CDK for Terraform (CDKTF) has been deprecated, leaving users to find alternatives. This article outlines options for migration, focusing on Pulumi as a viable path forward, detailing the conversion and import process for transitioning projects.

The author shares a misstep during an AWS migration from IAM Users to SSO, resulting in a permissions error related to a KMS key. After discovering the issue, they detail the recovery process and a new checklist step to prevent similar problems in the future.

The article discusses the migration of over 30 Kubernetes clusters to Terraform, detailing the challenges faced with previous tools like Sceptre and AWS CDK, and outlining a structured, iterative approach to the transition. Key strategies included automating processes, ensuring safety during rollbacks, and emphasizing hands-on knowledge transfer over traditional documentation. The authors share insights on tooling, risk management, and team collaboration throughout the migration journey.

Terraform migrate 1.1 introduces support for version control system (VCS) workspaces, enabling Git-based infrastructure management and automated migration flows. It improves user experience with features like token validation optimization and reliable branch pushing, streamlining the migration process to HCP Terraform and Terraform Enterprise.

Pulumi has introduced direct support for executing Terraform modules, enabling users to migrate from Terraform to Pulumi without the need for immediate conversion. This feature allows teams to preserve existing Terraform modules while leveraging Pulumi's advanced infrastructure as code capabilities, facilitating a smoother, incremental migration process.

Migrating Amazon EKS from version 1.29 to 1.32 involved careful planning and execution, focusing on upgrading Terraform modules, transitioning to Amazon Linux 2023, and implementing a blue/green deployment strategy to minimize downtime. The process highlighted the importance of addressing breaking changes, updating add-ons, and validating instance types to ensure a smooth upgrade across four clusters. Overall, the upgrade was completed in two weeks with minimal disruptions and enhanced security and performance.