You are viewing documentation for Kubernetes version: v1.24

Kubernetes v1.24 documentation is no longer actively maintained. The version you are currently viewing is a static snapshot. For up-to-date documentation, see the latest version.

Scheduler Configuration

FEATURE STATE: Kubernetes v1.19 [beta]

You can customize the behavior of the kube-scheduler by writing a configuration file and passing its path as a command line argument.

A scheduling Profile allows you to configure the different stages of scheduling in the kube-scheduler. Each stage is exposed in an extension point. Plugins provide scheduling behaviors by implementing one or more of these extension points.

You can specify scheduling profiles by running kube-scheduler --config <filename>, using the KubeSchedulerConfiguration (v1beta2 or v1beta3) struct.

A minimal configuration looks as follows:

apiVersion: kubescheduler.config.k8s.io/v1beta2
kind: KubeSchedulerConfiguration
clientConnection:
  kubeconfig: /etc/srv/kubernetes/kube-scheduler/kubeconfig

Profiles

A scheduling Profile allows you to configure the different stages of scheduling in the kube-scheduler. Each stage is exposed in an extension point. Plugins provide scheduling behaviors by implementing one or more of these extension points.

You can configure a single instance of kube-scheduler to run multiple profiles.

Extension points

Scheduling happens in a series of stages that are exposed through the following extension points:

  1. queueSort: These plugins provide an ordering function that is used to sort pending Pods in the scheduling queue. Exactly one queue sort plugin may be enabled at a time.
  2. preFilter: These plugins are used to pre-process or check information about a Pod or the cluster before filtering. They can mark a pod as unschedulable.
  3. filter: These plugins are the equivalent of Predicates in a scheduling Policy and are used to filter out nodes that can not run the Pod. Filters are called in the configured order. A pod is marked as unschedulable if no nodes pass all the filters.
  4. postFilter: These plugins are called in their configured order when no feasible nodes were found for the pod. If any postFilter plugin marks the Pod schedulable, the remaining plugins are not called.
  5. preScore: This is an informational extension point that can be used for doing pre-scoring work.
  6. score: These plugins provide a score to each node that has passed the filtering phase. The scheduler will then select the node with the highest weighted scores sum.
  7. reserve: This is an informational extension point that notifies plugins when resources have been reserved for a given Pod. Plugins also implement an Unreserve call that gets called in the case of failure during or after Reserve.
  8. permit: These plugins can prevent or delay the binding of a Pod.
  9. preBind: These plugins perform any work required before a Pod is bound.
  10. bind: The plugins bind a Pod to a Node. bind plugins are called in order and once one has done the binding, the remaining plugins are skipped. At least one bind plugin is required.
  11. postBind: This is an informational extension point that is called after a Pod has been bound.
  12. multiPoint: This is a config-only field that allows plugins to be enabled or disabled for all of their applicable extension points simultaneously.

For each extension point, you could disable specific default plugins or enable your own. For example:

apiVersion: kubescheduler.config.k8s.io/v1beta2
kind: KubeSchedulerConfiguration
profiles:
  - plugins:
      score:
        disabled:
        - name: PodTopologySpread
        enabled:
        - name: MyCustomPluginA
          weight: 2
        - name: MyCustomPluginB
          weight: 1

You can use * as name in the disabled array to disable all default plugins for that extension point. This can also be used to rearrange plugins order, if desired.

Scheduling plugins

The following plugins, enabled by default, implement one or more of these extension points:

  • ImageLocality: Favors nodes that already have the container images that the Pod runs. Extension points: score.
  • TaintToleration: Implements taints and tolerations. Implements extension points: filter, preScore, score.
  • NodeName: Checks if a Pod spec node name matches the current node. Extension points: filter.
  • NodePorts: Checks if a node has free ports for the requested Pod ports. Extension points: preFilter, filter.
  • NodeAffinity: Implements node selectors and node affinity. Extension points: filter, score.
  • PodTopologySpread: Implements Pod topology spread. Extension points: preFilter, filter, preScore, score.
  • NodeUnschedulable: Filters out nodes that have .spec.unschedulable set to true. Extension points: filter.
  • NodeResourcesFit: Checks if the node has all the resources that the Pod is requesting. The score can use one of three strategies: LeastAllocated (default), MostAllocated and RequestedToCapacityRatio. Extension points: preFilter, filter, score.
  • NodeResourcesBalancedAllocation: Favors nodes that would obtain a more balanced resource usage if the Pod is scheduled there. Extension points: score.
  • VolumeBinding: Checks if the node has or if it can bind the requested volumes. Extension points: preFilter, filter, reserve, preBind, score.
  • VolumeRestrictions: Checks that volumes mounted in the node satisfy restrictions that are specific to the volume provider. Extension points: filter.
  • VolumeZone: Checks that volumes requested satisfy any zone requirements they might have. Extension points: filter.
  • NodeVolumeLimits: Checks that CSI volume limits can be satisfied for the node. Extension points: filter.
  • EBSLimits: Checks that AWS EBS volume limits can be satisfied for the node. Extension points: filter.
  • GCEPDLimits: Checks that GCP-PD volume limits can be satisfied for the node. Extension points: filter.
  • AzureDiskLimits: Checks that Azure disk volume limits can be satisfied for the node. Extension points: filter.
  • InterPodAffinity: Implements inter-Pod affinity and anti-affinity. Extension points: preFilter, filter, preScore, score.
  • PrioritySort: Provides the default priority based sorting. Extension points: queueSort.
  • DefaultBinder: Provides the default binding mechanism. Extension points: bind.
  • DefaultPreemption: Provides the default preemption mechanism. Extension points: postFilter.

