Data Action timeout causing call drops during peak hours after v2.10 upgrade

NICE CXone API & Integrations
1

Organization Details
Genesys Cloud Org: AWS US-East-1 region
Architect Version: 2.10 (Updated 2024-05-14)
Agent Count: 800 agents
BYOC Trunk: Enabled

Problem Description
We observe call disconnects when executing a Data Action block at the start of the flow. The external CRM lookup via HTTP Request takes longer than expected during peak traffic hours.

Configuration

Error Log Entry
Timestamp: 2024-05-15 09:15 UTC
Level: ERROR
Message: Request Timed Out - HTTP 504 Gateway Timeout
Flow ID: 00000000-0000-0000-0000-000000000000

Impact
Calls disconnect before reaching Agent skill group. Customer data does not populate in the screen pop.

Question
Does increasing timeout affect concurrent call limits? Need solution for high latency API without global configuration change.

2

Reply by AnalyticsOwl_Lis27

Increasing timeout is viable but impacts concurrency on the platform. Try asynchronous Data Action pattern instead. This allows the flow to continue while backend processes data.

Use MuleSoft Anypoint Platform for better error handling with Kafka topic gc-architect-events. The external system should poll status or receive a webhook callback.

MuleSoft Flow Snippet:

<flow name="GC-DataAction-Lookup">
    <http:listener config-ref="HTTP_Listener_Configuration" path="/gc-dataaction"/>
    <logger message="Received Lookup Request" level="INFO"/>
    <set-payload value="#['Processing initiated']"/>
    <async:executor threadNamePattern="GC-Worker-%d">
        <java:class name="com.company.CRMIntegrationService" method="fetchCustomerData">
            <java:parameter expression="#payload"
            />
        </java:class>
    </async:executor>
</flow>

Refer to API documentation for Data Action timeout limits specifically. Ensure the callback returns within 5 seconds to maintain flow state.

3

Reply by QMAnalyst_Har40

Dispositions do not record correctly if call drops at the IVR stage before transfer. Recent A/B test with AI routing showed latency spikes correlated with dropped calls in QM queue.

ML model inference time added 2 seconds to average handle time during peak periods. Verify disposition codes in Reporting API endpoints for incomplete interactions.

A/B Test Observation:

  • Control Group: Standard timeout (30s)
  • Test Group: Async callback pattern
  • Result: 15% reduction in dropped calls in Test Group
  • Note: Data consistency matters for compliance reporting