kube-scheduler-Container Compute Service-阿里云帮助中心

kube-scheduler is the control plane component that decides where the pods in an ACS cluster run. Configure its custom parameters to align scheduling behavior with the resource, affinity, and zone requirements of your workloads.

kube-scheduler overview

Based on the resources and other properties of pods and nodes, kube-scheduler selects a suitable virtual node in the ACS cluster for each pod. For more information, see kube-scheduler in the official Kubernetes documentation.

The kube-scheduler component is installed by default when you create a cluster. For more information about the scheduling capabilities that it provides, see scheduling overview.

Component configuration

The kube-scheduler component supports custom parameters. To configure these parameters, complete the following steps:

Log on to the ACS console. In the left navigation pane, click Clusters.
On the Clusters page, click the name of the target cluster. In the left navigation pane, click Add-ons.

On the Core Components tab, find the kube-scheduler card and click Configuration in the lower-right corner of the card. In the dialog box, configure the parameters as prompted.

Different scheduler versions support different features, and therefore different custom parameters. For more information about the features that each scheduler version supports, see kube-scheduler. The following table describes the meaning and usage of each custom parameter.

Configuration item	Description	Type	Valid values	Default value
Enable Zone Recommendation	During scheduling, ACS automatically filters out zones whose resources are constrained. For more information, see zone resource recommendation feature.	bool	`true`, `false`	`true`
Enable custom labels and schedulers for GPU-HPN nodes	For more information, see Use a custom scheduler and Manage node labels and taints.	bool	`true`, `false`	`false`
`scorePluginWeights`	The weight of each scoring plugin of the scheduler. The weight affects which nodes the scheduler prioritizes. For more information about scoring weights, see the Score . Both native community plugins and Alibaba Cloud proprietary plugins are supported. Community plugins include NodeAffinity, PodTopologySpread, InnerPodAffinity, and TaintToleration. For more information, see the affinity and toleration policy documentation. Alibaba Cloud proprietary plugins include ServerlessScheduling. For the general-purpose, performance, and gpu compute classes, this plugin gives priority to the zones that have more idle platform resources, based on zone resource recommendations. Considerations: During scheduling, ACS automatically filters out zones whose resources are constrained, and the filtering stage is not affected by scoring policies. When you use affinity-related scoring policies, select multiple Recommended zones for the pod so that the scoring policies take full effect. For more information, see zone resource recommendation feature.	`plugin`: string. `weight`: int.	`plugin`: only the plugins in the public plugin list. `weight`: an integer in the range [1,10000].	The default weight of each scoring plugin is as follows: NodeAffinity: 1. PodTopologySpread: 1. InnerPodAffinity: 1. TaintToleration: 1. ServerlessScheduling: 2.

Change history

Starting with Kubernetes 1.32, the version number rules of the ACS scheduler are the same as those of ACK. For more information about how the version numbers in the following tables are composed, see kube-scheduler.

Version 1.32

Version number	Release date	Changes
v1.32.0-apsara.6.11.8.507bee55	December 22, 2025	New features: Added support for using the label alibabacloud.com/acs: "true" to directly specify ACS compute resources (general-purpose, performance, or gpu), ignoring pod affinity constraints.`alibabacloud.com/acs: "true"` to directly specify ACS compute resources (general-purpose, performance, or gpu) and ignore pod affinity constraints. Enhancements: Expanded the scenarios covered by GPU-HPN topology-aware scheduling. For more information, see GPU-HPN topology-aware scheduling. Enabled the Enable custom labels and schedulers for GPU-HPN nodes feature by default. Bug fixes: Fixed the issue where GPU-HPN pods were scheduled to nodes whose GPU resource types did not match. Trigger condition: the Enable custom labels and schedulers for GPU-HPN nodes feature is enabled and the pod uses the default scheduler of the cluster.
v1.32.0-aliyun.6.9.7.eb18a8a9	September 22, 2025	Bug fixes: Fixed the issue where pods remained Pending when PodAntiAffinity and TopologySpread were used for strict spreading. This issue was introduced in v1.32.0-aliyun.6.9.6.4e1407c2.
v1.32.0-aliyun.6.9.6.4e1407c2	September 17, 2025	New features: Added a switch for the zone resource recommendation feature. Enhancements: Optimized the scheduling retry efficiency for pods that use a Persistent Volume Claim (PVC). Optimized the PodTopologySpread policy. When `whenUnsatisfiable` is set to `ScheduleAnyway`, pods that are being created are also included in the topology distribution calculation.
v1.32.0-aliyun.6.9.5.ccfc0094	July 9, 2025	Initial release of the component. This version is compatible with ACS cluster 1.32 and includes all the ACS features of v1.31.0-aliyun-1.4.2.

