What’s the best way to handle concurrent screen recording uploads when a Premium App needs to aggregate media from agents across multiple secondary organizations?
We are developing a compliance module that requires immediate access to screen recording metadata and blobs. The current workflow involves the app listening to the screen-recording:completed event, extracting the recordingId, and then attempting to fetch the download URL via the GET /api/v2/analytics/screenrecordings/{screenRecordingId} endpoint. While this works reliably for single-tenant deployments, we are encountering intermittent 500 Internal Server Errors when scaling this pattern across a multi-org structure with high agent concurrency.
The issue appears to be tied to the asynchronous nature of the blob storage availability versus the event trigger. Sometimes the event fires before the recording blob is fully indexed in the secondary org’s storage bucket, leading to a race condition. We have implemented an exponential backoff strategy in our Node.js 18.x backend, but the 500 errors persist at a rate of approximately 15% during peak hours, which is unacceptable for a production compliance tool.
Environment details:
- Genesys Cloud API v2
- AppFoundry Premium App (Node.js 18.x)
- Multi-org tenant structure (Primary + 3 Secondary Orgs)
- SDK: @genesyscloud/genesyscloud-node-sdk 2.4.1
Key observations:
- The
screenRecordingIdis valid and returned correctly by the event payload. - Retrying the GET request after 5 seconds usually succeeds, suggesting a temporary storage latency issue.
- The 500 error response body is empty, providing no specific error code or reason.
Is there a recommended pattern for verifying blob availability before attempting to fetch the download URL? Or is there a specific header or query parameter that can be used to wait for the resource to be ready? We want to avoid polling indefinitely but need a robust method to ensure data integrity across all connected organizations. Any insights into how the platform handles the synchronization between the event bus and the storage layer would be greatly appreciated.