Release Date | Version Number | Change Content | Limitations and Impact |
2026-09-20 | v1.6.11 | Fixed the issue where the crane-descheduler ReduceHighLoadNode policy had low eviction throughput and failed to reduce load in a timely manner in some special business Pod scenarios. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2026-04-20 | v1.6.10 | The method of adding a Plugin supports crane-scheduler and the referencing of a pooled scheduler. Update the K8S dependency to v1.34 to fix the issue where the crane-scheduler-controller webhook is missing the Pod field. It supports building multi-architecture (amd64/arm64) images. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2026-01-21 | v1.6.9 | Optimized the logic for obtaining node loads in the crane-descheduler ReduceHighLoadNode policy. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2025-12-04 | v1.6.8 | Optimized related logic of crane-descheduler: Optimize the logic for obtaining Pod Owner References. Resolve the issue of abnormal operation of the auto-tidying policy. A new switch is added to control whether sensitive-class services are allowed to be evicted. Fixed the issue where newly added native nodes cannot use the dynamic scaling feature. Optimize the scheduling speed of crane-scheduler. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2025-09-10 | v1.6.7 | crane-scheduler supports preemptVerb. Fixed the issue where crane-descheduler cannot obtain node loads. crane-descheduler supports printing the request information when load data is obtained. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2025-08-11 | v1.6.6 | crane-scheduler supports prioritizing the scaling down of nodes with low GPU/CPU/MEMORY bin packing rates. crane-scheduler fixed the oversell ratio calculation issue. When calculating the RS replica quantity, crane-scheduler PlacementPolicy no longer includes Pods in the terminated state. crane-descheduler added support for automatic GPU tidying. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2025-07-22 | v1.6.5 | Optimized related logic of crane-descheduler: Optimize the Kubernetes object query logic. Optimize the logic for waiting for pod readiness after pod eviction. The configuration property name "Exceed Eviction Watermark Count" has been modified. crane-descheduler supports automatic tidying (disabled by default). | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2025-06-19 | v1.6.4 | The crane-scheduler PlacementPolicy supports scheduling across node groups based on workload proportions. Fixed the issue where the crane-scheduler webhook caused missing pod fields. crane-scheduler supports node registration for the PlacementPolicy. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2025-06-06 | v1.6.3 | The crane-scheduler PlacementPolicy supports compact/balanced scheduling within a single node group. Fixed the issue where the crane-scheduler PlacementPolicy lacked field descriptions during explanation. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2025-05-20 | v1.6.2 | crane-scheduler supports real-time scheduling interception. crane-scheduler and crane-descheduler have been migrated to Locksmith. crane-descheduler supports persistent configuration for the "Exceed Eviction Watermark Count" setting. crane-descheduler optimizes the dispersal and eviction of high-load nodes. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2025-04-15 | v1.6.1 | Fixed the bug where the crane-scheduler-controller exits abnormally when a node is deleted. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2025-04-02 | v1.6.0 | New feature: supports intelligent scheduling, allowing users to flexibly define scheduling rules through PlacementPolicy. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2025-03-19 | v1.5.3 | crane-scheduler-controller optimizes node load calculation and data retrieval methods to reduce data update latency. Optimize the logic for obtaining pod recommendation values within the pod usage deduction feature of crane-scheduler. The descheduler cannot re-elect a leader after the main logic of its primary replica program exits. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2024-11-26 | v1.5.2 | Optimized the algorithm for node allocation rate when adjusting the dynamic node scaling factor. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2024-11-20 | v1.5.1 | Optimize the access method to the APIServer during node scale-out. Enable nodeCache to optimize scheduling performance. Fixed the bug where only 20 nodes' scaling coefficients are updated at a time during dynamic node scale-out. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2024-10-30 | v1.5.0 | New Features: crane-scheduler supports the load awareness scheduling feature for all native nodes, which can be enabled on demand in the component management configuration page. crane-scheduler supports that when multiple cnrp select the same node, only the latest cnrp takes effect. crane-descheduler supports setting the policy execution interval and the minimum number of low-load nodes for high-load distribution. Optimizations: When pods are scheduled to native nodes, workloads of type ds are skipped to adapt to scenarios where the cluster contains both native nodes and regular nodes. Fixed the issue where, due to data latency, the scheduler reschedules pods back to the source node after the descheduler evicts them. Fixed the issue where newly created native nodes cannot be scheduled until 5 minutes after creation due to empty utilization data for the first 5 minutes. