DP-420 Sample Questions & Answers

The properties.locations property in an ARM template for a Cosmos DB account is an array of location objects. Each object specifies a locationName and a failoverPriority. The region with failoverPriority: 0 is the primary write region, and subsequent regions are assigned priorities of 1, 2, and so on, for the automatic failover order.

Transactional batch operations in Azure Cosmos DB for NoSQL are scoped to a single logical partition. This is because they are executed as a single transaction on the database engine replica set that hosts the logical partition. This constraint ensures ACID properties for the operations within the batch. If you attempt to include operations for documents in different logical partitions, the entire batch operation will fail.

This scenario describes a classic 'hot partition' problem. A single physical partition is receiving a disproportionate amount of requests, causing it to hit the maximum throughput limit for a single partition (10,000 RU/s), while the overall container throughput is underutilized. The fundamental solution is to change the partition key to one that distributes requests more evenly across all logical and physical partitions, such as using a synthetic key or a key with higher cardinality.

The requirement for granular, second-level point-in-time restore (PITR) can only be met by continuous backup mode. Periodic backup creates discrete snapshots at set intervals (e.g., every 4 hours). Continuous backup, on the other hand, constantly records changes, allowing for restoration to any point in time within the retention period (up to 30 days). The first step is to migrate the account's backup policy, which can be done via the Azure Portal, CLI, or ARM templates.

The 'Cosmos DB Operator' role is specifically designed for this scenario. It grants permissions to manage Azure Cosmos DB resources (accounts, databases, containers) but does not grant access to the data itself or the ability to use the Data Explorer. This effectively separates control plane management from data plane access, adhering to the principle of least privilege.

For a query to be a single-partition query, the WHERE clause must include a filter that specifies the exact value of the partition key. In this case, the partition key is deviceId combined with the month and year. The query only filters on deviceId and timestamp. Since the full partition key value (e.g., 'DEVICE123-2024-08') cannot be derived from the query filters, Cosmos DB must execute it as a cross-partition query, scanning all partitions that could potentially contain data for 'DEVICE123'.

Creating a new CosmosClient for each request can lead to creating new TCP connections, which can exhaust the available sockets on the client machine, leading to connectivity errors.

The CosmosClient instance performs initial metadata requests to get the routing table for partitions. Reusing the client allows this information to be cached, significantly reducing the latency of subsequent requests.

The CosmosClient is responsible for managing the underlying connections. A singleton pattern ensures this management is efficient and stable over the lifetime of the application, which is particularly important for the more complex connection pooling used in Direct mode.

The SDK's bulk support is highly dependent on the client's resources. It creates many concurrent tasks to saturate the provisioned throughput. If the client machine (e.g., a VM running the script) has limited CPU, memory, or network bandwidth, it can become the bottleneck, unable to generate requests fast enough to consume the available RUs on the server side. Since no throttling is occurring, the bottleneck is on the client side, not the server side.

Within a JavaScript stored procedure, the createDocument function is used to create a new document. It takes the collection's self-link, the document body, and an optional callback function as parameters. All operations performed within a single stored procedure execution are transactional, provided they target the same logical partition.

Azure Synapse Link uses an auto-sync process that propagates data from the transactional store to the analytical store. This process is designed to have minimal impact on the transactional workload and does not consume any of the provisioned RUs. The typical latency for data to become available in the analytical store is about 2 minutes.