Why should you care?

CAP theorem: network partitions happen; design for eventual consistency and idempotency.

See it live — copy this example

Paste into your MeanVerse project (Angular + Express + MongoDB), then run with ng serve / node / mongosh as noted.

// idempotent transfer using client idempotency key
router.post('/transfers', auth, async (req, res, next) => {
  const key = req.header('Idempotency-Key');
  if (!key) return res.status(400).json({ message: 'Idempotency-Key required' });
  const existing = await Transfer.findOne({ idempotencyKey: key, userId: req.user.id });
  if (existing) return res.status(200).json(existing);
  try {
    const transfer = await transferService.create({ ...req.body, idempotencyKey: key });
    res.status(201).json(transfer);
  } catch (e) { next(e); }
});

What happened?

• Client retries same key on timeout.
• Server returns original transfer instead of double debit.
• Essential when load balancer retries POST.

Practice next

1. Assume requests duplicate and order varies.
2. Use idempotency keys on financial writes.
3. build health checks and graceful shutdown.
4. Add saga orchestrator for order + payment + inventory.
5. Test split-brain scenario in staging chaos day.

Remember

Failures are routine at scale. Idempotency and sagas prevent duplicate effects. Observability traces cross service boundaries.

Mobile retry storm

User taps Pay twice; network retries POST three times.

Outcome: Idempotency-Key ensures one transfer; user sees consistent result.