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2024-08-22 | v1.4.3 | Optimize crane-descheduler to address the issue where the scheduler schedules pods onto high-load nodes while crane-descheduler is in the process of reducing node load (by evicting pods). Optimize the execution efficiency of the high-load distribution policy, reduce the latency in detecting high-load nodes, and optimize the policy execution interval. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2024-07-12 | v1.4.2 | Optimized the "Pod Load Deduction Scheduling" feature to avoid CA blocking. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2024-06-27 | v1.4.1 | Supports disabling the eviction stop watermark and optimizes node annotations after node auto-scaling is disabled. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2024-06-13 | v1.4.0 | Added the "Node Dynamic Scaling" feature, which can be enabled on demand via the console. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2024-05-09 | v1.3.0 | The "Pod Load Deduction Scheduling" feature is added and can be enabled on-demand in the console. Adjust the default weight for hotspots and optimize updates for metrics and other items. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2024-01-29 | v1.2.0 | Converged security permissions. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2023-10-13 | v1.1.11 | Components work with multiple replicas by default to improve availability. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2023-08-16 | v1.1.10 | Fix the issue of failing to obtain Pod load. Fix the issue of evicting excessive Pods when node memory reaches the eviction watermark. Add a new eviction stop watermark. Add the hot-value metric for external observation. When the hotspot value of a compute node is calculated, DaemonSet pods are ignored by default. It supports all components to work normally on super nodes. Improve the QoS and priority of all components on regular nodes to ensure they work with high priority. Decommission the hot-value solution implemented by node-annotator. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2023-07-27 | v1.1.9 | Implemented the hot-value feature using the PodInformer method. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2023-07-17 | v1.1.8 | Decommission the hot-value solution implemented via the PreBind extension point, and continue to use the original annotate node method. The PreBind solution has a single point of failure issue, which can block the scheduling of the entire cluster and poses an excessively high risk. Fix the issue of scheduler configuration failure when upgrading a cluster from version 1.18 to version 1.20. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2023-06-30 | v1.1.7 | The certificate algorithm for crane-scheduler-controller has been updated from sha1 to sha256, and it supports clusters of version 1.22 and above. Helm Management Optimization: The Hook deletePolicy supports reentrant operations. Add the post-install, post-upgrade, and pre-upgrade extension points for manageControlPlane to support in-place upgrades for PreBind. Redo the hot-value feature and add the PreBind extension point to avoid node hotspot issues: Currently, crane-scheduler only supports running in a single-replica mode. crane-scheduler-controller is compatible with the hot-value.weight configuration. Optimize metric reporting to mitigate OOM issues. Disable the hostNetwork networking mode for crane-scheduler-controller and descheduler to prevent liveness probe failures caused by port conflicts. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2023-06-21 | v1.1.6 | Remove the hostNetwork setting from the deployments of crane-scheduler-controller and descheduler. crane-scheduler-controller is compatible with the hot-value.weight configuration. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2023-06-20 | v1.1.5 | The certificate algorithm has been updated from sha1 to sha256. crane-scheduler-controller supports clusters of version 1.22 and above. The Helm Hook deletePolicy supports reentrant operations. Add the PreBind extension point and use the scheduler's cache hotspot to avoid node hotspot issues. Optimize metric reporting to mitigate OOM issues. Add the post-install, post-upgrade, and pre-upgrade extension points for manageControlPlane to support in-place upgrades for PreBind. Modify the default hot-value configuration for crane-scheduler. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2023-03-23 | v1.1.4 | Fixed the issue where newly added native nodes receive low scores during the scoring phase, making Pods difficult to schedule. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
2022-09-21 | v1.1.3 | Fix the issue where missing Helm repo parameters prevent node metrics from being updated to the node. Integrate the descheduler into CraneScheduler and add the eviction waterline for the rescheduler, aligning the dynamic scheduler waterline with the rescheduler target waterline. | This upgrade does not affect existing services. During the upgrade process, the component may be unavailable. It is recommended to upgrade during off-peak hours. |
Esta página foi útil?
Você também pode entrar em contato com a Equipe de vendas ou Enviar um tíquete em caso de ajuda.
comentários