Docs

Volcengine Kubernetes Engine

History feature release notes

History feature release notes (2025)

Volcengine Kubernetes Engine (VKE) feature updates released in 2025.

warning

For the latest release notes for VKE, refer to release notes.

Notes

Release region refers to the region where the feature was first released. The release regions coming up afterward will not affect the historical release notes. For the regions where features are actually available, please refer to the information on the console pages.

Historical beta records are not updated after a feature becomes generally available. For subsequent iterations of a feature, refer to the latest relevant release entry.

December 2025

Feature name	Description	Region	Related documentation
Pod networking	When you bind an independent EIP to a pod, you can use an IP address pool to allocate an EIP to the pod, which improves application deployment efficiency.	All	Binding dedicated EIP to pods
Ingresses	The information about Ingresses displayed in the console is optimized, which allows you to view Ingress forwarding paths.	All	Creating an ALB Ingress through the console
mGPU multi-GPU sharing	The mGPU multi-card sharing feature is optimized, which allows you to specify the number of cards for a container by using the container name.	All	mGPU multi-GPU sharing
New version of the RBAC authorization system	More permission management capabilities are added and optimized based on the original RBAC authorization system, which improves the flexibility and efficiency of permission management.	All	Using the new RBAC authorization system of VKE Granting RBAC permissions
Checking and self-healing	Custom node check items can be specified. This allows you to specify node failures based on node NodeCondition information and achieves automatic check and self-healing.	All	Configuring node check and self-healing Custom node check items
ALB Ingress	ALB Ingresses can be connected to Standard edition ALB instances to meet the load balancing requirements of clusters of different sizes. In addition, you can configure custom forwarding rules by using Standard edition ALB instances to achieve features, such as rewrite, redirect, traffic speed limit, and fixed response.	All	Overview of ALB instance editions Creating an ALB Ingress through the console Creating an ALB Ingress through kubectl Custom ALB Ingress forwarding rules
Subscription node pools supporting automatic binding of EIPs to nodes	EIPs can be automatically bound to nodes during the scale-out of subscription node pools, which reduces constraints in scenarios such as a pay-as-you-go node pool being switched to a subscription node pool.	All	Managing node pools
Load-aware scheduling takes effect at the cluster level	Load-aware scheduling can apply to all pods in a cluster. This reduces the need to modify pod configurations and improves your user experience.	All	Load-aware scheduling Installing add-ons
LoadBalancer service integrating with NLB	When you configure a LoadBalancer service, you can use an annotation to integrate with NLB. This achieves higher-performance elastic forwarding.	All	Configuring an NLB service using annotations

November 2025

Feature name	Description	Region	Related documentation
RBAC supporting account cluster-level authorization	[Beta · Apply for Trial] The RBAC authorization system supports cluster account-level authorization, which improves the flexibility and efficiency of permission management.	All	Granting RBAC permissions Using the new RBAC authorization system of VKE
Custom metric collection	The custom metric collection feature is added, which allows clusters to expose metrics of control plane add-ons. This helps you use self-built Prometheus to collect metrics and build a cluster control plane monitoring dashboard, which integrates the metrics with your self-built monitoring system.	All	Operation and maintenance configuration Using self-built Prometheus to collect control plane add-on metrics
Full release of Kubernetes v1.32	Kubernetes v1.32 released by VKE has ended beta and is officially released, which keeps pace with the native community.	All	Release notes of VKE based on Kubernetes v1.32
Cluster public access	You can disable the cluster public access feature and simultaneously clean up the managed NAT gateways. After you disable this feature, you can use a custom public NAT gateway to achieve cluster public access.	All	Cluster basics How do I manually enable Public NAT for Internet endpoint for a cluster?
Updates of node pool billing types	The billing type of nodes within existing node pools can be updated, which improves your user experience and meets various billing change requirements.	All	Managing node pools
CLB and ALB Ingresses supporting cross-project certificate selection	For CLB and ALB Ingresses, when you configure HTTPS certificates, you can select certificates from the Certificate Center of another project.	All	Creating an ALB Ingress through the console Creating CLB Ingresses through the console
Upgrading the node Kubernetes version at the node pool level	[Beta · Apply for Trial] When you upgrade the Kubernetes version of a cluster, you can upgrade nodes either at the node level or the node pool level. This improves the efficiency of node upgrades in large-scale clusters.	All	Upgrading clusters Best practices of cluster upgrade
EFS storage type	Container cluster storage adapts to EFS, which allows you to use static EFS PVs by using FSX or NFS. This meets your requirements for more storage types.	China (Beijing) China (Shanghai)	Accessing static EFS PV through the FSX client Using static EFS PV through the protocol service NFS
FSX authentication	In VKE clusters, when you use FSX to integrate with TOS, you can achieve authentication through IRSA. This achieves pod-level isolation in terms of access to OpenAPIs, which achieves fine-grained isolation of access permissions on cloud resources and reduces security risks.	All	FSX authentication methods Using IRSA to implement authentication for storage mounting Accessing TOS static PersistentVolume through the FSX client
Cloud-native basic observability	In the Observability Dashboard of the Observability Platform: - Period-over-period comparison and statement copying features are added. - These features help you understand data definitions in preset dashboards. - You can use these statements to build custom monitoring dashboards in systems like Grafana.	All	Using dashboards
Cloud-native full-stack observability	When you enable cloud-native full-stack observability, collected trace data will be saved in APMPlus.	All	Enabling full-stack observability

