Overview

A horizontal pod autoscaler, defined by a HorizontalPodAutoscaler object, specifies how the system should automatically increase or decrease the scale of a replication controller or deployment configuration, based on metrics collected from the pods that belong to that replication controller or deployment configuration.

Requirements for Using Horizontal Pod Autoscalers

In order to use horizontal pod autoscalers, your cluster administrator must have properly configured cluster metrics.

Supported Metrics

The following metrics are supported by horizontal pod autoscalers:

Table 1. Metrics
Metric Description API version

CPU utilization

Percentage of the requested CPU

autoscaling/v1, autoscaling/v2beta1

Memory utilization

Percentage of the requested memory.

autoscaling/v2beta1

Autoscaling

You can create a horizontal pod autoscaler with the oc autoscale command and specify the minimum and maximum number of pods you want to run, as well as the CPU utilization or memory utilization your pods should target.

Autoscaling for Memory Utilization is a Technology Preview feature only.

After a horizontal pod autoscaler is created, it begins attempting to query Heapster for metrics on the pods. It may take one to two minutes before Heapster obtains the initial metrics.

After metrics are available in Heapster, the horizontal pod autoscaler computes the ratio of the current metric utilization with the desired metric utilization, and scales up or down accordingly. The scaling will occur at a regular interval, but it may take one to two minutes before metrics make their way into Heapster.

For replication controllers, this scaling corresponds directly to the replicas of the replication controller. For deployment configurations, scaling corresponds directly to the replica count of the deployment configuration. Note that autoscaling applies only to the latest deployment in the Complete phase.

OKD automatically accounts for resources and prevents unnecessary autoscaling during resource spikes, such as during start up. Pods in the unready state have 0 CPU usage when scaling up and the autoscaler ignores the pods when scaling down. Pods without known metrics have 0% CPU usage when scaling up and 100% CPU when scaling down. This allows for more stability during the HPA decision. To use this feature, you must configure readiness checks to determine if a new pod is ready for use.

Autoscaling for CPU Utilization

Use the oc autoscale command and specify at least the maximum number of pods you want to run at any given time. You can optionally specify the minimum number of pods and the average CPU utilization your pods should target, otherwise those are given default values from the OKD server.

For example:

$ oc autoscale dc/frontend --min 1 --max 10 --cpu-percent=80
deploymentconfig "frontend" autoscaled

The above example creates a horizontal pod autoscaler with the following definition when using the autoscaling/v1 version of the horizontal pod autoscaler:

Example 1. Horizontal Pod Autoscaler Object Definition
apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
  name: frontend (1)
spec:
  scaleTargetRef:
    kind: DeploymentConfig (2)
    name: frontend (3)
    apiVersion: apps/v1 (4)
    subresource: scale
  minReplicas: 1 (5)
  maxReplicas: 10 (6)
  cpuUtilization:
    targetCPUUtilizationPercentage: 80 (7)
1 The name of this horizontal pod autoscaler object
2 The kind of object to scale
3 The name of the object to scale
4 The API version of the object to scale
5 The minimum number of replicas to which to scale down
6 The maximum number of replicas to which to scale up
7 The percentage of the requested CPU that each pod should ideally be using

Alternatively, the oc autoscale command creates a horizontal pod autoscaler with the following definition when using the v2beta1 version of the horizontal pod autoscaler:

apiVersion: autoscaling/v2beta1
kind: HorizontalPodAutoscaler
metadata:
  name: hpa-resource-metrics-cpu (1)
spec:
  scaleTargetRef:
    apiVersion: apps/v1 (2)
    kind: ReplicationController (3)
    name: hello-hpa-cpu (4)
  minReplicas: 1 (5)
  maxReplicas: 10 (6)
  metrics:
  - type: Resource
    resource:
      name: cpu
      targetAverageUtilization: 50 (7)
1 The name of this horizontal pod autoscaler object
2 The API version of the object to scale
3 The kind of object to scale
4 The name of the object to scale
5 The minimum number of replicas to which to scale down
6 The maximum number of replicas to which to scale up
7 The average percentage of the requested CPU that each pod should be using

Autoscaling for Memory Utilization

Autoscaling for Memory Utilization is a Technology Preview feature only.