Version 1.31

Version number	Release date	Changes
v1.31.0-aliyun-1.5.0	December 22, 2025	New features: Added support for using the label alibabacloud.com/acs: "true" to directly specify ACS compute resources (general-purpose, performance, or gpu), ignoring pod affinity constraints.`alibabacloud.com/acs: "true"` to directly specify ACS compute resources (general-purpose, performance, or gpu) and ignore pod affinity constraints. Enhancements: Expanded the scenarios covered by GPU-HPN topology-aware scheduling. For more information, see GPU-HPN topology-aware scheduling. Enabled the Enable custom labels and schedulers for GPU-HPN nodes feature by default. Bug fixes: Fixed the issue where GPU-HPN pods were scheduled to nodes whose GPU resource types did not match. Trigger condition: the Enable custom labels and schedulers for GPU-HPN nodes feature is enabled and the pod uses the default scheduler of the cluster.
v1.31.0-aliyun-1.4.4	September 22, 2025	Bug fixes: Fixed the issue where pods remained Pending when PodAntiAffinity and TopologySpread were used for strict spreading. This issue was introduced in v1.31.0-aliyun-1.4.3.
v1.31.0-aliyun-1.4.3	September 17, 2025	New features: Added a switch for the zone resource recommendation feature. Enhancements: Optimized the scheduling retry efficiency for pods that use a PVC. Optimized the PodTopologySpread policy. When `whenUnsatisfiable` is set to `ScheduleAnyway`, pods that are being created are also included in the topology distribution calculation.
v1.31.0-aliyun-1.4.2	July 9, 2025	Optimized the internal log printing implementation and the base image of the component.
v1.31.0-aliyun-1.4.1	May 19, 2025	Added support for custom weight parameters of scoring plugins. For more information, see kube-scheduler.
v1.31.0-aliyun-1.4.0	April 9, 2025	Optimized the scheduling process for pods that use WaitForFirstConsumer PVCs. When a pod uses a WaitForFirstConsumer PVC and also specifies a vSwitch, specify the corresponding zone in the pod affinity settings. The scheduler ensures that the zone allocated to the PVC is consistent with the vSwitch, which prevents the pod from remaining Pending. For more information, see Pod annotation features.
v1.31.0-aliyun-1.3.1	March 20, 2025	GPU sharing scheduling supports the static policy and Quality of Service (QoS) settings. Custom resource priority scheduling provides better fault tolerance for invalid configurations.
v1.31.0-aliyun-1.3.0	February 8, 2025	Fixed the cloud disk scheduling issue for pods that use WaitForFirstConsumer PVCs. Added support for configuring a custom scheduler for GPU-HPN pods.
v1.31.0-aliyun-1.2.0	November 29, 2024	Initial release of the component. This version is compatible with ACS cluster 1.31.

Version 1.30

Version number	Release date	Changes
v1.30.3-aliyun-1.6.0	December 22, 2025	New features: Added support for using the label alibabacloud.com/acs: "true" to directly specify ACS compute resources (general-purpose, performance, or gpu), ignoring pod affinity constraints.`alibabacloud.com/acs: "true"` to directly specify ACS compute resources (general-purpose, performance, or gpu) and ignore pod affinity constraints. Enhancements: Expanded the scenarios covered by GPU-HPN topology-aware scheduling. For more information, see GPU-HPN topology-aware scheduling. Enabled the Enable custom labels and schedulers for GPU-HPN nodes feature by default. Bug fixes: Fixed the issue where GPU-HPN pods were scheduled to nodes whose GPU resource types did not match. Trigger condition: the Enable custom labels and schedulers for GPU-HPN nodes feature is enabled and the pod uses the default scheduler of the cluster.
v1.30.3-aliyun-1.4.4	September 22, 2025	Bug fixes: Fixed the issue where pods remained Pending when PodAntiAffinity and TopologySpread were used for strict spreading. This issue was introduced in v1.30.3-aliyun-1.4.3.
v1.30.3-aliyun-1.4.3	September 17, 2025	New features: Added a switch for the zone resource recommendation feature. Enhancements: Optimized the scheduling retry efficiency for pods that use a PVC. Optimized the PodTopologySpread policy. When `whenUnsatisfiable` is set to `ScheduleAnyway`, pods that are being created are also included in the topology distribution calculation.
v1.30.3-aliyun-1.4.2	July 9, 2025	Optimized the internal log printing implementation and the base image of the component.
v1.30.3-aliyun-1.4.1	May 19, 2025	Added support for custom weight parameters of scoring plugins. For more information, see scorePluginWeights.
v1.30.3-aliyun-1.4.0	April 9, 2025	Optimized the scheduling process for pods that use WaitForFirstConsumer PVCs. When a pod uses a WaitForFirstConsumer PVC and also specifies a vSwitch, specify the corresponding zone in the pod affinity settings. The scheduler ensures that the zone allocated to the PVC is consistent with the vSwitch, which prevents the pod from remaining Pending. For more information, see Pod annotation features.
v1.30.3-aliyun-1.3.1	March 20, 2025	GPU sharing and scheduling supports the static policy and QoS settings. Custom resource priority scheduling provides better fault tolerance for invalid configurations.
v1.30.3-aliyun-1.3.0	February 8, 2025	Fixed the cloud disk scheduling issue for pods that use WaitForFirstConsumer PVCs. Added support for configuring a custom scheduler for GPU-HPN pods.
v1.30.3-aliyun-1.2.0	November 29, 2024	Optimized pod scheduling efficiency. Adapted to the new version of the `acs-virtual-node` add-on. The `acs-virtual-node` add-on must be v2.12.0-acs.9 or later.
v1.30.3-aliyun-1.1.1	September 9, 2024	Initial release of the component. This version is compatible with ACS cluster 1.30.

