Skip to content

Another Autoscaler is a Kubernetes controller that automatically starts, stops, or restarts pods from a deployment at a specified time using a cron expression.

License

dignajar/another-autoscaler

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

45 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Another Autoscaler

Another Autoscaler is a Kubernetes controller that automatically starts, stops, or restarts pods from a deployment at a specified time using a cron syntax.

Another Autoscaler read the annotation of each deployment and performs an increase or decrease in the number of replicas.

Docker image Kubernetes YAML manifests codebeat badge release license

The date and time must be in UTC.

The restart feature execute a rollout restart deployment.

Use cases

  • Cost savings by reducing the number of replicas after working hours or weekends.
  • Stop GPU deployments during non-working hours.

Another Autoscaler is a perfect combination with Cluster Autoscaler.

Installation

# Deploy Another Autoscaler into Kubernetes on "another" namespace
kubectl apply -f https://raw.githubusercontent.com/dignajar/another-autoscaler/master/kubernetes/full.yaml

# Check if Another Autoscaler is working
kubectl get pods -n another

Configuration

The following annotations for the deployments are valid (metadata.annotations).

  • another-autoscaler/stop-time: Define the date and time when the replica of the deployment will be set to 0.
  • another-autoscaler/start-time Define the date and time when the replica of the deployment will be set to 1.
  • another-autoscaler/restart-time:: Define the date and time when the rollout restart will be peformerd to a deployment.
  • another-autoscaler/stop-replicas: This is the number of replicas to set when Another Autoscaler scale down the deployment, by default is 0.
  • another-autoscaler/start-replicas: This is the number of replicas to set when Another Autoscaler scale up the deployment, by default is 1.

Examples

Stop pods at 6pm every day:

another-autoscaler/stop-time: "00 18 * * *"

Start pods at 1pm every day:

another-autoscaler/start-time: "00 13 * * *"

Start 3 pods at 2:30pm every day:

another-autoscaler/start-time: "30 14 * * *"
another-autoscaler/start-replicas: "3"

Restart pods at 9:15am every day:

another-autoscaler/restart-time: "15 09 * * *"

Restart pods at 2:30am, only on Saturday and Sunday:

another-autoscaler/restart-time: "00 02 * * 0,6"

Full example, how to start pods at 2pm and stop them at 3pm every day

The following example start 5 replicas in total at 2pm and stop 4 of them at 3pm every day, the deployment start with 0 replicas.

The start-replicas is not incremental, the value is the number of replicas will be setup by Another Autoscaler at the defined time by start-time.

The date and time must be in UTC.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx
  labels:
    app: nginx
  annotations:
    another-autoscaler/start-time: "00 14 * * *"
    another-autoscaler/start-replicas: "5"
    another-autoscaler/stop-time: "00 15 * * *"
    another-autoscaler/stop-replicas: "1"
spec:
  replicas: 0
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: nginx:latest
        ports:
        - containerPort: 80

GitOps / FluxCD

To avoid conflicts with Flux and Another Autoscaler, you can remove the field spec.replicas from your deployment manifest and leave Another Autoscaler to manage the number of replicas.