Version Overview
Component Change Description
Table 1 openFuyao v26.06 Component Change Description
| Component Name (Helm Chart Package Name) | Component Change Type (New, Enhanced, Fixed, Changed, Deprecated, Removed) | Change Description | SIG |
|---|---|---|---|
| hermes-router | Enhanced | Added routing strategies based on inference latency prediction and backend compute saturation, upgraded core component K8s GIE to the latest version 1.5.0, added tokenizer module, comprehensively enhanced request-level scheduling capabilities. | sig-ai-inference |
| cache-indexer | Enhanced | Rewritten in Go, optimized inference instance local HBM KVCache index maintenance, added memory-level KV Cache index awareness based on mooncake, migrated tokenizer functionality to intelligent routing hermes-router, enabling more accurate global request-level scheduling. | sig-ai-inference |
| pd-orchestrator | Enhanced | Further improved Elastic-scaler framework CRD semantics, implemented MetricsManager, Context Builder and other key components, supported elastic scaling of large-parameter LLM based on LWS, provided APA as default algorithm and supported user-developed custom scaling algorithms. | sig-ai-inference |
| infernex | Enhanced | 1. Deployment mode changed from deployment to the more mainstream and flexible LeaderWorkerSet (LWS) mode, supporting multi-DP architecture deployment and one-click deployment of mainstream MOE large models such as minimax 2.7, deepseek v4. 2. Added infernex-bridge: KServe integration plugin, supporting launching InferNex inference services on NPU environments through KServe. 3. Added infernex-checker: A systematic environment verification tool for InferNex pre-deployment, performing comprehensive checks on hardware, Kubernetes clusters, and business configurations before helm install, effectively identifying potential risks and ensuring deployment success rate. | sig-ai-inference |
| npu-dra-plugin | Enhanced | On top of 910B support, added support for 310P and 910C card models; on top of full-card mounting, added support for hard partition mounting. | sig-orchestration-engine |
| eagle-eye | Enhanced | Added weight distribution and Lingqu network dynamic metric collection: real-time collection of 30+ dynamic performance metrics including link status, actual transmission rate, remaining bandwidth, and packet loss rate for node-side RDMA network cards and card-side RoCE network cards, with periodic collection through Prometheus and near-real-time push via NATS. | sig-ai-inference |
| kubevirt | Enhanced | Added support for kubevirt advanced capabilities in Kunpeng environments, including macvlan, SR-IOV, SR-IOV live migration, network card hot-plug capabilities, and VM live migration. Verified that current kubevirt does not support Kunpeng UEFI secure boot. | sig-orchestration-engine |
| console-website | Enhanced | Added frontend page management of basic VM capabilities based on the openFuyao frontend management plane, reducing O&M costs. | sig-container-platform |
| sandbox-controller | Added | Added openFuyao-sandbox, with sandbox lifecycle management, warm-up pool mechanism, and enhanced checkpoint/restore capabilities. | sig-orchestration-engine |
| sandbox-agent | Added | Added openFuyao-sandbox, with sandbox lifecycle management, warm-up pool mechanism, and enhanced checkpoint/restore capabilities. | sig-orchestration-engine |
| node-agent | Added | Added openFuyao-sandbox, with sandbox lifecycle management, warm-up pool mechanism, and enhanced checkpoint/restore capabilities. | sig-orchestration-engine |
| ubs-k8s-enable | Fixed | Fixed issues caused by configuring kubelet path using native methods and correcting health probe return value exceeding length limits that caused kubelet errors, enhancing component availability. | sig-orchestration-engine |
| compliance-operator | Added | Compliance Operator is a newly added security configuration scanning component for K8s clusters, compatible with configuration scanning based on CIS K8s and STIG K8s baselines, and provides scanning result reports. | sig-installation |
| cluster-api-provider-bke | Enhanced | Added upgrade handling based on declarative components, supporting automatic orchestration, verification, and execution of upgrades according to community release package component lists and dependency relationships. | sig-installation |
| bke-manifests | Enhanced | Added upgrade handling based on declarative components, supporting automatic orchestration, verification, and execution of upgrades according to community release package component lists and dependency relationships. | sig-installation |
| bkeadm | Enhanced | Added upgrade handling based on declarative components, supporting automatic orchestration, verification, and execution of upgrades according to community release package component lists and dependency relationships. | sig-installation |
| release-image | Added | Added upgrade handling based on declarative components, supporting automatic orchestration, verification, and execution of upgrades according to community release package component lists and dependency relationships. | sig-installation |
| upgrade-path | Added | Added upgrade handling based on declarative components, supporting automatic orchestration, verification, and execution of upgrades according to community release package component lists and dependency relationships. | sig-installation |
| many-core-scheduler | Removed | Removed the many-core-scheduler scheduling plugin. | sig-orchestration-engine |
| many-core-orchestrator | Added | Added many-core-orchestrator component, supporting host interference metric collection, node interference analysis, Volcano interference-aware scheduling, and optional Kata Containers VM-level isolation functionality. | sig-orchestration-engine |
| weight-dispatcher | Added | Added weight-dispatcher component, supporting model weight file transmission from storage nodes to compute nodes, and accelerating the weight transmission process. | sig-ai-inference |
Note:
openFuyao-sandbox is in technical preview status for openFuyao v26.06. It will be iteratively updated in subsequent releases to enrich and refine its capabilities. Technical roadmap adjustments or breaking changes may occur, so please refer to the latest release notes for reference.
