Deploying Cloud-Native Apps with Azure Container Apps

Cloud-native applications can be deployed using Azure Container Apps, a managed solution that sits between Kubernetes and traditional container orchestration. Azure Container Apps provides Kubernetes-shaped scaling behavior without the need to operate a control plane, offering a granular build process with security features, autoscaling rules, continuous deployment via Azure Pipelines, and revision management.

Three key concepts in this build are the Container Apps Environment, which defines the shared networking, logging, and (optionally) Dapr configuration for a group of container apps; the Revision feature, which creates an immutable copy of every deployment; and the Scale rule, built on KEDA, which determines the number of replicas based on real demand signals such as concurrent requests or queue depth.

To connect to Azure Container Registry securely, Container Apps uses a managed identity instead of a password. The registry connection is set up by assigning the system-assigned identity the appropriate role and scope, with the registry server and identity information securely stored and referenced in the configuration. This approach avoids the need for long-lived credentials and eliminates the potential for credential rotation or leakage.

The Container App is created by defining its environment, image, ingress settings, container resource requirements, scaling options, and other configurations. It is then exposed to the internet through a built-in load balancer with TLS termination, or can be used as a backend service for other apps within the same environment.

Autoscaling in Azure Container Apps is achieved through scale rules defined with KEDA. HTTP-triggered autoscaling adjusts the number of replicas based on concurrent requests per replica, while queue-triggered autoscaling scales background worker applications based on the number of messages waiting in a queue. This allows for efficient resource utilization, with worker applications scaling down to zero when idle and scaling back up when new messages arrive.

Continuous deployment is supported through Azure Pipelines, using a YAML configuration file to define the build and deployment stages. The pipeline triggers on code pushed to the main branch, builds and pushes the container image to Azure Container Registry, and then updates the Container App with the new revision. This ensures a seamless and automated deployment process, allowing for canary rollouts or instant rollbacks without manual intervention.