To avoid resource wasting in some scenarios, Container Service for Kubernetes (ACK) provides the kubernetes-cronhpa-controller component to automatically scale resources based on predefined schedules. This topic describes how to use Cron Horizontal Pod Autoscaler (CronHPA) to scale your workloads based on a schedule. This topic also describes how to enable CronHPA and Horizontal Pod Autoscaler (HPA) to interact without conflicts.

Prerequisites

Background information

kubernetes-cronhpa-controller is a Kubernetes HPA controller that can scale pods in a Kubernetes cluster based on a schedule that is similar to a crontab. You can use CronHPA with Kubernetes objects whose subresources can be scaled. The objects include Deployments and StatefulSets. kubernetes-cronhpa-controller is an open source project on GitHub. For more information, see kubernetes-cronhpa-controller.

The following table describes the parameters in the CronHPA configuration.
apiVersion: autoscaling.alibabacloud.com/v1beta1
kind: CronHorizontalPodAutoscaler
metadata:
  labels:
    controller-tools.k8s.io: "1.0"
  name: cronhpa-sample
  namespace: default 
spec:
   scaleTargetRef:
      apiVersion: apps/v1
      kind: Deployment
      name: nginx-deployment-basic
   excludeDates:
   # exclude November 15th
   - "* * * 15 11 *"
   # exclude every Friday 
   - "* * * * * 5"
   jobs:
   - name: "scale-down"
     schedule: "30 */1 * * * *"
     targetSize: 1
   - name: "scale-up"
     schedule: "0 */1 * * * *"
     targetSize: 3
     runOnce: false
ParameterDescription
scaleTargetRef scaleTargetRef specifies the object that you want to scale. If the subresources of the object can be scaled, you can enable CronHPA for the object.
excludeDatesThe value of excludeDates must be an array of dates. Scaling jobs are not executed on the dates that are specified in excludeDates.
Note The minimum time period is one day.

The value is in the "* * * * * *" format, which indicates "<seconds> <minutes> <hours> <day of month> <month> <day of week>".

For example, if you do not want to execute scaling jobs on November 15, set excludeDates to the following value:
excludeDates:
  - "* * * 15 11 *"
jobsYou can set multiple CronHPA jobs in the spec section. You can set the following parameters for each CronHPA job:
  • name: Names are used to distinguish CronHPA jobs. Therefore, the name of each CronHPA job must be unique in the CronHPA configuration.
  • schedule: the scaling schedule, which is similar to a crontab. kubernetes-cronhpa-controller uses a Golang library to support a variety of rules. For more information, see go-cron. The format of cron expressions must conform to the following rules. Otherwise, you cannot create cron expressions.
    Create cron expressions based on the following rules:
    Field name   | Mandatory? | Allowed values  | Allowed special characters
      ----------   | ---------- | --------------  | --------------------------
      Seconds      | Yes        | 0-59            | * / , -
      Minutes      | Yes        | 0-59            | * / , -
      Hours        | Yes        | 0-23            | * / , -
      Day of month | Yes        | 1-31            | * / , - ?
      Month        | Yes        | 1-12 or JAN-DEC | * / , -
      Day of week  | Yes        | 0-6 or SUN-SAT  | * / , - ?
  • targetSize: the number of pods to which you want to scale at the scheduled time.
  • runOnce: If you set runOnce to true, the job is executed only once. The job exits after one run.

Install the CronHPA controller

You can install the CronHPA controller ack-kubernetes-cronhpa-controller by using one of the following methods.

  1. Log on to the ACK console and click Clusters in the left-side navigation pane.
  2. On the Clusters page, click the name of a cluster and choose Operations > Add-ons in the left-side navigation pane.
  3. On the Add-ons page, click the Manage Applications tab, find ack-kubernetes-cronhpa-controller, and then click Install.

You can uninstall the CronHPA controller if CronHPA is no longer used. For more information about how to uninstall ack-kubernetes-cronhpa-controller, see Manage system components or Manage releases by using Helm.

Create CronHPA jobs

Before you create and run CronHPA jobs for your application, make sure that the CronHPA controller runs as normal in your cluster and only one HPA task is created for your application. For more information about how to enable CronHPA and HPA to interact without conflicts, see Enable CronHPA and HPA to interact without conflicts. You can create CronHPA jobs in the following scenarios.

