릴리스 노트
제품 릴리스 기록
Update | Description | Documentation |
Update of the accidental deletion prevention policies for clusters | Default deny policies are enabled for newly created clusters, targeting "Namespaces with existing Pods cannot be deleted", "CRDs with existing CRs cannot be deleted", "Unblocked Nodes cannot be deleted", "CoreDNS component deletion protection", and "PV in binding state cannot be deleted". | |
Support for fault experiments by the cluster control plane | Overload and service outage experiments are supported for control plane components (such as etcd, kube-apiserver, and CoreDNS). By conducting periodic experiments and assessments, the capability to respond to control plane failures can be enhanced for the business. | |
Support for custom resource priority for scheduling | PlacementPolicy is a smart scheduling policy provided by TKE to help users manage Pod scheduling behavior more flexibly in a Kubernetes cluster. With PlacementPolicy, users can customize the resource priority according to business needs, preferentially use high-priority nodes during scale-out, and preferentially release low-priority nodes during scale-in. This balances cost reduction and efficiency improvement with business stability. | - |
New models for TKE native nodes | TKE native nodes support BF1 and M8 models. |
Update | Description | Documentation |
Support for the custom model scale-out policy by native nodes | The model priority policy is added for native node pools, which supports adjusting the primary and alternative models. If auto-scaling is enabled, the scale-out policy preferentially guarantees the purchase of primary model resources. Alternative models can be selected sequentially only when primary model resources are sold out. The scale-out policy can also be used together with AZ settings to implement dual-dimension scheduling. This effectively solves the problem that models for scale-out do not match expectations due to insufficient resources. | |
Basic monitoring capability improvement for native nodes | Native nodes support monitoring instance-related metrics, including CPU, memory, network, disk, storage, load balancing, and GPU. | - |
Collection configuration improvement for log shipping to Kafka | The original log format and various compression formats are supported if the log consumer is Kafka. | - |
Update | Description | Documentation |
Cluster upgrade capability iteration | TKE supports upgrading the Kubernetes version from 1.28 to 1.30. | |
New models for TKE native nodes | TKE native nodes support BMIA2m, PNV6, and HCCPNV6 models. | |
Network monitoring support for workloads | Workloads support monitoring inbound and outbound network traffic and bandwidth. |
Update | Description | Reference |
Cluster storage supports CBS encryption | CBS encryption can protect the data to the maximum extent, requiring no additional modifications on business and user applications. It supports the first-generation and second-generation encryption. | |
Native nodes support one-click login via OrcaTerm | The one-click login to native nodes via OrcaTerm feature supports SSH and TAT. Users can log in to nodes directly through the TKE console without manually configuring a public IP address or SSH key. This feature optimizes Ops experience, resolves the issue of deep login entry and complex operation procedures for native nodes, and offers a more convenient access method for high-frequency Ops scenarios such as data analysis and AI. | - |
Super nodes support multi-path collection | Super nodes can be configured with multi-path for Log collection. |
Update | Description | Reference |
Cluster storage supports GooseFS | By adding Data Accelerator Goose FileSystem (GooseFS) as cluster storage, users can improve the storage access performance for services such as massive data analysis, machine learning, and AI. | - |
Cluster control plane component monitoring | TKE provides monitoring capability for cluster control plane components, allowing cluster administrators to view Kubernetes cluster control plane performance, quickly detect, troubleshoot, and fix issues. |
피드백