Deprecating Stopped Clusters: A Better Approach for Reliability and Cost Optimization
For some time, we've supported stopping your Kubernetes clusters by shutting down the control plane. However, after careful consideration, we've decided to deprecate and remove this feature. While the intent was to provide a cost-saving mechanism, we've identified several downsides that make this approach less than ideal.
Why Are We Removing Stopped Clusters?
- Security & Maintenance Risks – A stopped control plane does not receive updates, security patches, or bug fixes. This creates a liability for users who restart their clusters after extended periods.
- Resource Reservation – Even when stopped, a cluster reserves its place in our infrastructure.
- Hidden Costs – Stopped clusters still incur costs, particularly for cloud resources like load balancers, storage, and reserved IPs.
What Should You Do Instead?
If you're looking to optimize costs without completely removing your cluster, consider these alternatives:
- Scale Nodes to Zero – Keep the control plane running but remove all worker nodes. This reduces compute costs while allowing control plane maintenance to continue.
- Delete and Recreate with Automation – If you need a temporary cluster, delete it when not in use and use Terraform or another infrastructure-as-code tool to recreate it when needed.
When Will This Change Take Effect?
We will be rolling this out starting from June 2025. At that time, we will automatically start all currently stopped clusters to ensure they remain updated and secure. If you currently rely on stopped clusters, we recommend transitioning to one of the alternative approaches as soon as possible.
If you have any questions or need guidance on automating your cluster lifecycle, feel free to reach out to our support team!
Thanks for being part of our managed Kubernetes platform. We're always working to improve performance, security, and usability for our users. 🚀