Question

A developer reports that an application writing to an Azure Cosmos DB for NoSQL container intermittently fails. Inspecting the SDK diagnostics, you notice the failed requests return HTTP status code 503 (Service Unavailable), not 429. The container uses autoscale throughput and the Normalized RU Consumption metric peaks at only 62%. What is the MOST likely cause of these failures, and the appropriate action?

Accepted Answer

Transient connectivity or service-side unavailability is occurring; ensure the SDK's built-in retry policy is enabled and implement retry logic for 503 responses.

Answer

The account is exceeding provisioned RU/s; increase the maximum autoscale throughput to eliminate the errors.

Answer

The partition key is causing a hot partition; redesign the partition key to distribute writes more evenly.

Answer

The requests violate the consistency level; lower the consistency from Strong to Session to allow the writes to succeed.