October 2025

Feature name	Description	Region	Related documentation
Elastic scheduled scale-out for node pools	The best effort feature for elastic scheduled scale-out of node pools is added. When the number of scheduled instances cannot meet your scheduling requirements, the system attempts to create multiple scheduled orders to reach the required number of scheduled instances based on the scheduled instance type and availability zone specified for the node pool.	All	Elastic scheduled scale-out Elastic scheduled scale-out (time-period type)
Official commercialization of the pod-identity-webhook add-on	The pod-identity-webhook add-on is officially available for commercial use, which can automatically inject OIDC-related environment variables into pods that use the IRSA feature.	All	pod-identity-webhook Using IRSA in clusters Add-on overview
ALB Ingress	The HTTPS mutual authentication feature is added, which allows you to enable ALB listener mutual authentication during the configuration of the ALB Ingress. This ensures server and client security in HTTPS scenarios.	All	Configuring mutual authentication on HTTPS listeners for ALB Ingresses
Application performance observability	When you configure workloads in a cluster, APMPlus can be integrated to provide end-to-end performance monitoring services for applications.	All	Application performance observability
Instant snapshot availability of EBS	The instant snapshot availability feature of EBS is supported. After this feature is enabled, you do not need to wait for the snapshot creation to complete before using the snapshot to create a cloud disk.	All	Using EBS snapshots

September 2025

Feature name	Description	Region	Related documentation
Full release of Kubernetes v1.30	Kubernetes v1.30 released by VKE has ended public beta and is officially fully released, which keeps pace with the native community.	All	Release notes of VKE based on Kubernetes v1.30
ALB Ingress	When configuring ALB Ingress through the console or YAML, you can configure WAF protection and access log delivery for ALB instances, as well as access control (ACL) and personalization configuration of listeners.	All	Creating an ALB Ingress through the console Creating an ALB Ingress through kubectl
Access static PersistentVolume through the FSX client	TOS and vePFS storage allow access to static PersistentVolumes through the FSX client. Unlike the method of directly using native storage, the FSX client provides high-performance capabilities with better applicability, providing greater flexibility and improved compatibility.	All	Accessing static TOS PersistentVolume through the FSX client Accessing static vePFS PersistentVolume through the FSX client
Node pool supports RDMA communication mode	[Beta · Apply for Trial] RDMA is now available as a new network communication mode for node pools, allowing nodes to exchange data based on RDMA. This enables faster data transmission and improved communication efficiency, which is applicable to applications that require large-scale data transfer and high-performance network communication.	All	Creating a node pool Creating a managed node pool
Optimizing multi-card sharing configuration	When using mGPU multi-card sharing, the number of GPU cards is configured based on container names instead of containers themselves. This can eliminate the issue that the number of GPU cards needs to be modified frequently because the number of containers in pods keeps changing.	All	mGPU multi-GPU sharing
PersistentVolume backup and restoration	Backup Center now supports PersistentVolumes backup and recovery. CSI snapshots can be backed up to TOS and used for restoration in a specified cluster with one click.	All	Creating backup plans
AI profiling	The AI profiling feature is added for the observability feature of VKE. When you deploy LLMs in a VKE cluster, you can analyze the AI profiling results of specified containers. This facilitates detection and optimization throughout the LLM life cycle, including training, inference, and optimization.	All	AI profiling
Pod event	VKE clusters can be integrated with TLS, which provides event persistence capabilities and supports the generation of charts that record pod events based on log indexes. The charts help you view the events of a specified pod.	All	Configuring and viewing persistent events
RDMA resource allocation mode	[Beta · Apply for Trial] When you use RDMA resources in RoCE high-performance computing clusters, the shared-host mode is added. This mode allows all RDMA NICs on a host to be inserted into containers as IPVLAN sub-interfaces. These NICs are then available for pods within the container cluster.	All	Using RDMA resources (RoCE) in VKE clusters
Registered clusters supporting cloud-native observability	[Beta · Apply for Trial] The high-availability collector deployed in the registered cluster provides basic observability capabilities by using VMP.	All	Observability of registered clusters

