Docs

Volcengine Kubernetes Engine

Kubernetes version release notes

(EOM) Release notes of VKE based on Kubernetes v1.20

Volcengine Kubernetes Engine (VKE) strictly adheres to the Certified Kubernetes Conformance Program. This topic describes the changes introduced in the VKE release based on Kubernetes v1.20.

warning

VKE will stop providing technical support for Kubernetes v1.20 starting from June 30, 2024 (UTC+8). For details, see Cluster version announcement.

Release notes

Table 1: Core components and descriptions

Core component	Current version	Considerations
Kubernetes	v1.20.15	None
Containerd Runtime	1.6.21	None
ETCD	3.4.13	None
CoreDNS	1.7.0	None

Table 2: Kubernetes versions and the corresponding VKE versions

Kubernetes version	VKE version	Release notes
v1.20.15	vke.37	Fixed an issue where the pvcprotection controller in kube-controller-manager caused PVCs to remain in the Terminating state for an extended period during deletion and fail to be released.
v1.20.15	vke.36	The expansion scheduler is enhanced, and the mGPU cross-GPU sharing feature supports the full computing power allocation strategy.
v1.20.15	vke.35	Fixed the issue that occurs during initialization of the extended scheduler add-on.
v1.20.15	vke.34	Optimized the release specifications for core components.
v1.20.15	vke.33	The issue where a kubelet restart causes the container Ready status to change is fixed.
v1.20.15	vke.32	The extended scheduler is enhanced to support mGPU preemptive scheduling.
v1.20.15	vke.31	The extended scheduler is enhanced to support VCI-aware scheduling.
v1.20.15	vke.30	The self-signed certificate of the API server is optimized.
v1.20.15	vke.29	Kubelet is enhanced, which allows you to configure container resource views. Kubelet is enhanced, which allows you to configure the graceful shutdown period for evicted pods by using annotations. The load-aware scheduling issue of the expansion scheduler is fixed.
v1.20.15	vke.28	The runc container escape vulnerability CVE-2024-21626 is fixed.
v1.20.15	vke.27	Kubelet is enhanced, which performs device allocation based on the scheduler allocation results.
v1.20.15	vke.25	Enhanced the elastic resource priority scheduling feature to support proportional scheduling and to trigger scale-out by Cluster Autoscaler.
v1.20.15	vke.24	The API server parameter configuration is enhanced to ensure consistency with the community version.
v1.20.15	vke.23	Enhanced the API Server and optimized the way parameters are configured.
v1.20.15	vke.22	The API server is strengthened and the vulnerabilities CVE-2023-44487 and CVE-2023-39325 are fixed.
v1.20.15	vke.21	Enhanced the extended scheduler to avoid cache penetration to etcd.
v1.20.15	vke.20	The expansion scheduler is enhanced, with the load-aware scheduling feature improved.
v1.20.15	vke.19	The issue with synchronization of the actual load when the expansion scheduler reschedules pods is fixed.
v1.20.15	vke.18	The expansion scheduler is enhanced, which achieves load-aware hotspot descheduling. The expansion scheduler is enhanced, which can inject node information into pods during scheduling.
v1.20.15	vke.17	Kubelet is enhanced, with the initialization logic of the node network add-on optimized.
v1.20.15	vke.16	The Kubernetes cluster installation process is optimized.
v1.20.15	vke.15	The issue that occurred during the expansion scheduler initialization is fixed.
v1.20.15	vke.14	The issue in kubelet caused by node resource contention is fixed. The issue where kubelet fails to update the NetworkUnavailable condition when the node network is not ready is fixed.
v1.20.15	vke.13	kube-controller-manager is enhanced to support the non-graceful node shutdown feature. The kube-controller-manager force-detach feature is enhanced. The expansion scheduler is enhanced, which supports the mGPU cross-GPU sharing feature.
v1.20.15	vke.12	kubelet is enhanced, and reserved PIDs are optimized.
v1.20.15	vke.11	The API server is enhanced to prevent cache penetration into ETCD.
v1.20.15	vke.10	The expansion scheduler is enhanced, which supports load-aware scheduling.
v1.20.15	vke.9	The fine-grained request observability capability of the API server is enhanced. The configuration of scoring plugins for the extended scheduler is enhanced.
v1.20.15	vke.8	Capacity scheduling is supported.
v1.20.15	vke.7	The performance of the extended scheduler in scenarios with a large number of pending Pods is enhanced.
v1.20.15	vke.6	The extended scheduler is enhanced to support mGPU computing power allocation policies.
v1.20.15	vke.5	kubelet is enhanced to support scenarios where mGPUs on nodes have large amounts of video memory.
v1.20.15	vke.4	The extended scheduler is enhanced to support gang, binpack, and mGPU sharing scheduling policies. kubelet is enhanced to support bind mounts for its directory. NodeLifeCycleController is enhanced so that Pods are no longer evicted when a virtual node becomes unreachable.
v1.20.15	vke.3	The API server proxy allowlist feature is enhanced.
v1.20.15	vke.2	The API server is enhanced to prohibit proxying to 0.0.0.0.
v1.20.12	vke.1	The initial Kubernetes v1.20 version provided for VKE. In IPVS mode, conn_reuse_mode is not set to 0 by default.