Scenario 1: Create CronHPA jobs when you create an application

In the Scaling section on the Advanced wizard page, select Enable on the right side of CronHPA to create CronHPA jobs for the application. For more information about how to create an application, see Create a stateless application by using a Deployment or Use a StatefulSet to create a stateful application. cronhpa
The ACK console automatically checks whether the CronHPA controller is installed in the cluster. If the CronHPA controller is not installed, the Install button appears on the page. After the CronHPA controller is installed, the CronHPA parameters appear on the page. The following table describes the parameters.
ParameterDescription
Job NameEnter a name for the CronHPA job. The name of each CronHPA job must be unique.
Desired Number of ReplicasReplicated pods are scaled to the desired number at the scheduled time.
Scaling ScheduleSet the scaling schedule.

For more information about how to set the scaling schedule for a CronHPA job, see AliyunContainerService/kubernetes-cronhpa-controller.

Scenario 2: Create CronHPA jobs for an existing application

The following example demonstrates how to create CronHPA jobs for an existing application. A stateless application is used in this example.

  1. Log on to the ACK console and click Clusters in the left-side navigation pane.
  2. On the Clusters page, click the name of a cluster and choose Workloads > Deployments in the left-side navigation pane.
  3. On the Deployments page, find the application that you want to manage and click Details in the Actions column.
  4. Click the Pod Scaling tab and configure CronHPA jobs.
    • If the CronHPA controller is not installed, the Install button appears on the page. Click Install and perform the following steps.
    • If the CronHPA controller is installed, perform the following steps.
  5. Click Create on the right side of CronHPA. In the Create dialog box, configure CronHPA jobs. create
    The following table describes the parameters.
    ParameterDescription
    Job NameEnter a name for the CronHPA job. The name of each CronHPA job must be unique.
    Desired Number of ReplicasReplicated pods are scaled to the desired number at the scheduled time.
    Scaling ScheduleSet the scaling schedule.

    For more information about how to set the scaling schedule for a CronHPA job, see AliyunContainerService/kubernetes-cronhpa-controller.

Create or modify CronHPA jobs

  1. Go to the Pod Scaling tab by performing the steps that are described in the preceding Create CronHPA jobs section.
  2. On the Pod Scaling tab, find the CronHPA job that you created in the CronHPA section and click Add or Modify Job in the Actions column.
  3. In the Edit dialog box, click Add Job to create CronHPA jobs. You can also modify existing CronHPA jobs. Then, click OK.
    modifyTo delete CronHPA jobs, perform the following steps:
    In the Edit dialog box, click the delete icon in the upper-right corner of the job that you want to delete. Then, click OK. delete

Templates of CronHPA and HPA

CronHPA

apiVersion: autoscaling.alibabacloud.com/v1beta1
kind: CronHorizontalPodAutoscaler
metadata:
  labels:
    controller-tools.k8s.io: "1.0"
  name: cronhpa-sample
spec:
   scaleTargetRef:
      apiVersion: apps/v1
      kind: Deployment
      name: nginx-deployment-basic
   jobs:
   - name: "scale-down"
     schedule: "30 */1 * * * *"
     targetSize: 1
   - name: "scale-up"
     schedule: "0 */1 * * * *"
     targetSize: 11    
    

HPA

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: php-apache
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: nginx-deployment-basic
  minReplicas: 4
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50    
    

Enable CronHPA and HPA to interact without conflicts

CronHPA triggers horizontal scaling for containers based on schedules. HPA is used to scale pods to ensure the availability of your applications when network traffic spikes. If ACK detects that both CronHPA and HPA are deployed, ACK sets HPA as the scaling object of CronHPA. CronHPA triggers HPA to scale pods at the scheduled time.

The configurations of CronHPA and HPA indicate the following information:
  • The scaleTargetRef field is used in the configurations of both CronHPA and HPA to specify the object to be scaled.
  • The crontab rules in the jobs section of the CronHPA configuration specify the number to which pods are scaled at the scheduled time.
  • HPA triggers scaling activities based on resource usage.

If both CronHPA and HPA are deployed, CronHPA and HPA may scale pods for the same application that is specified by scaleTargetRef. CronHPA and HPA are independent and unaware of each other. As a result, the CronHPA controller and the HPA controller scale pods for the application separately. The later scaling activity overwrites the earlier one.

