# Change Data Capture

## Purpose And Business Outcome

Expose transactional change events so consumers can react to committed inserts, updates, and deletes.

## Scope And Non-Goals

Scope:
- Collection-level change stream subscriptions
- Event publication after successful commit

Non-goals:
- Cross-process event transport guarantees
- External message broker delivery semantics

## User And System Workflows

1. Consumer subscribes to a collection change stream.
2. Application performs data mutations.
3. On commit, CDC publishes events to active subscribers.
4. Subscriber handlers process entity and metadata payloads.

## Interfaces And APIs

- `Watch(...)` collection API
- `ChangeStreamObservable`
- `ChangeStreamDispatcher`
- `ChangeStreamEvent`

## Permissions And Data Handling

- CDC payloads can include document identifiers and entity data.
- Restrict subscription access to trusted application components.

## Dependencies And Failure Modes

Dependencies:
- Transaction commit lifecycle
- Subscriber channel health

Failure modes:
- Subscriber backpressure or handler exceptions
- Event handling assumptions that conflict with rollback behavior

## Monitoring, Alerts, And Troubleshooting

- Review CDC behavior in integration and scalability tests.
- Follow [`../runbook.md`](../runbook.md) for incident response.
- Follow [`../security.md`](../security.md) for event payload handling controls.
- Use [`../troubleshooting.md`](../troubleshooting.md) for diagnosis guidance.

## Rollout And Change Considerations

- Behavioral CDC changes require explicit release-note callouts.
- Maintain compatibility expectations for event payload shape.

## Validation Guidance

- Run `CdcTests` and `CdcScalabilityTests` before release.
- Validate commit-only emission behavior in regression tests.