August 2025

Feature name	Description	Region	Related documentation
Optimized upgrade checks	Optimized the check process for cluster upgrade, allowing performing checks for both control plane and node upgrade simultaneously. Optimized the upgrade report content, making it easier for users to quickly address anomalies and complete the upgrade process for clusters.	All	Upgrading clusters Cluster upgrade checks Best practices of cluster upgrade
Allows custom containerd configurations	Node pools support custom containerd configuration, which includes the Container Registry proxy address for container images and disabling self-signed certificate verification. This can meet the custom requirements of user business.	All	Configuring containerd for node pools Creating a cluster. Creating a node pool Creating a managed node pool
RBAC allows cluster-level permissions under an account	Grant RBAC permissions at the cluster level under your account to IAM users or roles. These users or roles can access all existing clusters and any future clusters created under your root user.	All	Granting RBAC permissions
Compatible with veLinux 2.0 and Ubuntu 22.04 series images at public beta stage	[Public beta] Volcengine Kubernetes Engine now supports veLinux 2.0 and Ubuntu 22.04 series images. All CPU instance families, including General-purpose, Compute-optimized, and Memory-optimized, can work with these images.	All	Overview of operating system (OS) images
Backup center	[Public beta] The Volcengine Kubernetes Engine (VKE) has newly added a Backup center. It can back up container application metadata to TOS and restore backup data to the specified cluster with one click. This feature meets the needs of scenarios including container business disaster recovery, IDC-hosted cloud applications, and cross-cloud application migration.	China (Beijing) Asia Pacific (Johor)	Backup center overview

July 2025

Feature name	Description	Region	Related documentation
The node pool supports selecting multiple specifications	When creating or updating node pools, this feature allows you to select multiple computing specifications, thus improving the success rate of scaling.	All	Creating a node pool Creating a managed node pool
Instant node elasticity (Karpenter)	This feature supports the instant autoscaler of Karpenter-based nodes, achieving the autoscaler feature of multiple specifications.	All	Cluster Autoscaler overview Configuring the instant elasticity of nodes (Karpenter)
kubelet parameters of nodes	Supports new kubelet parameters for customizing nodes such as max-pods and cpu-manager-policy.	All	Configuring kubelet parameters for nodes
Optimized application template configuration in Application center	In the console, Application center now provides optimized configuration for AI cloud-native inference service templates, simplifying console operations and helping users deploy inference models quickly.	China (Beijing) China (Shanghai)	Deepseek practices Qwen practices Kimi practice
Application center newly adds two cloud-native inference service templates	The Application center in the console newly adds two inference service templates of AI cloud-native: DeepSeek-R1-W4AFP8 and Kimi-K2-Instruct. We have enriched the inference templates to meet the demands of users for multiple inference scenarios.	China (Beijing) China (Shanghai)	Quickly deploy SGLang-based DeepSeek-R1 quantitative version. Quickly deploy SGLang-based Kimi-K2-Instruct
The Kubernetes version 1.30 clusters support NUMA and RDMA	Kubernetes version 1.30 clusters support NUMA-aware scheduling and RDMA topology-aware scheduling. This feature has perfected the capabilities of Kubernetes version 1.30 and enhanced user experience.	All	NUMA-aware scheduling RDMA topology-aware scheduling
The operating system supports the veLinux 2.0 image	[Beta·Apply for Trial] for the clusters based on Kubernetes Version 1.30 and later, the new nodes can work with veLinux 2.0 image and veLinux 2.0 CentOS compatible-version image. The veLinux images of a higher version provide a broader compatibility with the native community.	All	Overview of operating system (OS) images

