Transactional batch operations in Azure Cosmos DB

APPLIES TO: NoSQL

Transactional batch describes a group of point operations that need to either succeed or fail together with the same partition key in a container. Operations are defined, added to the batch, and the batch is executed. If all operations succeed in the order they're described within the transactional batch operation, the transaction will be committed. However, if any operation fails, the entire transaction is rolled back.

What's a transaction in Azure Cosmos DB

A transaction in a typical database can be defined as a sequence of operations performed as a single logical unit of work. Each transaction provides ACID (Atomicity, Consistency, Isolation, Durability) property guarantees.

  • Atomicity guarantees that all the operations done inside a transaction are treated as a single unit, and either all of them are committed or none of them are.
  • Consistency makes sure that the data is always in a valid state across transactions.
  • Isolation guarantees that no two transactions interfere with each other � many commercial systems provide multiple isolation levels that can be used based on the application needs.
  • Durability ensures that any change that is committed in a database will always be present. Azure Cosmos DB supports full ACID compliant transactions with snapshot isolation for operations within the same logical partition key.

Transactional batch operations and stored procedures

Azure Cosmos DB currently supports stored procedures, which also provide the transactional scope on operations. However, transactional batch operations offer the following benefits:

  • Language option � Transactional batch is supported on the SDK and language you work with already, while stored procedures need to be written in JavaScript.
  • Code versioning � Versioning application code and onboarding it onto your CI/CD pipeline is much more natural than orchestrating the update of a stored procedure and making sure the rollover happens at the right time. It also makes rolling back changes easier.
  • Performance � Reduced latency on equivalent operations by up to 30% when compared to the stored procedure execution.
  • Content serialization � Each operation within a transactional batch can use custom serialization options for its payload.

How to create a transactional batch operation

When creating a transactional batch operation, start with a container instance and call CreateTransactionalBatch:

PartitionKey partitionKey = new PartitionKey("road-bikes");

TransactionalBatch batch = container.CreateTransactionalBatch(partitionKey);

Next, add multiple operations to the batch:

Product bike = new (
    id: "68719520766",
    category: "road-bikes",
    name: "Chropen Road Bike"
);

batch.CreateItem<Product>(bike);

Part part = new (
    id: "68719519885",
    category: "road-bikes",
    name: "Tronosuros Tire",
    productId: bike.id
);

batch.CreateItem<Part>(part);

Finally, call ExecuteAsync on the batch:

using TransactionalBatchResponse response = await batch.ExecuteAsync();

Once the response is received, examine if the response is successful. If the response indicates a success, extract the results:

if (response.IsSuccessStatusCode)
{
    TransactionalBatchOperationResult<Product> productResponse;
    productResponse = response.GetOperationResultAtIndex<Product>(0);
    Product productResult = productResponse.Resource;

    TransactionalBatchOperationResult<Part> partResponse;
    partResponse = response.GetOperationResultAtIndex<Part>(1);
    Part partResult = partResponse.Resource;
}

Important

If there's a failure, the failed operation will have a status code of its corresponding error. All the other operations will have a 424 status code (failed dependency). If the operation fails because it tries to create an item that already exists, a status code of 409 (conflict) is returned. The status code enables one to identify the cause of transaction failure.