Customize kube-scheduler parameters

Prerequisites

An ACK Pro cluster, ACK Edge Pro cluster, ACK Lingjun cluster, or ACK Serverless Pro cluster running Kubernetes 1.20 or later is created. To update the Kubernetes version, manually upgrade the cluster.

Limitations

Procedure

1. Log on to the ACK console. In the left navigation pane, click Clusters.

2. On the Clusters page, find the one you want to manage and click its name. In the left navigation pane, click Add-ons.

3. On the Core Components tab, find Kube Scheduler and click Configuration. Configure the parameters in the Kube Scheduler Parameters dialog box.

   Available parameters vary by kube-scheduler version. For version-specific features, see Container Service for Kubernetes:kube-scheduler.

   Parameter	Description	Type	Valid value	Default value
   Enable Virtual Node-based Pod Scheduling	Specifies whether to schedule pods on virtual nodes based on node affinity and pod spread constraints.	bool	true false	true
   podMaxBackoffSeconds	Maximum backoff interval (in seconds) between scheduling attempts for a failed pod. The interval between two scheduling attempts must exceed this value.	int	[1,100000]	10
   Preferably Use Bin Packing During Pod Scheduling	Specifies whether to enable the binpack algorithm. For sample configurations, see Custom binpack parameters.	bool	false true	false
   binpackPluginWeight	Weight of the binpack plug-in node score. Requires Preferably Use Bin Packing During Pod Scheduling.	int	[1,100000]	100
   binpackResourceWeight	Weight of each resource in the binpack plug-in node score calculation. Requires Preferably Use Bin Packing During Pod Scheduling.	resourceName: string resourceWeight: int	resourceName: letters, digits, periods (.), forward slashes (/), and hyphens (-) only. resourceWeight: integer from 1 to 10000.	cpu:1 memory:1
   scorePluginWeights	Weight of each scoring plug-in for node score calculation. Note The NodeResourceFit plug-in weight conflicts with the binpack plug-in weight. If you select Preferably Use Bin Packing During Pod Scheduling (Pods Are Evenly Distributed among Nodes When Unselected), do not set a NodeResourceFit weight for this parameter.	plugin: string weight: int	Only plug-ins in the plugin drop-down list are available. weight: integer from 1 to 10000.	plugin: NodeAffinity weight: 100
   percentageOfNodesToScore	The percentage of nodes suitable for pod scheduling. Default value: 0. When set to 0, 5%–50% of nodes are scored based on cluster size.	int	[0,100]	0
   Node scoring for GPU sharing	When using shared GPU scheduling (requires the AI suite), schedules GPU-accelerated pods to GPUs with higher memory and compute requests.	bool	false true	true
   Load-aware scoring during pod scheduling (loadAwareResourceWeight)	Specifies whether to enable load-aware scheduling. Requires the ack-koordinator component.	bool	false true	false
   loadAwareThreshold	This parameter specifies the threshold for node filtering.	The value consists of the resourceName and resourceWeight fields.	resourceName: Valid values are cpu and memory. threshold: Valid values range from 0 to 100. By default, this parameter is left empty, which disables node filtering.	resourceName: cpu threshold: 80
   loadAwareResourceWeight	This parameter specifies the resource weight used to calculate the node score for node sorting. This parameter is available after you select Specifies whether to enable load-aware node scoring during pod scheduling.	The value consists of the resourceName and resourceWeight fields.	resourceName: The schema of the resourceName parameter is verified. Values can only be cpu or memory. resourceWeight: Valid values are integers ranging from 1 to 100.	cpu=1 memory=1
   loadAwareAggregatedUsageAggregationType	This parameter specifies the type of data aggregation for the statistics. Valid values: avg: calculates the average value. p50: calculates 50% of the statistics. p90, p95, and p99: calculates 90% of the statistics, calculates 95% of the statistics, and calculates 99% of the statistics.	enum	avg p50 p90 p95 p99	avg
   preemptionAlgorithm	ACK Scheduler determines whether to evict lower-priority pods through resource simulation, prioritizing rapid startup of high-priority workloads. Supported strategies: Default: Kubernetes community-standard preemption ElasticQuota: Resource preemption based on ElasticQuotaTree Auto: Adaptive preemption policy based on cluster specifications None: Disables preemption See Enable preemption.	enum	Default ElasticQuota Auto None	Auto
   enableReservation	Specifies whether to enable resource reservation.	boolean	true false	false
   featureGates	Feature gates enabled by the scheduler. For version-specific support, see the kube-scheduler documentation.	string	N/A	ACK uses the same feature gates as open source Kubernetes.

The following example shows a sample binpack configuration.

Custom binpack parameters

Binpack vs. spread algorithm

Configure custom binpack parameters

Select Preferably Use Bin Packing During Pod Scheduling and set the binpack plug-in weight. A larger weight increases the chance of co-locating pods on the same node. Configure resource names and their weights for node score calculation. A larger resource weight gives that resource greater influence on pod scheduling.

On the Core Components tab of the Add-ons page, find Kube Scheduler and click Configuration to configure the binpack parameters.

If you select Preferably Use Bin Packing During Pod Scheduling without configuring resourceName and resourceWeight, the default CPU and memory settings shown below are used for node score calculation. The defaults are resourceName cpu with resourceWeight 1, and resourceName memory with resourceWeight 1..

References

To schedule pods based on actual resource usage rather than requested amounts, enable load-aware scheduling.