June 2025

Feature name	Description	Region	Related documentation
Application center now offers five new AI cloud-native inference service templates	In the console, Application Center now provides five new AI cloud-native inference service templates. Backed by ByteDance's large-scale AI practices and ModelArk's extensive production experience, and built on container orchestration, AI gateways, end-to-end inference observability, and distributed caching, these templates help enterprises run mainstream inference models efficiently, stably, and cost-effectively on large-scale GPU clusters.	China (Beijing)	Cloud-native best practices
Packet capture available in full-stack observability	Cloud-native full-stack observability is able to perform network packet capture on the specified workload or pod, facilitating troubleshooting of container networks.	All	Capturing network packets

May 2025

Feature name	Description	Region	Related documentation
Custom node names are allowed in node pools	Custom node names and increment expressions are allowed in node pools. The node metadata can be named with the hostname. Interactive guidance is available on the console page to help configure node names, hostnames, and metadata. It provides a better experience in the configurations of nodes.	All	Creating a cluster. Creating a node pool Creating a managed node pool
Monitoring Center now supports observability for registered clusters	Registered clusters can now integrate with cloud-native observability to provide cluster resource observability, improving the observability capabilities of registered clusters.	All	No
IRSA officially online	IRSA (IAM Roles for Service Accounts), based on Identity and Access Management (IAM), is released in VKE, allowing access control at pod-level in multi-tenant scenarios.	All	IRSA overview
Specified data volumes allowed to be mounted when adding existing nodes to a custom node pool	Verification is no longer required when adding existing nodes to a custom node pool. Data volumes can be mounted by using device names. It facilitates resource migration and allocation with high efficiency by eliminating the concern of data volume size.	All	Managing node pools
Diagnosis of failures when scaling out a node pool is now officially released.	Diagnosis of failures when scaling out a node pool is now available. It provides abundant diagnosis scenarios to help clients increase maintenance efficiency.	All	Scaling out/in node pools
Node resource reservation on the console page	Reservation of resources such as CPUs and memory for nodes can be configured on the console page, increasing resource management efficiency.	All	Configuring kubelet parameters for nodes
User-defined domain names are allowed when issuing KubeConfig	User-defined domain names are allowed when issuing KubeConfig, which facilitates cross-cloud access and provides strong security.	All	Creating a cluster.

April 2025

Feature name	Description	Region	Related documentation
Enhanced node checking and self-healing	The feature of node checking and self-healing can detect typical ECS failures and activate related protocols accordingly (e.g., cold migration), including encapsulating troubleshooting policies based on failure types, and intervening with auto-fix protocol.	All	Node check and self-healing
Ubuntu 22.04 operating system is supported	[ Beta · Apply for Trial ] Ubuntu 22.04 operating system is supported now in clusters based on Kubernetes v1.30 and above, with cgroupv2 to implement isolation of container resources. This feature brings a new version of the operating system to the image family supported by VKE.	All	Overview of operating system (OS) images
EOM of node images with GPU Driver 470 series	Public OS images that include the GPU Driver 470 series are no longer maintained. We recommend upgrading to the latest driver version as soon as possible.	All	[Product changes]

March 2025

Feature name	Description	Region	Related documentation
ALB Ingress allows native certificates in Secret	[ Beta · Apply for Trial ] certificate information can be configured directly in Secret. The ALB Controller in the Certificate Center will create a certificate according to that information and then bind the certificate ID on the ALB side. It simplifies certificate configurations.	All	Enabling HTTPS for ALB Ingress
Multiple listening ports on an ALB Ingress	[ Beta · Apply for Trial ] A single ALB Ingress allows multiple listening ports, which reduces the number of Ingress and management cost.	All	Configuring ALB Ingress using annotation
Optimized layout for container commands on the console	It provides a better layout for typing container commands on the console page for deployment, allowing each command to be put in a separate textbox and ensuring bash commands can be parsed correctly.	All	Workload
Service with support for 7-layer CLB	The service can work with 7-layer CLBs and HTTPS listeners. Additionally, it can automatically create and maintain the back-end server groups to which the listeners are attached. It satisfies the scenarios that need to expose 7-layer services outside of nginx-ingress and configure SSL/TLS certificates for 7-layer web services for secured communication between clients and CLBs.	All	Configuring CLB services by using annotations
Custom ECS hostnames are allowed in node pools.	Custom ECS hostnames are allowed in node pools. It provides a flexible configuration for hostnames.	All	Creating a cluster. Creating a node pool Creating a managed node pool
CR password-free add-on applicable to repositories in multiple regions	The password-free add-on (cr-credential-controller) is applicable to Container Registry (CR) in multiple regions. It meets the needs for cross-region use of the password-free add-on.	All	Using password-free add-on to pull private images to create workloads cr-credential-controller Installing add-on
Tags allowed to be added when creating LoadBalancer services	User tags can be added to the annotations of the LoadBalancer service by CCM (Cloud Controller Manager) of the VKE when creating that service. It expands the capabilities of the LoadBalancer service.	All	Configuring CLB services by using annotations
ACL is available at the access endpoint of APIServer for registered clusters	ACL is available at the access endpoint of APIServer for registered clusters, providing a higher security level for registered cluster access and permission control.	All	Registration of existing clusters through proxy

