Question

You are building an order-processing service on Azure Cosmos DB for NoSQL. A container named 'Orders' is partitioned on '/customerId'. When a customer submits an order, your code must atomically insert one order document and update the customer's running-balance document — both belonging to the same customer — so that either both operations succeed or both fail. Throughput and latency must remain low. Which approach should you use?

Accepted Answer

Execute a TransactionalBatch that includes both operations, scoped to the customer's partition key value.

Answer

Use the bulk executor to send both operations concurrently with EnableBulkExecution set to true.

Answer

Wrap both operations in a client-side try/catch and manually delete the order if the balance update fails.

Answer

Perform the two writes across two separate containers and rely on the multi-region write consistency level to keep them atomic.