– Karpenter

SKS: Container registry mirroring for Karpenter nodes

Wed, 19 Aug 2026 08:00:00 +0000

Exoscale SKS now supports configuring OCI registry mirrors for nodes provisioned by Karpenter. Registry mirror endpoints and pull credentials can be defined in an ExoscaleNodeClass and applied consistently to new nodes created from that class.

What this enables

Faster and more reliable image pulls: Route image requests through private mirrors located closer to your workloads.
Lower egress costs: Reduce repeated downloads from public registries by caching commonly used container images.
Fewer upstream limits: Avoid relying solely on public registries and reduce the impact of their rate limits or availability issues.
Greater supply-chain control: Use private registries, authenticated mirrors, and trusted TLS configurations for container images.

Why it matters

Registry mirroring gives teams more control over how container images reach their SKS clusters. It improves reliability and supports production environments that require private access, predictable image delivery, or tighter control over external dependencies.

Learn how to configure OCI registry mirroring for Karpenter nodes.

SKS: CPU Manager policies for Karpenter-provisioned nodes

Mon, 17 Aug 2026 08:00:00 +0000

Exoscale SKS now supports configuring Kubernetes CPU Manager policies for nodes provisioned by Karpenter. You can define the kubelet CPU Manager configuration in an ExoscaleNodeClass and apply it consistently to every node created from that class.

What this enables

More predictable performance: Assign exclusive CPUs to eligible workloads that are sensitive to latency or CPU contention.
Consistent node configuration: Apply the selected CPU policy automatically to new nodes provisioned by Karpenter.
Flexible workload isolation: Reserve CPU capacity for Kubernetes and system components while keeping application resources separate.

Why it matters

CPU Manager policies give teams greater control over how processor resources are allocated across their SKS workloads. This is especially useful for high-performance, real-time, and other compute-intensive applications that benefit from stable CPU placement.

Learn how to configure CPU Manager policies with Karpenter.

SKS: Karpenter autoscaling integration is now Generally Available

Thu, 19 Mar 2026 10:00:00 +0000

We’re happy to announce that the Karpenter autoscaling integration for Exoscale Scalable Kubernetes Service (SKS) is now Generally Available.

With this release, Exoscale becomes the first European cloud provider to offer a fully integrated Karpenter experience for managed Kubernetes, enabling modern, workload-aware autoscaling for production environments.

Check it out: https://github.com/kubernetes-sigs/karpenter?tab=readme-ov-file#karpenter-implementations

Why move beyond traditional autoscaling?

Karpenter doesn’t just scale, it optimizes, provisioning the exact compute resources needed in real time.

Simplified autoscaling: No manual node pool configuration.
Continuous auto-healing: Automatic replacement of unhealthy nodes.
Cost efficiency: Reduced overprovisioning and lower cloud costs.

This provides a powerful, sovereign foundation for elastic workloads like AI pipelines, CI/CD, and high-traffic applications, all while keeping your data strictly within European infrastructure.

Ready to streamline your Kubernetes operations?

Check out our documentation to get started with Karpenter on SKS today

SKS: Karpenter Autoscaling Integration (Public Preview)

Tue, 11 Nov 2025 13:03:02 +0000

We’re bringing next-generation Kubernetes autoscaling to Exoscale, now in Public Preview.

Our new Karpenter integration brings advanced, real-time autoscaling to Exoscale Kubernetes clusters. It automatically provisions and removes compute or GPU nodes as demand changes, improving efficiency and reducing idle costs.

Source: Karpenter

The integration extends Exoscale’s existing autoscaling to full node lifecycle management, delivering hyperscaler-level scalability within a sovereign European cloud.

Karpenter automatically selects the optimal instance type for each workload, supports GPU-based AI and ML use cases, and scales clusters down to zero when idle, so users pay only for the resources they use.

Based on the open source Karpenter project, this add-on is included in SKS PRO Cluster at no additional cost.

Note: Karpenter is currently in Public Preview. Some features may change before General Availability, and you may encounter bugs. If you experience any issues, please open a support ticket in the Exoscale Portal.

For more details, check out our blog: https://www.exoscale.com/blog/sks-karpenter/

Learn how to enable Karpenter