You can also enable the following plugins, through the component config APIs, that are not enabled by default:

  • SelectorSpread: Favors spreading across nodes for Pods that belong to Services, ReplicaSets and StatefulSets. Extension points: preScore, score.
  • CinderLimits: Checks that OpenStack Cinder volume limits can be satisfied for the node. Extension points: filter.

Multiple profiles

You can configure kube-scheduler to run more than one profile. Each profile has an associated scheduler name and can have a different set of plugins configured in its extension points.

With the following sample configuration, the scheduler will run with two profiles: one with the default plugins and one with all scoring plugins disabled.

apiVersion: kubescheduler.config.k8s.io/v1beta2
kind: KubeSchedulerConfiguration
profiles:
  - schedulerName: default-scheduler
  - schedulerName: no-scoring-scheduler
    plugins:
      preScore:
        disabled:
        - name: '*'
      score:
        disabled:
        - name: '*'

Pods that want to be scheduled according to a specific profile can include the corresponding scheduler name in its .spec.schedulerName.

By default, one profile with the scheduler name default-scheduler is created. This profile includes the default plugins described above. When declaring more than one profile, a unique scheduler name for each of them is required.

If a Pod doesn't specify a scheduler name, kube-apiserver will set it to default-scheduler. Therefore, a profile with this scheduler name should exist to get those pods scheduled.

Plugins that apply to multiple extension points

Starting from kubescheduler.config.k8s.io/v1beta3, there is an additional field in the profile config, multiPoint, which allows for easily enabling or disabling a plugin across several extension points. The intent of multiPoint config is to simplify the configuration needed for users and administrators when using custom profiles.

Consider a plugin, MyPlugin, which implements the preScore, score, preFilter, and filter extension points. To enable MyPlugin for all its available extension points, the profile config looks like:

apiVersion: kubescheduler.config.k8s.io/v1beta3
kind: KubeSchedulerConfiguration
profiles:
  - schedulerName: multipoint-scheduler
    plugins:
      multiPoint:
        enabled:
        - name: MyPlugin

This would equate to manually enabling MyPlugin for all of its extension points, like so:

apiVersion: kubescheduler.config.k8s.io/v1beta3
kind: KubeSchedulerConfiguration
profiles:
  - schedulerName: non-multipoint-scheduler
    plugins:
      preScore:
        enabled:
        - name: MyPlugin
      score:
        enabled:
        - name: MyPlugin
      preFilter:
        enabled:
        - name: MyPlugin
      filter:
        enabled:
        - name: MyPlugin

One benefit of using multiPoint here is that if MyPlugin implements another extension point in the future, the multiPoint config will automatically enable it for the new extension.

Specific extension points can be excluded from MultiPoint expansion using the disabled field for that extension point. This works with disabling default plugins, non-default plugins, or with the wildcard ('*') to disable all plugins. An example of this, disabling Score and PreScore, would be:

apiVersion: kubescheduler.config.k8s.io/v1beta3
kind: KubeSchedulerConfiguration
profiles:
  - schedulerName: non-multipoint-scheduler
    plugins:
      multiPoint:
        enabled:
        - name: 'MyPlugin'
      preScore:
        disabled:
        - name: '*'
      score:
        disabled:
        - name: '*'

In v1beta3, all default plugins are enabled internally through MultiPoint. However, individual extension points are still available to allow flexible reconfiguration of the default values (such as ordering and Score weights). For example, consider two Score plugins DefaultScore1 and DefaultScore2, each with a weight of 1. They can be reordered with different weights like so:

apiVersion: kubescheduler.config.k8s.io/v1beta3
kind: KubeSchedulerConfiguration
profiles:
  - schedulerName: multipoint-scheduler
    plugins:
      score:
        enabled:
        - name: 'DefaultScore2'
          weight: 5

In this example, it's unnecessary to specify the plugins in MultiPoint explicitly because they are default plugins. And the only plugin specified in Score is DefaultScore2. This is because plugins set through specific extension points will always take precedence over MultiPoint plugins. So, this snippet essentially re-orders the two plugins without needing to specify both of them.

