Why should you care?

PostgresVerse group runs orders in Mumbai, inventory in Singapore, payments in PCI zone — distributed not one giant database.

See it live — copy this example

Run in pgAdmin or psql.

-- Orders service (Mumbai PostgresVerse)
CREATE TABLE orders (
  order_id uuid PRIMARY KEY,
  global_id text UNIQUE,  -- idempotency across regions
  status text NOT NULL
);

-- Inventory service (separate cluster) references global_id not FK
CREATE TABLE reservations (
  reservation_id bigserial PRIMARY KEY,
  global_order_id text UNIQUE,
  sku text,
  qty int
);

What happened?

• No cross-cluster FK — global_id correlates sagas.
• UNIQUE on global_order_id makes inventory reservation idempotent when order event retries.
• Each service owns its postgres.

Practice next

1. Create orders table with global_id UUID.
2. Simulate inventory reservation on second database with same global_id.
3. Document saga: reserve → pay → confirm with compensating cancel.
4. Simulate payment failure triggering reservation release event.
5. Map RPO per region postgres cluster on architecture diagram.

Remember

Database per bounded context. Global correlation id not cross-db FK. Eventual consistency with outbox/CDC.

PostgresVerse global commerce

Order in IN region; inventory SG; payment HK PCI cluster; saga ties global_id.

Outcome: Teams deploy independently; blast radius contained per domain database.