Unlike CPU-based autoscaling, memory-based autoscaling requires specifying the autoscaler using YAML instead of using the oc autoscale command. Optionally, you can specify the minimum number of pods and the average memory utilization your pods should target as well, otherwise those are given default values from the OKD server.

  1. Memory-based autoscaling is only available with the v2beta1 version of the autoscaling API. Enable memory-based autoscaling by adding the following to your cluster’s master-config.yaml file:

    ...
    apiServerArguments:
      runtime-config:
      - apis/autoscaling/v2beta1=true
    ...
  2. Place the following in a file, such as hpa.yaml:

    apiVersion: autoscaling/v2beta1
    kind: HorizontalPodAutoscaler
    metadata:
      name: hpa-resource-metrics-memory (1)
    spec:
      scaleTargetRef:
        apiVersion: apps/v1 (2)
        kind: ReplicationController (3)
        name: hello-hpa-memory (4)
      minReplicas: 1 (5)
      maxReplicas: 10 (6)
      metrics:
      - type: Resource
        resource:
          name: memory
          targetAverageUtilization: 50 (7)
    1 The name of this horizontal pod autoscaler object
    2 The API version of the object to scale
    3 The kind of object to scale
    4 The name of the object to scale
    5 The minimum number of replicas to which to scale down
    6 The maximum number of replicas to which to scale up
    7 The average percentage of the requested memory that each pod should be using
  3. Then, create the autoscaler from the above file:

    $ oc create -f hpa.yaml

For memory-based autoscaling to work, memory usage must increase and decrease proportionally to the replica count. On average:

  • An increase in replica count must lead to an overall decrease in memory (working set) usage per-pod.

  • A decrease in replica count must lead to an overall increase in per-pod memory usage.

Use the OpenShift web console to check the memory behavior of your application and ensure that your application meets these requirements before using memory-based autoscaling.

Viewing a Horizontal Pod Autoscaler

To view the status of a horizontal pod autoscaler:

  • Use the oc get command to view information on the CPU utilization and pod limits:

    $ oc get hpa/hpa-resource-metrics-cpu
    NAME                         REFERENCE                                 TARGET    CURRENT  MINPODS        MAXPODS    AGE
    hpa-resource-metrics-cpu     DeploymentConfig/default/frontend/scale   80%       79%      1              10         8d

    The output includes the following:

    • Target. The targeted average CPU utilization across all pods controlled by the deployment configuration.

    • Current. The current CPU utilization across all pods controlled by the deployment configuration.

    • Minpods/Maxpods. The minimum and maximum number of replicas that can be set by the autoscaler.

  • Use the oc describe command for detailed information on the horizontal pod autoscaler object.

    $ oc describe hpa/hpa-resource-metrics-cpu
    Name:                           hpa-resource-metrics-cpu
    Namespace:                      default
    Labels:                         <none>
    CreationTimestamp:              Mon, 26 Oct 2015 21:13:47 -0400
    Reference:                      DeploymentConfig/default/frontend/scale
    Target CPU utilization:         80% (1)
    Current CPU utilization:        79% (2)
    Min replicas:                   1 (3)
    Max replicas:                   4 (4)
    ReplicationController pods:     1 current / 1 desired
    Conditions: (5)
      Type                  Status  Reason                  Message
      ----                  ------  ------                  -------
      AbleToScale           True    ReadyForNewScale        the last scale time was sufficiently old as to warrant a new scale
      ScalingActive         True    ValidMetricFound        the HPA was able to successfully calculate a replica count from pods metric http_requests
      ScalingLimited        False   DesiredWithinRange      the desired replica count is within the acceptable range
    Events:
    1 The average percentage of the requested memory that each pod should be using.
    2 The current CPU utilization across all pods controlled by the deployment configuration.
    3 The minimum number of replicas to scale down to.
    4 The maximum number of replicas to scale up to.
    5 If the object used the v2alpha1 API, status conditions are displayed.

