Routing Pattern Production Drill
Routing Pattern Production Drill
Routing Pattern Production Drill
Use the exchange topology as an operations decision, not just a publisher setting. A production routing review should answer these checks before the queue names are frozen:
| Check | Evidence to collect | Production risk if skipped |
|---|---|---|
| Topic hierarchy | Example keys such as factory.line1.temperature and binding patterns using * or # |
Consumers receive too much data and reimplement filtering in application code |
| Offline consumer buffering | Durable queue, persistent messages, and a measured backlog drain rate | Maintenance windows silently drop telemetry or create an unbounded queue |
| Failure routing | Dead-letter exchange, rejection policy, retry owner, and poison-message sample | Bad payloads loop forever or block healthy messages |
| Fan-out requirement | Explicit reason for a fanout exchange or separate topic bindings | Audit, analytics, and alert consumers interfere with one another |
For an offline analytics consumer, record the expected backlog in messages and bytes. For example, 10 msg/s over a 30 minute maintenance window creates 18,000 queued messages; at 200 bytes each, the broker needs roughly 3.6 MB of payload storage before metadata and index overhead. That calculation belongs beside the durable queue declaration, not in a post-incident note.
Test example keys such as factory.line1.temperature against bindings using * or #. Record durable queues, persistent messages and a measured backlog drain rate. Record the backlog calculation beside the durable queue declaration.