Version 1.28

Version number	Release date	Changes
v1.28.12-aliyun-1.4.5	December 22, 2025	Bug fixes: Fixed the issue where GPU-HPN pods were scheduled to nodes whose GPU resource types did not match. Trigger condition: the Enable custom labels and schedulers for GPU-HPN nodes feature is enabled and the pod uses the default scheduler of the cluster.
v1.28.12-aliyun-1.4.4	September 22, 2025	Bug fixes: Fixed the issue where pods remained Pending when PodAntiAffinity and TopologySpread were used for strict spreading. This issue was introduced in v1.28.12-aliyun-1.4.3.
v1.28.12-aliyun-1.4.3	September 17, 2025	New features: Added a switch for the zone resource recommendation feature. Enhancements: Optimized the scheduling retry efficiency for pods that use a PVC. Optimized the PodTopologySpread policy. When `whenUnsatisfiable` is set to `ScheduleAnyway`, pods that are being created are also included in the topology distribution calculation.
v1.28.12-aliyun-1.4.2	July 9, 2025	Optimized the internal log printing implementation and the base image of the component.
v1.28.12-aliyun-1.4.1	May 19, 2025	Added support for custom weight parameters of scoring plugins. For more information, see scorePluginWeights.
v1.28.12-aliyun-1.4.0	April 9, 2025	Optimized the scheduling process for pods that use WaitForFirstConsumer PVCs. When a pod uses a WaitForFirstConsumer PVC and also specifies a vSwitch, specify the corresponding zone in the pod affinity settings. The scheduler ensures that the zone allocated to the PVC is consistent with the vSwitch, which prevents the pod from remaining Pending. For more information, see Pod annotation features.
v1.28.9-aliyun-1.3.1	March 20, 2025	GPU sharing and scheduling supports the static policy and QoS settings. Custom resource priority scheduling provides better fault tolerance for invalid configurations.
v1.28.9-aliyun-1.3.0	February 8, 2025	Fixed the cloud disk scheduling issue for pods that use WaitForFirstConsumer PVCs. Added support for configuring a custom scheduler for GPU-HPN pods.
v1.28.9-aliyun-1.2.0	November 29, 2024	Optimized pod scheduling efficiency. Adapted to the new version of the `acs-virtual-node` add-on. The `acs-virtual-node` add-on must be v2.12.0-acs.9 or later.
v1.28.9-aliyun-1.1.1	September 6, 2024	Optimized event messages.
v1.28.9-aliyun-1.1.0	August 22, 2024	Added support for custom resource priority scheduling.
v1.28.9-aliyun-1.0.1.dfd9511f	July 12, 2024	Optimized internal interfaces.
v1.28.9-aliyun-1.0.0b72f000	June 15, 2024	Initial release of the component. This version is compatible with ACS cluster 1.28.

Version 1.26

Version number	Release date	Changes
v1.26.3-aliyun-1.3.1	March 20, 2025	GPU sharing and scheduling supports the static policy and QoS settings. Custom resource priority scheduling provides better fault tolerance for invalid configurations.
v1.26.3-aliyun-1.3.0	February 8, 2025	Fixed the cloud disk scheduling issue for pods that use WaitForFirstConsumer PVCs. Added support for configuring a custom scheduler for GPU-HPN pods.
v1.26.3-aliyun-1.2.0	November 29, 2024	Optimized pod scheduling efficiency. Adapted to the new version of the `acs-virtual-node` add-on. The `acs-virtual-node` add-on must be v2.12.0-acs.9 or later.
v1.26.3-aliyun-1.1.1	September 6, 2024	Optimized event messages.
v1.26.3-aliyun-1.1.0	August 22, 2024	Added support for custom resource priority scheduling.
v1.26.3-aliyun-1.0.1.4d326e78	July 12, 2024	Optimized internal interfaces.
v1.26.3-aliyun-1.0.fa88f9ae	June 25, 2024	Initial release of the component. This version is compatible with ACS cluster 1.26.