Horizontal Pod Autoscaler

The Horizontal Pod Autoscaler automatically scales the number of Pods in a replication controller, deployment, replica set or stateful set based on observed CPU utilization (or, with custom metrics support, on some other application-provided metrics). Note that Horizontal Pod Autoscaling does not apply to objects that can't be scaled, for example, DaemonSets.

The Horizontal Pod Autoscaler is implemented as a Kubernetes API resource and a controller. The resource determines the behavior of the controller. The controller periodically adjusts the number of replicas in a replication controller or deployment to match the observed metrics such as average CPU utilisation, average memory utilisation or any other custom metric to the target specified by the user.

How does the Horizontal Pod Autoscaler work?

The Horizontal Pod Autoscaler is implemented as a control loop, with a period controlled by the controller manager's --horizontal-pod-autoscaler-sync-period flag (with a default value of 15 seconds).

During each period, the controller manager queries the resource utilization against the metrics specified in each HorizontalPodAutoscaler definition. The controller manager obtains the metrics from either the resource metrics API (for per-pod resource metrics), or the custom metrics API (for all other metrics).

• For per-pod resource metrics (like CPU), the controller fetches the metrics from the resource metrics API for each Pod targeted by the HorizontalPodAutoscaler. Then, if a target utilization value is set, the controller calculates the utilization value as a percentage of the equivalent resource request on the containers in each Pod. If a target raw value is set, the raw metric values are used directly. The controller then takes the mean of the utilization or the raw value (depending on the type of target specified) across all targeted Pods, and produces a ratio used to scale the number of desired replicas.

  Please note that if some of the Pod's containers do not have the relevant resource request set, CPU utilization for the Pod will not be defined and the autoscaler will not take any action for that metric. See the algorithm details section below for more information about how the autoscaling algorithm works.

• For per-pod custom metrics, the controller functions similarly to per-pod resource metrics, except that it works with raw values, not utilization values.

• For object metrics and external metrics, a single metric is fetched, which describes the object in question. This metric is compared to the target value, to produce a ratio as above. In the autoscaling/v2beta2 API version, this value can optionally be divided by the number of Pods before the comparison is made.

The HorizontalPodAutoscaler normally fetches metrics from a series of aggregated APIs (metrics.k8s.io, custom.metrics.k8s.io, and external.metrics.k8s.io). The metrics.k8s.io API is usually provided by metrics-server, which needs to be launched separately. For more information about resource metrics, see Metrics Server.

See Support for metrics APIs for more details.

The autoscaler accesses corresponding scalable controllers (such as replication controllers, deployments, and replica sets) by using the scale sub-resource. Scale is an interface that allows you to dynamically set the number of replicas and examine each of their current states. More details on scale sub-resource can be found here.

Algorithm Details

From the most basic perspective, the Horizontal Pod Autoscaler controller operates on the ratio between desired metric value and current metric value:

desiredReplicas = ceil[currentReplicas * ( currentMetricValue / desiredMetricValue )]

For example, if the current metric value is 200m, and the desired value is 100m, the number of replicas will be doubled, since 200.0 / 100.0 == 2.0 If the current value is instead 50m, we'll halve the number of replicas, since 50.0 / 100.0 == 0.5. We'll skip scaling if the ratio is sufficiently close to 1.0 (within a globally-configurable tolerance, from the --horizontal-pod-autoscaler-tolerance flag, which defaults to 0.1).

When a targetAverageValue or targetAverageUtilization is specified, the currentMetricValue is computed by taking the average of the given metric across all Pods in the HorizontalPodAutoscaler's scale target. Before checking the tolerance and deciding on the final values, we take pod readiness and missing metrics into consideration, however.

All Pods with a deletion timestamp set (i.e. Pods in the process of being shut down) and all failed Pods are discarded.

If a particular Pod is missing metrics, it is set aside for later; Pods with missing metrics will be used to adjust the final scaling amount.

When scaling on CPU, if any pod has yet to become ready (i.e. it's still initializing) or the most recent metric point for the pod was before it became ready, that pod is set aside as well.

Due to technical constraints, the HorizontalPodAutoscaler controller cannot exactly determine the first time a pod becomes ready when determining whether to set aside certain CPU metrics. Instead, it considers a Pod "not yet ready" if it's unready and transitioned to unready within a short, configurable window of time since it started. This value is configured with the --horizontal-pod-autoscaler-initial-readiness-delay flag, and its default is 30 seconds. Once a pod has become ready, it considers any transition to ready to be the first if it occurred within a longer, configurable time since it started. This value is configured with the --horizontal-pod-autoscaler-cpu-initialization-period flag, and its default is 5 minutes.

