Migration to AWS without rollback

The context

A software platform needed to move its production workload to AWS while protecting continuity for the people and systems that depended on it. A migration reaches its most sensitive point during the cutover phase, when production traffic moves from the previous environment to the new one.

The client asked Das Meta to plan and deliver the transition, then support the platform during the first week of live operation. The customer remains anonymous, but the operational evidence can still show what happened after launch.

What Das Meta delivered

The work covered the AWS foundation and the production platform running on it. That included account and network configuration, IAM, Kubernetes, image management, GitOps, databases and integrations, monitoring, and operational documentation.

The aim was to make the change controlled and observable. Teams need to know which version is deployed, where traffic is going, and where to look if a service behaves differently after cutover. Pinning container images by digest is one example of the operational practices that help make a deployment traceable.

The migration outcome

The production cutover completed without a rollback. During the following seven days, the application gateway recorded 336,763 requests.

The platform reported 10 HTTP 500 responses, equal to 0.002969% of the application requests observed through that gateway, or approximately 0.003% when rounded. The authentication service returned 628,094 successful token responses and reported zero application HTTP 5xx responses in the same period.

Those figures matter because they describe a production system under real use after the move, rather than a migration that only completed in a test environment.

Reading the numbers carefully

The gateway figures measure requests that reached the application gateway. They do not prove every end-to-end transaction succeeded, and they do not measure uptime, latency percentiles, downtime, customer experience, cost savings, or an SLA. They also do not include failures that happened before a request reached the gateway, such as DNS, load balancer, network, or client-side failures.

The authentication figure records successful token responses, which is different from a complete authentication request count. It confirms that the service handled a substantial number of successful responses without reported application HTTP 5xx errors, but it does not provide a full authentication error rate.

What a controlled cutover requires

Migration success relies on more than moving workloads into a new account. A useful plan makes ownership, dependencies, rollback conditions, monitoring, and the first days of support explicit before production traffic moves.

That preparation gives a team a practical way to make decisions under pressure. When the platform goes live, the people responsible can see whether traffic is arriving, whether services are responding, and whether the path back is still available if it is needed.

Das Meta helps software teams build those foundations, from cloud architecture and migration delivery to the operating practices that keep the platform understandable after launch.