Design WhatsApp: Real-Time Chat System Design
Step 6 β Store and Retry for Offline Users
User B is offline. Messages must be safely stored and reliably flushed when they reconnect. This is the 'store and forward' pattern β what gives WhatsApp its ββ delivery receipts.
Request Flow
Cache miss = offline, park in Queue
The Message Service does a Cache lookup. Cache miss = offline. Message goes into the Queue with 'pending' status.
server = redis.get("presence:userB") # β nil (offline)
# Park message:
queue.push({ msgId, to: "userB", status: "pending" })User B reconnects β flush pending
When User B's app opens, the WebSocket server sets presence in Cache, then pulls all pending messages from the Queue and sends them in order.
// On WS reconnect from User B:
redis.set("presence:userB", "server-3")
pending = db.query('SELECT * FROM msgs WHERE to=B AND status=pending ORDER BY ts')
for msg in pending:
ws.send(msg); db.mark_delivered(msg)Why We Need This
This is why WhatsApp shows the clock icon (sent, not delivered). The message is safely persisted β it just hasn't been pushed yet. The double-tick appears only when User B receives it.
Key Insight
Store-and-forward is the offline contract: any message persisted before the sender's ACK MUST eventually be delivered. The Queue makes this promise maintainable under failures.
Step 6 of 7
