GlobalPodAutoscaler (GPA) is a Kubernetes extension designed to provide high availability for your deployments by enabling automatic scaling of pods across multiple clusters. GPA actively scales deployments on a primary cluster while maintaining a backup cluster to ensure uninterrupted service in the event of resource outages on the primary cluster.
- Automatically scale pods across multiple clusters.
- Active scaling on the primary cluster with a secondary cluster as a backup.
- Seamless failover to secondary cluster in case of resource outages on the primary cluster.
- Automatic recovery of deployments from the secondary cluster back to the primary cluster once resources become available.
Global pod autoscaler is deployed in primary cluster and watches the state of the deployment. The state of the deployment is reported back to KEDA every 5 second. Here is the KEDA scaler to talk to global pod autoscaler.
#examples/external-scalar.yaml
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
name: scaler-name
spec:
scaleTargetRef:
name: secondary-deployment-name
minReplicaCount: 1
maxReplicaCount: 20
pollingInterval: 1
triggers:
- type: external-push
metadata:
scalerAddress: global-pod-autoscaler-service-link
governer: main-cluster-deployment-to-watch
namespace: main-cluster-deployment-namespaceKEDA is reported back the number of pods required every 5 second and KEDA provisions the necessary HPA to scale the secondary deployment.
In theory then you can set up multiple cluster and create a chain of deployments to make sure you are always up and running.
Create the necessary RBAC required to run GlobalPodAutoscaler
kubectl apply -f ./examples/rbac.yaml
Deploy the global pod autoscaler
kubectl apply -f ./examples/global-pod-autoscaler.yaml
We welcome contributions to the GlobalPodAutoscaler project. Please feel free to submit issues, fork the repository, and send pull requests.
Read more about KEDA Extenal Scaler here.
This project is licensed under the MIT License.