February 2025

Feature name	Description	Region	Related documentation
CLB Ingress allows certificates in Certificate Center	CLB Ingress can work with the certificates in the Certificate Center with the current Volcano Engine root user. It expands the scope of certificates that CLB Ingress can work with and meets the needs in more scenarios.	All	Certificate management
Creating data volumes using snapshots	The feature of creating data volumes using snapshots now goes online officially. It realizes quick start of nodes from snapshots, providing a higher efficiency.	All	Creating node pools
Batch deletion of Pods	Batch deletion of pods from the pod list on the console page is allowed, increasing resource management efficiency.	All	Managing pods
An allowlist of public IPs is supported for login to pods through terminals.	[Public beta] An allowlist of public IPs is supported for login to pods through terminals, providing more specific security control for terminal login.	All	Access control based on IAM for the VKE console
Optimized interaction when creating clusters	It provides default values, tips and guidance when creating clusters on the console page, delivering a better experience.	All	Creating a cluster.
IPVS is supported for clusters that use the VPC-CNI network model	[ Beta · Apply for Trial ] You can specify a PVS LoadBalancer when creating a cluster with the VPC-CNI network model. In autoscaling scenarios, the least-connections algorithm helps direct new connections to newly scaled-out pods whenever possible.	All	Creating a cluster.
Taint configuration available for node pool autoscaler	Two taint parameters for node pool autoscaler (CA) are available to meet various scheduling requirements in scale-out scenarios: Startup taint: if a node has this taint, the node is considered not yet initialized. Ignore taint: the taint scheduling attribute of a node or node pool will be ignored during scaling out.	All	Creating a cluster autoscaler
TSSD TL0 cloud disk for container storage	TSSD TL0 cloud disk is available for container storage to meet the needs for more cloud storage scenarios.	All	Using static PersistentVolume of cloud disks Using dynamic PersistentVolume of cloud disks
Optimization of the Deployment list page	The deployment list page is optimized in terms of display and the search tool, allowing sorting by creation date and filtering by label. It provides a better experience on the console page.	All	Deployments
Console configuration load-aware scheduling parameter	[ Beta · Apply for Trial ] Load-aware scheduling can be configured on the console page with parameters such as usage threshold and scoring weight. It provides a better experience and higher efficiency during configuration.	All	Load-aware scheduling Scheduler-plugin add-on
Supports Kruise add-on	[ Beta · Apply for Trial ] A Kruise add-on built based on OpenKruise is available in the Add-on Center, allowing automatic deployment, release, and maintenance of native cloud applications. It meets the requirements for enhanced add-on management capabilities for workloads.	All	kruise Add-on overview Installing add-ons
Cluster with shared VPC	[ Beta · Apply for Trial ] Clusters can be created based on shared VPC. It allows a single account to manage network resources and meets the requirements of communication and isolation between business networks.	All	Using a shared VPC to create clusters
Capability of working with KEDA add-on	The KEDA add-on is available in the add-on hub, providing an autoscaler driven by Kubernetes events to enhance the capability of elastic scaling.	All	Add-on overview keda
Online/offline hybrid deployment with mGPU	[ Beta · Apply for Trial ] mGPU allows hybrid deployment of online and offline services based on the same GPU card, increasing utilization of GPU resources.	All	Using mGPU to implement online-offline hybrid deployment
Public beta based on Kubernetes v1.30 version	[Public beta] Volcengine Kubernetes Engine based on Kubernetes v1.30 is available at public beta stage.	All	Release notes of VKE based on Kubernetes v1.30
Deployment set for node pool	ECS instances for nodes in node pools can be deployed using the deployment set. The deployment set is a policy that controls distributed ECS at the hardware layer and provides secure physical isolation, ensuring underlying fault tolerance and high availability.	All	Creating a cluster. Creating a node pool Creating a managed node pool

Last updated: 2026.08.20 17:07:47