About Kubernetes community versions

Dockershim is being deprecated.

Dockershim is officially deprecated. Docker is no longer supported, and the community will remove Dockershim in subsequent versions. This change is unrelated to the images used by clusters. Docker-generated images that comply with the OCI specification can continue to run in CRI-compatible runtimes. For more information, see Dockershim Deprecation FAQ.

Exec probe timeout handling.

The issue with Exec probe timeouts has been fixed. This issue could affect existing Pod definitions. Previously, Exec probes did not consider the timeoutSeconds field. With this fix, if no value is specified, the default value of 1 second is used. If the probe time exceeds 1 second, the existing Pod definitions must be updated to specify the timeoutSeconds field. The fix provided by the newly introduced ExecProbeTimeout feature gate can restore cluster operations to the previous behavior, but this behavior will be locked and removed in subsequent versions. To restore the previous behavior, set ExecProbeTimeout to false. For more information, see Configure Liveness, Readiness and Startup Probes - Configure Probes.

API Priority and Fairness (APF) graduates to Beta.

API Priority and Fairness (APF) is enabled by default in v1.20, allowing the kube-apiserver to classify incoming requests by priority. For more information, see API Priority and Fairness.

The kubectl debug feature graduates to Beta.

The kubectl alpha debug feature has reached Beta and is now kubectl debug. This feature provides support for common debugging workflows directly from kubectl. kubectl alpha debug is deprecated and will be removed in subsequent versions. For more information, see Debug Running Pods.

The volume snapshot operation feature graduates to GA.

This feature provides a standard way to trigger volume snapshot operations and enables users to incorporate these operations into any Kubernetes environment and supported storage provider in a portable manner. In addition, these Kubernetes snapshot primitives serve as fundamental building blocks that enable the development of advanced enterprise-grade storage management capabilities for Kubernetes.

The PID limits feature graduates to GA.

SupportNodePidsLimit (node-to-pod PID isolation) and SupportPodPidsLimit (ability to limit PIDs per pod) have both reached GA.

Some API versions are changed.

The extensions/v1beta1 and networking.k8s.io/v1beta1 API versions of Ingress and IngressClass resources are deprecated and will be removed in v1.22. Use networking.k8s.io/v1 instead.

Alpha feature update: IPv4/IPv6.

IPv4/IPv6 has been reimplemented to support dual-stack Services. You can assign both IPv4 and IPv6 cluster IP addresses to a single Service and convert Services between single-stack and dual-stack configurations.

Alpha feature: Graceful node shutdown.

Users and cluster administrators expect Pods to follow the expected Pod lifecycle, including Pod termination. GracefulNodeShutdown enables kubelet to listen for node system shutdown events, allowing Pods to be gracefully terminated during system shutdown.

Reference links

For more information about the performance and feature evolution of Kubernetes v1.20, see CHANGELOG-1.20.md.

Last updated: 2026.09.17 10:39:30