Ensure your application meets these integration specifications:

• Your application must support secure (HTTPS) connections to the endpoint.

• Events are encoded as JSON and separated by a newline character. For sample JSON format events, see Sample Events JSON format.

• For on-premise applications, you can request a replica of the Firehose stream for standby or secondary instances by using the replicaId query parameter (integer; default 1).

You can request to replay events from a specific timestamp by using the from Timestamp parameter:

• Provide the timestamp as the number of milliseconds since epoch.

• If you do not specify fromTimestamp, the HTTP Pull endpoint sends only events received after the HTTP connection is established.

To manage event de-duplication and ensure data continuity, follow these best practices:

• Use the unique identifier (record UID) present in all events to de-duplicate and avoid processing duplicate events.

• During production deployments, if a restart occurs, use the EventsStreamRequest with the fromTimestamp value in combination with de-duplication methods to prevent missing events or data loss.

These guidelines to scale your application and optimize throughput for the Firehose API:

• To distribute load across multiple receivers in your application, or to increase throughput for high-latency connections, you can open multiple connections to the Firehose API HTTP endpoint.

• Your application can have up to 12 receivers.

• When you use more than one connection, you must specify the partitions each connection receives by setting the minPartition (default one) and maxPartition (default 12) parameters in the GET request.