Configuring event-driven integration on Salesforce spans Platform Event definition, CDC enablement, subscriber registration, and ongoing monitoring. The walkthrough below covers the standard sequence for a typical real-time replication use case where Salesforce data needs to flow to an external system in near real time.
- Define the event type or enable CDC
Decide whether the use case is better served by a Platform Event (custom event shape, fired by Apex or Flow) or by Change Data Capture (automatic events on standard or custom object changes). For replication, CDC is usually right; for custom business signals (a new milestone reached), Platform Events are right. Configure the chosen approach: create the Platform Event definition with its fields, or enable CDC on the target object from the Change Data Capture page in Setup.
- Configure subscribers
Identify the subscribers that need to consume the events. For internal subscribers, build the Apex trigger or Flow that processes each event. For external subscribers, configure the external system to subscribe via the streaming API, CometD, or the Pub/Sub gRPC API. Each subscriber needs a clear contract for what it does with each event and how it handles retries on failure. Document the subscribers in the org's integration runbook before going live.
- Validate end-to-end in a sandbox
In a sandbox, fire a sample event (publish a Platform Event record or update a CDC-enabled object). Verify the subscriber receives it, processes it correctly, and any downstream changes happen as expected. Test failure modes: subscriber temporarily offline, subscriber slow to process. Confirm the replay-based catch-up works correctly. Iterate until the integration handles every scenario from the test plan.
- Promote to production and monitor through Event Manager
Deploy the Platform Event definition or CDC configuration to production. Activate the subscribers. Open Event Manager and monitor for the first hour to confirm events are flowing and subscribers are keeping up. Set up alerts (via the platform's notification settings or a custom monitoring tool) for any subscriber that falls behind. Schedule a weekly review of event throughput and subscription lag for the first month, then settle into the steady-state cadence.
- Event retention is 72 hours for most event types. A subscriber that is down for longer loses any events it missed.
- Platform Event publish limits are hourly and daily. Exceeding them causes events to be rejected at publish time without an error to the calling code.
- CDC events represent changes, not full record state. Subscribers that need a full record snapshot must query the record separately after receiving the change event.
- Subscriber lag is silent. Without monitoring, a subscriber that falls behind goes unnoticed until events start dropping past the retention window.
- Sandbox event throughput limits are lower than production. Sandbox tests for high-volume scenarios need careful scaling considerations.