The currentMetricValue / desiredMetricValue base scale ratio is then calculated using the remaining pods not set aside or discarded from above.

If there were any missing metrics, we recompute the average more conservatively, assuming those pods were consuming 100% of the desired value in case of a scale down, and 0% in case of a scale up. This dampens the magnitude of any potential scale.

Furthermore, if any not-yet-ready pods were present, and we would have scaled up without factoring in missing metrics or not-yet-ready pods, we conservatively assume the not-yet-ready pods are consuming 0% of the desired metric, further dampening the magnitude of a scale up.

After factoring in the not-yet-ready pods and missing metrics, we recalculate the usage ratio. If the new ratio reverses the scale direction, or is within the tolerance, we skip scaling. Otherwise, we use the new ratio to scale.

Note that the original value for the average utilization is reported back via the HorizontalPodAutoscaler status, without factoring in the not-yet-ready pods or missing metrics, even when the new usage ratio is used.

If multiple metrics are specified in a HorizontalPodAutoscaler, this calculation is done for each metric, and then the largest of the desired replica counts is chosen. If any of these metrics cannot be converted into a desired replica count (e.g. due to an error fetching the metrics from the metrics APIs) and a scale down is suggested by the metrics which can be fetched, scaling is skipped. This means that the HPA is still capable of scaling up if one or more metrics give a desiredReplicas greater than the current value.

Finally, right before HPA scales the target, the scale recommendation is recorded. The controller considers all recommendations within a configurable window choosing the highest recommendation from within that window. This value can be configured using the --horizontal-pod-autoscaler-downscale-stabilization flag, which defaults to 5 minutes. This means that scaledowns will occur gradually, smoothing out the impact of rapidly fluctuating metric values.

API Object

The Horizontal Pod Autoscaler is an API resource in the Kubernetes autoscaling API group. The current stable version, which only includes support for CPU autoscaling, can be found in the autoscaling/v1 API version.

The beta version, which includes support for scaling on memory and custom metrics, can be found in autoscaling/v2beta2. The new fields introduced in autoscaling/v2beta2 are preserved as annotations when working with autoscaling/v1.

When you create a HorizontalPodAutoscaler API object, make sure the name specified is a valid DNS subdomain name. More details about the API object can be found at HorizontalPodAutoscaler Object.

Support for Horizontal Pod Autoscaler in kubectl

Horizontal Pod Autoscaler, like every API resource, is supported in a standard way by kubectl. We can create a new autoscaler using kubectl create command. We can list autoscalers by kubectl get hpa and get detailed description by kubectl describe hpa. Finally, we can delete an autoscaler using kubectl delete hpa.

In addition, there is a special kubectl autoscale command for creating a HorizontalPodAutoscaler object. For instance, executing kubectl autoscale rs foo --min=2 --max=5 --cpu-percent=80 will create an autoscaler for replication set foo, with target CPU utilization set to 80% and the number of replicas between 2 and 5. The detailed documentation of kubectl autoscale can be found here.

Autoscaling during rolling update

Kubernetes lets you perform a rolling update on a Deployment. In that case, the Deployment manages the underlying ReplicaSets for you. When you configure autoscaling for a Deployment, you bind a HorizontalPodAutoscaler to a single Deployment. The HorizontalPodAutoscaler manages the replicas field of the Deployment. The deployment controller is responsible for setting the replicas of the underlying ReplicaSets so that they add up to a suitable number during the rollout and also afterwards.

If you perform a rolling update of a StatefulSet that has an autoscaled number of replicas, the StatefulSet directly manages its set of Pods (there is no intermediate resource similar to ReplicaSet).

Support for cooldown/delay

When managing the scale of a group of replicas using the Horizontal Pod Autoscaler, it is possible that the number of replicas keeps fluctuating frequently due to the dynamic nature of the metrics evaluated. This is sometimes referred to as thrashing.

Starting from v1.6, a cluster operator can mitigate this problem by tuning the global HPA settings exposed as flags for the kube-controller-manager component:

Starting from v1.12, a new algorithmic update removes the need for the upscale delay.

• --horizontal-pod-autoscaler-downscale-stabilization: Specifies the duration of the downscale stabilization time window. Horizontal Pod Autoscaler remembers the historical recommended sizes and only acts on the largest size within this time window. The default value is 5 minutes (5m0s).

