Why should you care?

DataFlow at national scale cannot stay one mysqld — distributed design combines shards, replicas, cache, and queue for cohesive platform.

See it live — copy this example

Run in MySQL Workbench or the mysql CLI.

-- Conceptual health check across layers:
SELECT 'primary' AS role, @@hostname AS host, COUNT(*) AS orders FROM orders
UNION ALL
SELECT 'replica', @@hostname, COUNT(*) FROM orders; -- run on replica connection
-- App router: tenant_id -> shard_id -> host

What happened?

• Primary and replica should eventually match row counts (lag aside).
• Router directs tenant to shard.
• Distributed system = MySQL + replication + sharding + observability together.

Practice next

1. Draw DataFlow target architecture diagram.
2. List single-node limits you hit first (disk, QPS, HA).
3. Map each component: primary, replica, cache, router.
4. Compare InnoDB Cluster vs manual primary/replica ops.
5. Document saga flow for checkout across payment + inventory services.

Remember

Distribution = replicas + shards + failover. MySQL is one layer in bigger system. Design for ops complexity upfront.

DataFlow scale blueprint

Architecture review approves 4 shards + 2 replicas each + Redis cache for national launch.

Outcome: Team knows path from tutorial DB to production topology.