Looking for advice on why our Architect IVR flow is dropping calls with a timeout error when we simulate high concurrency using JMeter. We are trying to validate the call capacity limits for a new customer onboarding process, but the system seems to choke unexpectedly.
Background
We are running a load test to determine the maximum concurrent calls our Genesys Cloud organization can handle before performance degrades. The test environment is configured with the standard US-East region settings. We are using JMeter 5.6.2 to simulate SIP trunk traffic. The test plan creates 50 concurrent calls that route to a specific IVR flow. This flow is relatively simple. It plays a welcome prompt, collects a 4-digit extension, and then transfers the call to a queue. The prompt is stored in Genesys Cloud and is an MP3 file with standard sample rates. We have verified the audio codec settings are correct based on previous discussions.
Issue
When the concurrent call count reaches approximately 45, we start seeing a significant number of calls failing. The JMeter sampler reports a 504 Gateway Timeout error. In the Genesys Cloud interaction logs, these calls show as “Abandoned” or “Timeout” before they even reach the play prompt step. The error occurs consistently at the same concurrency level. We are not hitting the API rate limits for the REST endpoints because this is media traffic, not API calls. The WebSocket connection limits seem to be within the acceptable range for our license tier. However, the media server appears to be unable to process the new call requests fast enough.
Troubleshooting
We have checked the following:
- The SIP trunk configuration is correct and allows for the expected bandwidth. We increased the max session limit on the trunk to 100.
- The IVR flow does not have any complex logic or long-running actions. It is a linear flow with one prompt and one transfer.
- We monitored the CPU usage on our JMeter server. It stays below 40% during the test, so the load generator is not the bottleneck.
- We reviewed the Genesys Cloud system health dashboard. All services are green. There are no reported outages.
We suspect there might be a hidden limit on concurrent media streams for our specific license or configuration. Or perhaps the media server is experiencing resource contention. Has anyone seen similar timeout errors during high-concurrency IVR load tests? What configuration changes or limits should we investigate next?