Solution

The reason for the conflict between CronHPA and HPA is that the CronHPA controller and the HPA controller are unaware of each other. To resolve the conflict, CronHPA needs only to detect the status of HPA. In the HPA configuration, scaleTargetRef specifies the Deployment that is managed by HPA. The Deployment is related to a ReplicaSet. When HPA scales the Deployment, the ReplicaSet scales pods to the desired number. ACK modifies the CronHPA configuration by setting scaleTargetRef to HPA. CronHPA can find the application that is specified by scaleTargetRef in the HPA configuration. This enables CronHPA to detect the state of HPA.

cronhpa
The following YAML template shows the configurations that enable CronHPA and HPA to interact without conflicts:
apiVersion: autoscaling.alibabacloud.com/v1beta1
kind: CronHorizontalPodAutoscaler
metadata:
  labels:
    controller-tools.k8s.io: "1.0"
  name: cronhpa-sample
spec:
   scaleTargetRef:
      apiVersion: autoscaling/v1
      kind: HorizontalPodAutoscaler
      name:  nginx-deployment-basic-hpa
   jobs:
   - name: "scale-down"
     schedule: "30 */1 * * * *"
     targetSize: 1
     runOnce: false
   - name: "scale-up"
     schedule: "0 */1 * * * *"
     targetSize: 3
     runOnce: false

After you deploy the preceding YAML template, CronHPA is aware of the values of minReplicas, maxReplicas, and desiredReplicas in the HPA configuration. CronHPA is also aware of the current number of pods provisioned for the application that is specified by scaleTargetRef. CronHPA can detect the status of HPA by modifying the HPA configurations. CronHPA compares the desired number of pods with the current number of pods and then determines whether to trigger scaling activities and change the maximum number of pods in the HPA configuration. CronHPA also compares the desired number of pods with the maximum and minimum numbers of pods that are specified in the HPA configuration and then determines whether to change the minimum numbers of pods in the HPA configuration.

The following table describes the rules that enable CronHPA and HPA to interact without conflicts.
HPA (min/max)CronhpaDeploymentScaling resultDescription
1/1055
  • HPA (min/max): 1/10
  • Deployment: 5
If the number of pods desired by CronHPA equals the current number of pods, CronHPA does not change the maximum and minimum numbers of pods in the HPA configuration. In addition, no scaling activity is triggered.
1/1045
  • HPA (min/max): 1/10
  • Deployment: 5
If the number of pods desired by CronHPA is smaller than the current number of pods, no scaling activity is triggered.
1/1065
  • HPA (min/max): 6/10
  • Deployment: 6
  • If the number of pods desired by CronHPA is greater than the current number of pods, CronHPA adds pods to reach the desired number.
  • If the number of pods desired by CronHPA is greater than the value of minReplicas in the HPA configuration, CronHPA changes the value of minReplicas.
5/1045
  • HPA (min/max): 4/10
  • Deployment: 5
  • If the number of pods desired by CronHPA is smaller than the current number of pods, no scaling activity is triggered.
  • If the number of pods desired by CronHPA is smaller than the value of minReplicas in the HPA configuration, CronHPA changes the value of minReplicas.
5/10115
  • HPA (min/max): 11/11
  • Deployment: 11
  • If the number of pods desired by CronHPA is greater than the current number of pods, CronHPA adds pods to reach the desired number.
  • If the number of pods desired by CronHPA is greater than the value of maxReplicas in the HPA configuration, CronHPA changes the value of maxReplicas.
The following list describes the parameters in the table:
  • HPA (min/max): the minimum and maximum numbers of pods that are specified in the HPA configuration.
  • CronHPA: the desired number of pods that are specified in the CronHPA configuration.
  • Deployment: the current number of pods that are provisioned for the application.

CronHPA does not directly change the number of pods for the Deployment. It triggers HPA to scale the pods. This resolves the conflict between CronHPA and HPA.

FAQ

How do I uninstall ack-kubernetes-cronhpa-controller?

  1. Log on to the ACK console and click Clusters in the left-side navigation pane.
  2. On the Clusters page, click the name of a cluster and choose Applications > Helm in the left-side navigation pane.
  3. On the Helm page, check whether the ack-kubernetes-cronhpa-controller component is displayed.
    • If the ack-kubernetes-cronhpa-controller component is displayed, click Delete in the Actions column.
      Important After you delete the component on the Helm page, the related CustomResourceDefinition (CRD) objects and tasks are retained. You must manually delete the CRD objects and tasks.
    • If the ack-kubernetes-cronhpa-controller component is not displayed, go to the next step.
  4. In the left-side navigation pane of the details page, choose Operations > Add-ons.
  5. On the Add-ons page, find the ack-kubernetes-cronhpa-controller component and click Uninstall.