The general hierarchy for precedence when configuring MultiPoint plugins is as follows:

  1. Specific extension points run first, and their settings override whatever is set elsewhere
  2. Plugins manually configured through MultiPoint and their settings
  3. Default plugins and their default settings

To demonstrate the above hierarchy, the following example is based on these plugins:

Plugin Extension Points
DefaultQueueSort QueueSort
CustomQueueSort QueueSort
DefaultPlugin1 Score, Filter
DefaultPlugin2 Score
CustomPlugin1 Score, Filter
CustomPlugin2 Score, Filter

A valid sample configuration for these plugins would be:

apiVersion: kubescheduler.config.k8s.io/v1beta3
kind: KubeSchedulerConfiguration
profiles:
  - schedulerName: multipoint-scheduler
    plugins:
      multiPoint:
        enabled:
        - name: 'CustomQueueSort'
        - name: 'CustomPlugin1'
          weight: 3
        - name: 'CustomPlugin2'
        disabled:
        - name: 'DefaultQueueSort'
      filter:
        disabled:
        - name: 'DefaultPlugin1'
      score:
        enabled:
        - name: 'DefaultPlugin2'

Note that there is no error for re-declaring a MultiPoint plugin in a specific extension point. The re-declaration is ignored (and logged), as specific extension points take precedence.

Besides keeping most of the config in one spot, this sample does a few things:

  • Enables the custom queueSort plugin and disables the default one
  • Enables CustomPlugin1 and CustomPlugin2, which will run first for all of their extension points
  • Disables DefaultPlugin1, but only for filter
  • Reorders DefaultPlugin2 to run first in score (even before the custom plugins)

In versions of the config before v1beta3, without multiPoint, the above snippet would equate to this:

apiVersion: kubescheduler.config.k8s.io/v1beta2
kind: KubeSchedulerConfiguration
profiles:
  - schedulerName: multipoint-scheduler
    plugins:
    
      # Disable the default QueueSort plugin
      queueSort:
        enabled:
        - name: 'CustomQueueSort'
        disabled:
        - name: 'DefaultQueueSort'
        
      # Enable custom Filter plugins
      filter:
        enabled:
        - name: 'CustomPlugin1'
        - name: 'CustomPlugin2'
        - name: 'DefaultPlugin2'
        disabled:
        - name: 'DefaultPlugin1'
        
      # Enable and reorder custom score plugins
      score:
        enabled:
        - name: 'DefaultPlugin2'
          weight: 1
        - name: 'DefaultPlugin1'
          weight: 3

While this is a complicated example, it demonstrates the flexibility of MultiPoint config as well as its seamless integration with the existing methods for configuring extension points.

Scheduler configuration migrations

  • With the v1beta2 configuration version, you can use a new score extension for the NodeResourcesFit plugin. The new extension combines the functionalities of the NodeResourcesLeastAllocated, NodeResourcesMostAllocated and RequestedToCapacityRatio plugins. For example, if you previously used the NodeResourcesMostAllocated plugin, you would instead use NodeResourcesFit (enabled by default) and add a pluginConfig with a scoreStrategy that is similar to:

    apiVersion: kubescheduler.config.k8s.io/v1beta2
    kind: KubeSchedulerConfiguration
    profiles:
    - pluginConfig:
      - args:
          scoringStrategy:
            resources:
            - name: cpu
              weight: 1
            type: MostAllocated
        name: NodeResourcesFit
    
  • The scheduler plugin NodeLabel is deprecated; instead, use the NodeAffinity plugin (enabled by default) to achieve similar behavior.

  • The scheduler plugin ServiceAffinity is deprecated; instead, use the InterPodAffinity plugin (enabled by default) to achieve similar behavior.

  • The scheduler plugin NodePreferAvoidPods is deprecated; instead, use node taints to achieve similar behavior.

  • A plugin enabled in a v1beta2 configuration file takes precedence over the default configuration for that plugin.

  • Invalid host or port configured for scheduler healthz and metrics bind address will cause validation failure.

  • Three plugins' weight are increased by default:
    • InterPodAffinity from 1 to 2
    • NodeAffinity from 1 to 2
    • TaintToleration from 1 to 3

What's next

Last modified May 10, 2022 at 11:01 AM PST: Move Pod Topology Spread Constraints into scheduling (6700656cb4)