Kubernetes upgrades don’t have to break things: How EKS is making cluster lifecycle management simpler and safer

Kubernetes frequently releases minor updates, with three new versions rolling out each year. Staying up-to-date is crucial for security, support, and new features, yet previous upgrades often caused irreversible issues. This shaped how teams managed their cluster lifecycles, with extensive preparation, parallel infrastructure, and complex procedures leading to multi-day processes.

Some organizations built blue-green deployment strategies to mitigate upgrade risks, doubling infrastructure and implementing traffic mirroring, while others considered upgrades as a "point of no return." As a result, teams delayed upgrades, using older versions longer than intended until forced by end-of-support deadlines.

Recognizing this problem, the EKS (Elastic Kubernetes Service) team aimed to make staying current with Kubernetes more straightforward. In late 2023, EKS launched Extended Support, which extended the availability of each Kubernetes version to 26 months, accommodating various industries' testing cycles and giving teams more flexibility.

Meanwhile, EKS introduced Upgrade Insights, automated checks that scan clusters for issues and provide actionable remediation steps, reducing preparation time and making upgrade readiness clearer. However, automated checks had limitations, particularly for large-scale production environments where certain incompatibilities only surfaced under heavy API usage patterns.

In July 2026, EKS introduced reversible in-place upgrades, allowing teams to roll back to the previous minor version within 7 days of the upgrade. This 7-day window provided enough time for validation and observation of behaviors under real production traffic while ensuring a reasonable limit on the divergence between the cluster state and the previous version.

The ability to roll back changed how teams approached upgrades, transforming them from high-stakes events to routine lifecycle tasks and reducing the reliance on costly mitigation patterns like parallel clusters and duplicated nodes. Salesforce, a large EKS user, found this change particularly impactful, incorporating rollback into its upgrade workflow to achieve faster rollouts with a built-in safety net.