Moving from ECS Fargate to Azure Container Apps. What should I watch for?

Think of Container Apps as a managed place to run containers, rather than a one-to-one mapping of every ECS concept. Start with the container images, listening ports, ingress, environment variables and scaling settings for each service.

Revisions are worth learning early. They give you a way to run different versions of an app and shift traffic between them, which is useful for blue/green releases. Check the revision and traffic settings rather than assuming every deployment automatically gives you a safe zero-downtime rollout.

For multi-region, separate the deployment question from the failover question. You can deploy the app in more than one region, but you also need a way to route users to a healthy region and a plan for data and other dependencies when one region is unavailable.

Decide whether both regions should normally receive traffic or whether one is mainly a standby. Then test the failure and recovery path, including what happens to requests and state during a switch. A blue/green release and a regional failover solve different problems, even if both involve directing traffic.

Watch the scaling behavior under a sudden spike. New replicas aren't instant, so existing ones can get overloaded before the app scales out. A higher minimum replica count can give you more headroom, but it also changes the cost calculation. Load test with the kind of traffic ramp you expect rather than relying on scale settings alone.

Terraform and Azure DevOps are reasonable starting points for provisioning and deployment. You don't have to build everything manually in the portal first. Have a container image available in a registry as part of the deployment flow, and expect private networking to add more moving parts than a public setup.

Also check certificate options against the ingress design. Managed certificates for custom domains have public accessibility requirements, so don't assume the same setup will work unchanged for a private app.

Keep an eye on cost, but don't assume a container platform will automatically be cheaper. Minimum replicas, traffic patterns and the resources allocated to each app all matter. Compare the expected setup against what you're running today, especially if extra replicas are needed for spike handling.