Viewing Horizontal Pod Autoscaler Status Conditions

You can use the status conditions set to determine whether or not the horizontal pod autoscaler is able to scale and whether or not it is currently restricted in any way.

The horizontal pod autoscaler status conditions are available with the v2beta1 version of the autoscaling API:

kubernetesMasterConfig:
  ...
  apiServerArguments:
    runtime-config:
    - apis/autoscaling/v2beta1=true

The following status conditions are set:

  • AbleToScale indicates whether the horizontal pod autoscaler is able to fetch and update scales, and whether any backoff conditions are preventing scaling.

    • A True condition indicates scaling is allowed.

    • A False condition indicates scaling is not allowed for the reason specified.

  • ScalingActive indicates whether the horizontal pod autoscaler is enabled (the replica count of the target is not zero) and is able to calculate desired scales.

    • A True condition indicates metrics is working properly.

    • A False condition generally indicates a problem with fetching metrics.

  • ScalingLimited indicates that autoscaling is not allowed because a maximum or minimum replica count was reached.

    • A True condition indicates that you need to raise or lower the minimum or maximum replica count in order to scale.

    • A False condition indicates that the requested scaling is allowed.

If you need to add or edit this line, restart the OKD services:

# systemctl restart origin-master-api origin-master-controllers

To see the conditions affecting a horizontal pod autoscaler, use oc describe hpa. Conditions appear in the status.conditions field:

$ oc describe hpa cm-test
Name:                           cm-test
Namespace:                      prom
Labels:                         <none>
Annotations:                    <none>
CreationTimestamp:              Fri, 16 Jun 2017 18:09:22 +0000
Reference:                      ReplicationController/cm-test
Metrics:                        ( current / target )
  "http_requests" on pods:      66m / 500m
Min replicas:                   1
Max replicas:                   4
ReplicationController pods:     1 current / 1 desired
Conditions: (1)
  Type                  Status  Reason                  Message
  ----                  ------  ------                  -------
  AbleToScale       True      ReadyForNewScale    the last scale time was sufficiently old as to warrant a new scale
  ScalingActive     True      ValidMetricFound    the HPA was able to successfully calculate a replica count from pods metric http_request
  ScalingLimited    False     DesiredWithinRange  the desired replica count is within the acceptable range
Events:
1 The horizontal pod autoscaler status messages.
  • The AbleToScale condition indicates whether HPA is able to fetch and update scales, as well as whether any backoff-related conditions would prevent scaling.

  • The ScalingActive condition indicates whether the HPA is enabled (for example, the replica count of the target is not zero) and is able to calculate desired scales. A`False` status generally indicates problems with fetching metrics.

  • The ScalingLimited condition indicates that the desired scale was capped by the maximum or minimum of the horizontal pod autoscaler. A`False` status generally indicates that you might need to raise or lower the minimum or maximum replica count constraints on your horizontal pod autoscaler.

The following is an example of a pod that is unable to scale:

Conditions:
  Type           Status    Reason            Message
  ----           ------    ------            -------
  AbleToScale    False     FailedGetScale    the HPA controller was unable to get the target's current scale: replicationcontrollers/scale.extensions "hello-hpa-cpu" not found

The following is an example of a pod that could not obtain the needed metrics for scaling:

Conditions:
  Type                  Status    Reason                    Message
  ----                  ------    ------                    -------
  AbleToScale           True     SucceededGetScale          the HPA controller was able to get the target's current scale
  ScalingActive         False    FailedGetResourceMetric    the HPA was unable to compute the replica count: unable to get metrics for resource cpu: no metrics returned from heapster

The following is an example of a pod where the requested autoscaling was less than the required minimums:

Conditions:
  Type              Status    Reason              Message
  ----              ------    ------              -------
  AbleToScale       True      ReadyForNewScale    the last scale time was sufficiently old as to warrant a new scale
  ScalingActive     True      ValidMetricFound    the HPA was able to successfully calculate a replica count from pods metric http_request
  ScalingLimited    False     DesiredWithinRange  the desired replica count is within the acceptable range
Events: