gcp Google Cloud Blog ·

GKE agentic migration open-sourced for AI-assisted EKS-to-GKE Kubernetes migrations

blogaiawsgcpdeprecationengineergcp-gkegcp-cloud-storage
feature announcement

Google has open-sourced GKE agentic migration, an AI-assisted tool designed to streamline the migration of Kubernetes environments from AWS EKS to Google Kubernetes Engine. It addresses challenges like manual IaC dissection, cloud-specific differences, and the risks of LLM hallucination in migration efforts. This tool provides deterministic guardrails and GitOps-native workflows, enabling platform and application teams to collaboratively manage complex migrations through an open-source agent plugin.

  • →Introducing GKE agentic migration as an open-source agent plugin
  • →Combats hallucinations and ensures GitOps compliance with hybrid verification
  • →Supports multi-persona state management and contextual runbook generation
  • →Addresses common EKS-to-GKE migration challenges with AI and governance
  • →Orchestrates migration lifecycle with deep EKS discovery and offline validation
Features (3) ›
  • Introducing GKE agentic migration as an open-source agent plugin

    Google has released GKE agentic migration, an open-source agent plugin designed to simplify and secure the process of migrating complex Kubernetes environments from AWS EKS to Google Kubernetes Engine. It replaces ad-hoc LLM prompting with a purpose-built AI-assisted pipeline that includes deterministic guardrails.

  • Combats hallucinations and ensures GitOps compliance with hybrid verification

    It combines LLM-generated infrastructure-as-code (IaC) with deterministic validations and human-in-the-loop approvals to prevent dangerous hallucinations. All changes are managed through Pull Request workflows, ensuring adherence to standard CI/CD and GitOps practices without direct cluster mutations.

  • Supports multi-persona state management and contextual runbook generation

    The plugin persists long-running migration state, enabling asynchronous handoffs between platform and application teams. It intentionally separates architectural translation from stateful data transport, guiding users to purpose-built tools for sensitive data migration via contextual runbooks.

Notes (2) ›
  • Addresses common EKS-to-GKE migration challenges with AI and governance

    The tool tackles issues such as the 'automation trust gap' from generic AI hallucination, the risks of 'ClickOps' bypassing GitOps workflows, siloed handoffs between teams, and fragmented toolchains that lack translation capabilities for cloud-specific configurations.

  • Orchestrates migration lifecycle with deep EKS discovery and offline validation

    The tool utilizes a migration state graph for systematic context passing, performing deep EKS repository discovery and live cluster scans. It generates readiness reports, scaffolds foundational GCP Terraform modules, and handles proprietary shifts like AWS IRSA to Workload Identity mapping, all with offline validation.

Read the original announcement →

https://cloud.google.com/blog/products/containers-kubernetes/gke-agentic-migration/

Related releases