Interface Change Description
None
Version Feature Description
openFuyao V26.06 main features are shown in Table 2, Table 3 and Table 4. For detailed information on feature characteristics, please refer to User Guide.
| Feature Name | Brief Description (English) | Feature Introduction | SIG |
|---|---|---|---|
| Installation and Deployment | Installation and Deployment | An installation and deployment tool compatible with standard Cluster-API, supporting one-click business cluster installation. The management cluster provides multi-scenario interactive business cluster lifecycle management capabilities on the unified management plane, including single/multi-node installation (with high availability), online/offline installation, cluster scaling, and Kubernetes in-place upgrades. | sig-installation |
| Container Orchestration | Container Orchestration | Provides openFuyao Kubernetes, compatible with K8s 1.34, offering enhanced features such as high-density deployment, startup acceleration, log enhancement, and certificate management enhancement. | sig-orchestration-engine |
| Management Plane | Management Plane | Provides an out-of-the-box console, supporting application management, app marketplace, extension component management, resource management, repository management, monitoring, alerting, user management, and CLI interaction. | sig-container-platform |
Table 3 Independent Components
| Feature Name | Brief Description (English) | Feature Introduction | SIG |
|---|---|---|---|
| infernex | AI Inference Acceleration Suite | InferNex is an end-to-end integrated deployment solution for AI inference in cloud-native scenarios. Through Helm Chart, it seamlessly integrates core components including open-source gateway, intelligent routing, high-performance inference backend, KVCache index management, scaling decision framework, and inference observability system, providing a complete acceleration chain from request admission, dynamic routing, inference execution to resource management and monitoring. It aims to improve inference throughput and reduce inference latency, achieving a one-stop efficient AI service deployment and usage experience. | sig-ai-inference |
| eagle-eye | AI Inference Observability | Eagle Eye is an observability system for AI inference scenarios, implementing full-link metric collection, near-real-time transmission, and intelligent diagnosis from AI gateway, inference engine, Mooncake to infrastructure (Ray, K8s, hardware). The system integrates Prometheus periodic metric collection with the low-latency push mechanism of distributed message queue systems, supporting trend analysis for scaling decisions as well as meeting the needs of time-sensitive modules (such as intelligent routing) for second-level data updates. Through an independent hardware health diagnosis module, it achieves continuous monitoring and anomaly identification of underlying metrics including NPU/GPU, temperature, power consumption, and error codes, building a closed-loop monitoring capability of "collection — transmission — diagnosis — evaluation", providing solid data support for the stability, performance optimization, and resource scheduling of AI inference systems. | sig-ai-inference |
| hermes-router | AI Inference Intelligent Routing | Hermes-router is an AI inference intelligent routing component based on the K8s GIE framework, providing various routing strategies including KVCache aware and latency prediction, used to receive inference requests and forward them to the optimal inference service backend, helping users improve AI inference performance, cluster resource utilization, and service stability across multiple scenarios. | sig-ai-inference |
| cache-indexer | AI Inference KVCache Index Management | cache-indexer is a global KVCache index management component for AI inference scenarios, uniformly maintaining both local HBM and memory-level KVCache views of inference instances, supporting intelligent routing to query candidate inference instances' KVCache hit rates for more accurate request-level scheduling. | sig-ai-inference |
| pd-orchestrator | Elastic Scheduling Component Set | An elastic scheduling component set for Kubernetes, containing 3 core components: Elastic Scaler provides general scaling decision capabilities, ResourceScalingGroup provides workload group scaling and resource orchestration capabilities, Tidal provides time-rule-based tidal scaling capabilities. | sig-ai-inference |
| weight-dispatcher | Model Weight Distribution | Weight-Dispatcher is a model weight preheating and distribution component for AI inference scenarios, used to distribute model weights from source nodes or external model repositories to target inference nodes' local cache directories before inference instances start, reducing the wait time and bandwidth pressure caused by repeatedly downloading large model weights during instance startup. | sig-ai-inference |
| kae-operator | KAE Access Enablement | Implements minute-level automated management capabilities for Kunpeng KAE hardware, including KAE hardware feature discovery, automated management and installation of drivers, firmware, and hardware device plugins. KAE can be deployed and ready for use within five minutes. | sig-orchestration-engine |
| logging-package | Logging Component | As an extension component, logging allows viewing Pod and container log information, and supports configuring log collection sources, collection tasks, and alert rules. The openFuyao logging system effectively improves the platform's error identification efficiency and enhances the overall platform monitoring capabilities. | sig-container-platform |
