can anyone clarify why sip trunk registration fails with 408 request timeout when hitting 50 concurrent calls in ap-southeast-1? running jmeter 5.6.2 against the telephony api and the logs show connection refused after the threshold. is there a hard limit on sip endpoints per tenant for load testing?
Take a look at at your SIP trunk configuration limits. Zendesk lacks native telephony, so this feels different, but GC has specific concurrency caps per edge. Check Admin > Voice > SIP Trunks. The 408 suggests the edge is overwhelmed, not a tenant-wide hard limit. Verify your BYOC edge capacity in ap-southeast-1.
I’d recommend looking at at the edge capacity configuration in Admin > Voice > SIP Trunks. The 408 Request Timeout usually indicates that the edge node in ap-southeast-1 cannot allocate resources for the new session, rather than a hard tenant limit. When running load tests with jmeter, ensure the callRate matches the configured concurrency limit. If the limit is 50, the system may drop registrations when the buffer fills. Check the edgeCapacity settings for your specific BYOC trunk. Sometimes, increasing the maxSessions value or distributing load across multiple edges helps. Also, verify that the SIP Keep-Alive interval is aligned with the edge timeout settings. This prevents stale connections from consuming capacity. If you see connection refused in logs, it often points to a local port exhaustion on the edge server. Review the auditTrail for failed registration attempts to identify patterns. Adjusting the timeout values in the SIP trunk configuration might also provide a buffer during peak load.