What Is New in Kubernetes 1.11
Kubernetes 1.11 delivers significant enhancements in core networking, custom resource handling, and hardware acceleration. This release stabilizes key features that have been in development and introduces new capabilities for running diverse workloads.
| Category | Key Updates |
|---|---|
| Core Networking | IPVS-based load balancing graduates to General Availability (GA) |
| Custom Resources | CustomResourceDefinitions (CRD) moves to GA, replacing the beta API |
| Hardware Acceleration | Device plugins for hardware resources like GPUs and FPGAs reach GA |
| Cluster Lifecycle | kubeadm reference implementation improvements for cluster management |
| Storage | CoreDNS is now the default DNS server; Persistent Local Volumes enter beta |
| Security | PodPriority and PodReadinessGate features are promoted to beta |
How does IPVS improve production load balancing?
IPVS (IP Virtual Server) based kube-proxy has moved to General Availability, making it the default for new clusters in many environments. This is a big deal for production-scale services because IPVS handles higher network throughput with lower latency compared to the older iptables mode.
In practice, IPVS uses more efficient hash tables for load balancing rules, which reduces the CPU overhead on nodes running many services. You'll see better performance during connection storms and when managing thousands of Service endpoints.
What changed with Custom Resource Definitions?
CustomResourceDefinitions (CRD) has officially graduated to GA, replacing the older ThirdPartyResources (TPR) API. The apiextensions.k8s.io/v1 API is now stable, meaning you can build your custom resources on a solid foundation without worrying about breaking changes.
This matters because CRDs are the backbone of the Kubernetes operator pattern. The GA status brings validation and defaulting schemas, making custom resources behave more like native Kubernetes objects. You should migrate from the beta API to the stable version.
Can Kubernetes now better manage my specialized hardware?
Yes, the device plugin framework has reached GA, providing a stable API for hardware vendors to integrate specialized resources like NVIDIA GPUs, FPGAs, and high-performance NICs. This means your machine learning workloads and other hardware-accelerated applications get first-class citizen treatment.
The device plugin model handles the entire lifecycle - discovery, advertising, allocation, and health checks. In practice, you'll see more reliable scheduling and better resource accounting for pods that need these specialized devices.
What storage improvements should I know about?
CoreDNS has replaced kube-dns as the default cluster DNS provider, offering better performance and a more flexible plugin architecture. For storage, Persistent Local Volumes are now in beta, letting you request locally-attached storage directly through the PVC interface.
This is particularly useful for stateful workloads that benefit from local SSD performance. The scheduler understands which nodes have available local storage and can place pods accordingly, making local storage much easier to manage at scale.
How has cluster management evolved with kubeadm?
The kubeadm reference implementation received several improvements for both initial cluster creation and ongoing management. You now have better control over the Kubernetes control plane components and their configuration.
These changes make it easier to build conformant clusters that meet your specific requirements while maintaining the simplicity of the kubeadm workflow. The tooling continues to mature as the community's preferred method for bootstrapping clusters.
FAQ
Should I migrate from iptables to IPVS load balancing?
Yes, if you're running at scale. IPVS offers better performance and scalability for clusters with many services. The transition is seamless for existing services as it's handled transparently by kube-proxy.
What's the immediate impact of CRD going GA?
You should update your CustomResourceDefinition manifests to use the apiextensions.k8s.io/v1 API version instead of the beta version. The new version provides structural schemas and validation features that make your CRDs more robust.
Do I need to do anything about the CoreDNS change?
New clusters will default to CoreDNS automatically. For existing clusters, the migration from kube-dns to CoreDNS is handled during cluster upgrades if you're using cluster management tools that support this transition.
How do device plugins affect my GPU workloads?
The GA device plugin framework means more stable GPU support through the NVIDIA device plugin. You'll continue to request GPUs using the same resource name (nvidia.com/gpu), but with improved reliability and resource management.
What are the benefits of Persistent Local Volumes?
This beta feature lets you use locally attached SSDs through the standard PVC interface instead of manual node assignment. The scheduler understands which nodes have available local storage, making it easier to deploy stateful applications that need fast local storage.