Start here
Overview
Event-driven design decouples producers and consumers—critical when CRM, billing, and messaging evolve independently.
Core concept
Definition
Events are facts: LeadCreated, InvoicePaid, TicketEscalated. Automation reacts asynchronously with retries and ordering where needed.
Business impact
Why it matters
Synchronous chains fail together; events isolate blast radius and enable replay for recovery.
Practical model
Framework
Idempotent consumers
Same event delivered twice should not double-effect.
Dead-letter queues
Poison messages go somewhere humans can inspect.
Implementation detail
Detailed breakdown
Ordering
When order matters, partition by business key—not global ordering.
In practice
Real-world example
An e-commerce operator replayed a day of order events after a partial outage—reconciling CRM and warehouse without manual CSVs.
Avoid these
Common mistakes
- Chatty synchronous calls from UI to five APIs—fragile and slow.
- No schema versioning on events—consumers break silently.
Engineering layer
Technical patterns
Outbox pattern
- Business transaction writes domain row + outbox row atomically.
- Relay publishes to broker; marks outbox `sent`—never lose events on crash.
Consumer idempotency
- Natural key `event_id` dedupe table per consumer group.
- Retries safe: same outcome on redelivery.
Build patterns
Code examples
Idempotent consumer
Skips duplicate deliveries from Kafka/SQS.
export async function handleMessage(msg) {
const ok = await db.insertIgnore('processed_events', {
id: msg.id,
consumer: 'workflow-worker',
});
if (!ok) return; // duplicate
await applyDomainLogic(msg.payload);
}Outbox relay loop
Polls unsent rows and publishes.
export async function relayOutbox() {
const batch = await db.query('SELECT * FROM outbox WHERE sent_at IS NULL LIMIT 100');
for (const row of batch) {
await broker.publish(row.topic, row.payload);
await db.update('outbox', { id: row.id }, { sent_at: new Date() });
}
}System view
System architecture
[Producer service: DB txn]
→ [Outbox table]
→ [Relay → message broker]
→ [Consumers: idempotent handlers]
→ [Workflow / projections / CRM]Keep learning
Related topics
Next step
PrimeAxiom builds event-driven automation back ends—book an architecture whiteboard session.