Support for resource metrics

Any HPA target can be scaled based on the resource usage of the pods in the scaling target. When defining the pod specification the resource requests like cpu and memory should be specified. This is used to determine the resource utilization and used by the HPA controller to scale the target up or down. To use resource utilization based scaling specify a metric source like this:

type: Resource
resource:
  name: cpu
  target:
    type: Utilization
    averageUtilization: 60

With this metric the HPA controller will keep the average utilization of the pods in the scaling target at 60%. Utilization is the ratio between the current usage of resource to the requested resources of the pod. See Algorithm for more details about how the utilization is calculated and averaged.

Container Resource Metrics

HorizontalPodAutoscaler also supports a container metric source where the HPA can track the resource usage of individual containers across a set of Pods, in order to scale the target resource. This lets you configure scaling thresholds for the containers that matter most in a particular Pod. For example, if you have a web application and a logging sidecar, you can scale based on the resource use of the web application, ignoring the sidecar container and its resource use.

If you revise the target resource to have a new Pod specification with a different set of containers, you should revise the HPA spec if that newly added container should also be used for scaling. If the specified container in the metric source is not present or only present in a subset of the pods then those pods are ignored and the recommendation is recalculated. See Algorithm for more details about the calculation. To use container resources for autoscaling define a metric source as follows:

type: ContainerResource
containerResource:
  name: cpu
  container: application
  target:
    type: Utilization
    averageUtilization: 60

In the above example the HPA controller scales the target such that the average utilization of the cpu in the application container of all the pods is 60%.

Support for multiple metrics

Kubernetes 1.6 adds support for scaling based on multiple metrics. You can use the autoscaling/v2beta2 API version to specify multiple metrics for the Horizontal Pod Autoscaler to scale on. Then, the Horizontal Pod Autoscaler controller will evaluate each metric, and propose a new scale based on that metric. The largest of the proposed scales will be used as the new scale.

Support for custom metrics

Kubernetes 1.6 adds support for making use of custom metrics in the Horizontal Pod Autoscaler. You can add custom metrics for the Horizontal Pod Autoscaler to use in the autoscaling/v2beta2 API. Kubernetes then queries the new custom metrics API to fetch the values of the appropriate custom metrics.

See Support for metrics APIs for the requirements.

Support for metrics APIs

By default, the HorizontalPodAutoscaler controller retrieves metrics from a series of APIs. In order for it to access these APIs, cluster administrators must ensure that:

• The API aggregation layer is enabled.

• The corresponding APIs are registered:

  • For resource metrics, this is the metrics.k8s.io API, generally provided by metrics-server. It can be launched as a cluster addon.

  • For custom metrics, this is the custom.metrics.k8s.io API. It's provided by "adapter" API servers provided by metrics solution vendors. Check with your metrics pipeline, or the list of known solutions. If you would like to write your own, check out the boilerplate to get started.

  • For external metrics, this is the external.metrics.k8s.io API. It may be provided by the custom metrics adapters provided above.

For more information on these different metrics paths and how they differ please see the relevant design proposals for the HPA V2, custom.metrics.k8s.io and external.metrics.k8s.io.

For examples of how to use them see the walkthrough for using custom metrics and the walkthrough for using external metrics.

Support for configurable scaling behavior

Starting from v1.18 the v2beta2 API allows scaling behavior to be configured through the HPA behavior field. Behaviors are specified separately for scaling up and down in scaleUp or scaleDown section under the behavior field. A stabilization window can be specified for both directions which prevents the flapping of the number of the replicas in the scaling target. Similarly specifying scaling policies controls the rate of change of replicas while scaling.

Scaling Policies

One or more scaling policies can be specified in the behavior section of the spec. When multiple policies are specified the policy which allows the highest amount of change is the policy which is selected by default. The following example shows this behavior while scaling down:

behavior:
  scaleDown:
    policies:
    - type: Pods
      value: 4
      periodSeconds: 60
    - type: Percent
      value: 10
      periodSeconds: 60

periodSeconds indicates the length of time in the past for which the policy must hold true. The first policy (Pods) allows at most 4 replicas to be scaled down in one minute. The second policy (Percent) allows at most 10% of the current replicas to be scaled down in one minute.

Since by default the policy which allows the highest amount of change is selected, the second policy will only be used when the number of pod replicas is more than 40. With 40 or less replicas, the first policy will be applied. For instance if there are 80 replicas and the target has to be scaled down to 10 replicas then during the first step 8 replicas will be reduced. In the next iteration when the number of replicas is 72, 10% of the pods is 7.2 but the number is rounded up to 8. On each loop of the autoscaler controller the number of pods to be change is re-calculated based on the number of current replicas. When the number of replicas falls below 40 the first policy (Pods) is applied and 4 replicas will be reduced at a time.