| colocation-package | Online/Offline Colocation Scheduling Enhancement | Supports online/offline business mixed deployment, ensuring online business scheduling and suppression of offline business during peak usage periods, while enabling offline business to use overcommitted resources during online business low periods to improve cluster resource utilization, with utilization improvement of 30%~50% and no significant QoS impact, jitter below 5%. | sig-orchestration-engine |
| many-core-orchestrator | Many-Core Colocation Scheduling | For many-core colocation clusters, provides host interference metric collection, node interference analysis, Volcano interference-aware scheduling, and optional Kata Containers VM-level isolation capabilities. The goal is to reduce long-tail latency jitter of online business under I/O, LLC cache, and memory pressure colocation scenarios, and improve node secure colocation density. | sig-orchestration-engine |
| matrixagent | Memory Borrowing Node-Side Function | matrixagent periodically collects node memory usage and reports to matrixcontroller, while monitoring escape decisions or memory borrowing, calling VirtAgent for memory borrowing, balancing cluster memory usage, and improving resource utilization. | sig-ub-enable |
| matrixcontroller | Memory Borrowing Center-Side Function | matrixcontroller monitors data reported by matrixagent, calculates memory watermarks, and automatically initiates memory borrowing requests based on thresholds, balancing cluster memory usage and improving resource utilization. | sig-ub-enable |
| matrixshm | Shared Memory CSI | Provides a CR declarative isolation-based memory sharing solution for Pods, supporting only full mapping and unmapping, with capacity constrained by both configured block size and NUMA lending ratio, and Pod scheduling controlled by node label sharing domains. Its lifecycle management is tightly bound to Pod exit, with multi-layer mechanisms including shm-agent monitoring, reference counting, operation failure rollback, and startup state recovery to prevent resource leakage, combined with path verification, strict permission control, atomic write caching, and task queue overload protection to ensure security and high availability. | sig-ub-enable |
| monitoring-dashboard | Custom Monitoring Dashboard | Used to present extensible metrics, meeting enterprise needs for highly customizable and extensible monitoring systems. | sig-container-platform |
| multi-cluster-service | Multi-Cluster Management | Provides a cluster list interface and full lifecycle management capabilities, supporting cluster adoption, scaling, configuration updates, and secure destruction. Users can achieve quick access through cluster credentials and flexibly edit cluster labels through label management. This component implements cross-cluster secure access, helping efficiently and uniformly manage multiple Kubernetes clusters in hybrid environments. | sig-container-platform |
| npu-operator | NPU Access Enablement | NPU Operator uses the Operator Framework in Kubernetes to automatically manage all Ascend drivers and firmware required for configuring Ascend devices, enabling the full cluster operation process, supporting cluster job scheduling, O&M monitoring, fault recovery, and other functions. By installing corresponding components, NPU resource management, optimized workload scheduling, containerized support for training and inference tasks can be achieved, enabling AI jobs to be deployed and run on NPU devices in container form. | sig-orchestration-engine |
| numa-affinity-package | NUMA Affinity Scheduling | Implements hardware NUMA topology awareness at both cluster and node levels, and performs NUMA affinity scheduling for applications based on NUMA affinity, improving application performance. | sig-container-platform |
| ray-package | openFuyao/ray | A "parallel universe" project derived from ray-project/ray, used to extend the unified distributed computing framework for AI and Python applications, dedicated to providing stable, reliable, and domestically hardware-compatible LTS (Long-Term Support) versions for Ray users and operators. | sig-distributed-framework |
| compliance-operator | Compliance Scanning | Compliance Operator is an automated compliance scanning tool based on the Kubernetes Operator pattern, supporting security compliance scanning of Kubernetes clusters using CIS Benchmark and DISA STIG standards. | sig-installation |
| npu-dra-plugin | Ascend NPU-Affine DRA Plugin | Based on the Kubernetes DRA mechanism, supports 310P, 910B, 910C card models; supports full-card and hard partition mounting. | sig-orchestration-engine |
| kubevirt | Container-VM Co-Management Component | kubevirt provides VM-container co-management capabilities in Kunpeng environments, supporting full VM lifecycle management, macvlan, SR-IOV, SR-IOV live migration, network card hot-plug capabilities, and VM live migration. Verified that current kubevirt does not support Kunpeng UEFI secure boot. | sig-orchestration-engine |
Table 4 Independent Components (technical preview status)
| Feature Name | Brief Description (English) | Feature Introduction | SIG |
|---|---|---|---|
| sandbox-controller | Agent Sandbox Management | Provides sandbox lifecycle management, warm-up pool mechanism, and enhanced checkpoint/restore capabilities. | sig-orchestration-engine |
| sandbox-agent | Agent Sandbox Management | Provides sandbox lifecycle management, warm-up pool mechanism, and enhanced checkpoint/restore capabilities. | sig-orchestration-engine |
| node-agent | Agent Sandbox Management | Provides sandbox lifecycle management, warm-up pool mechanism, and enhanced checkpoint/restore capabilities. | sig-orchestration-engine |
Licensed under the MulanPSL2
Copyright © 2026 openFuyao. All rights reserved.ICP Record Number: YueA2-20044005-306
YueGongWangAnBei No.44030002007300
YueGongWangAnBei No.44030002007300Licensed underthe MulanPSL2
Copyright © 2026 openFuyao. All rights reserved.
Bug