing? Any insights on how to tune the JMeter config or Architect flow to handle this load would be appreciated. We are trying to avoid hitting 504s in production during peak hours.