The policy selection can be changed by specifying the selectPolicy field for a scaling direction. By setting the value to Min which would select the policy which allows the smallest change in the replica count. Setting the value to Disabled completely disables scaling in that direction.

Stabilization Window

The stabilization window is used to restrict the flapping of replicas when the metrics used for scaling keep fluctuating. The stabilization window is used by the autoscaling algorithm to consider the computed desired state from the past to prevent scaling. In the following example the stabilization window is specified for scaleDown.

scaleDown:
  stabilizationWindowSeconds: 300

When the metrics indicate that the target should be scaled down the algorithm looks into previously computed desired states and uses the highest value from the specified interval. In above example all desired states from the past 5 minutes will be considered.

Default Behavior

To use the custom scaling not all fields have to be specified. Only values which need to be customized can be specified. These custom values are merged with default values. The default values match the existing behavior in the HPA algorithm.

behavior:
  scaleDown:
    stabilizationWindowSeconds: 300
    policies:
    - type: Percent
      value: 100
      periodSeconds: 15
  scaleUp:
    stabilizationWindowSeconds: 0
    policies:
    - type: Percent
      value: 100
      periodSeconds: 15
    - type: Pods
      value: 4
      periodSeconds: 15
    selectPolicy: Max

For scaling down the stabilization window is 300 seconds (or the value of the --horizontal-pod-autoscaler-downscale-stabilization flag if provided). There is only a single policy for scaling down which allows a 100% of the currently running replicas to be removed which means the scaling target can be scaled down to the minimum allowed replicas. For scaling up there is no stabilization window. When the metrics indicate that the target should be scaled up the target is scaled up immediately. There are 2 policies where 4 pods or a 100% of the currently running replicas will be added every 15 seconds till the HPA reaches its steady state.

Example: change downscale stabilization window

To provide a custom downscale stabilization window of 1 minute, the following behavior would be added to the HPA:

behavior:
  scaleDown:
    stabilizationWindowSeconds: 60

Example: limit scale down rate

To limit the rate at which pods are removed by the HPA to 10% per minute, the following behavior would be added to the HPA:

behavior:
  scaleDown:
    policies:
    - type: Percent
      value: 10
      periodSeconds: 60

To ensure that no more than 5 Pods are removed per minute, you can add a second scale-down policy with a fixed size of 5, and set selectPolicy to minimum. Setting selectPolicy to Min means that the autoscaler chooses the policy that affects the smallest number of Pods:

behavior:
  scaleDown:
    policies:
    - type: Percent
      value: 10
      periodSeconds: 60
    - type: Pods
      value: 5
      periodSeconds: 60
    selectPolicy: Min

Example: disable scale down

The selectPolicy value of Disabled turns off scaling the given direction. So to prevent downscaling the following policy would be used:

behavior:
  scaleDown:
    selectPolicy: Disabled

Implicit maintenance-mode deactivation

You can implicitly deactivate the HPA for a target without the need to change the HPA configuration itself. If the target's desired replica count is set to 0, and the HPA's minimum replica count is greater than 0, the HPA stops adjusting the target (and sets the ScalingActive Condition on itself to false) until you reactivate it by manually adjusting the target's desired replica count or HPA's minimum replica count.

Migrating Deployments and StatefulSets to horizontal autoscaling

When an HPA is enabled, it is recommended that the value of spec.replicas of the Deployment and / or StatefulSet be removed from their manifest(s). If this isn't done, any time a change to that object is applied, for example via kubectl apply -f deployment.yaml, this will instruct Kubernetes to scale the current number of Pods to the value of the spec.replicas key. This may not be desired and could be troublesome when an HPA is active.

Keep in mind that the removal of spec.replicas may incur a one-time degradation of Pod counts as the default value of this key is 1 (reference Deployment Replicas. Upon the update, all Pods except 1 will begin their termination procedures. Any deployment application afterwards will behave as normal and respect a rolling update configuration as desired. You can avoid this degradation by choosing one of the following two methods based on how you are modifying your deployments:

• Client Side Apply (this is the default)
• Server Side Apply

What's next

• Design documentation: Horizontal Pod Autoscaling.
• kubectl autoscale command: kubectl autoscale.
• Usage example of Horizontal Pod Autoscaler.