Skip to main content
Version: 13.x (Current)

Replicas

What are Replicas for?​

In this section you can define a range of desired replicas number for your services designed to automatically scale them based on CPU usage.
The feature is thought to generate the replicas only for the productions environments where the Environment Variable ENABLE_HPA is set to true.

caution

Based on your release pipeline, the ENABLE_HPA environment variable could need the environment prefix (e.g. PROD_) or not.

danger

The replicas configured will not work properly if the static replicas is set to 0.

Setting both static replicas and HPA Replicas may lead to undefined behaviors

Configure Replicas​

Replicas landing page

Upon accessing the Console Design area and selecting the Replicas section in the left menu, you'll see the list of your services for which you can set Replicas.

caution

For supporting Replicas a service must have both the minimum and the maximum of CPU limit set

In the table rows, for each service, you can directly change the following parameters:

  • Min Replica: the minimum number of replicas that must be running.
  • Max Replica: the maximum number of replicas that can be concurrent running.
  • CPU Threshold: this number represents the percentage that kubernetes should consider to trigger a pod scale, for further details about the algorithm used by Kubernetes Pod Autoscaler check out the Kubernetes documentation page
  • Generate: enable this flag to actually creating the replicas configuration when saving the configuration.

How it works​

info

This process is entrusted to the Horizontal Pod Autoscaler Kubernetes resource

The Kubernetes Horizontal Pod Autoscaler uses the metrics collected by Kubernetes itself to evaluate the need for instantiation of new pods to divide the work up to a maximum replication number.

When the total current CPU utilization by the replicas set exceeds the CPU Threshold, if the Max Replica number allows it, a new replica is scheduled for creation.
If the total CPU usage drops below the CPU Threshold, after a period of time useless replicas are removed.

How to scale services manually​

You can also scale services manually writing yourself the configuration.

You have to enable the horizontal auto-scaling of the services managed by the Console, setting the environment variable ENABLE_HPA to true for the production environments. Then create the hpa files in configuration folder, inside the configurations/{environment_name} directory. The horizontal auto-scaling is entrusted to HPA Kubernetes resource.

info

Check out this page for manually set the number of replicas of a Platform Services

The resource that is taken into consideration is the use of CPU. If the value ​​exceeds the threshold expressed as a percentage, the system will create a new instance of the pod to divide the load. The instance will come deleted after a period of time when all available instances remain below the set threshold.

This percentage is calculated with this expression:

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

Example​

Service name: bff-service

HPA file path: configurations/configuration/PROD/bff-service.hpa.yml

Content:

apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
  name: bff-service
  annotations:
    mia-platform.eu/deployment-date: '{{RELEASE_DATE}}'
    mia-platform.eu/commit-author: '{{AUTHOR_EMAIL}}'
    mia-platform.eu/commit-sha: '{{COMMIT_SHA}}'
    mia-platform.eu/version: 6.3.0
  labels:
    app: bff-service
    app.kubernetes.io/name: bff-service
    app.kubernetes.io/component: custom
    app.kubernetes.io/part-of: project
    app.kubernetes.io/managed-by: mia-platform

spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: bff-service
  minReplicas: 2
  maxReplicas: 4
  targetCPUUtilizationPercentage: 85
info

For more information concerning the Kubernetes feature check out the Kubernetes documentation page