Autoscaling Example

Prerequisites

• The cluster should have metric-server running in the kube-system namespace

• For Kind install ../../testing/scripts/metrics.yaml See https://github.com/kubernetes-sigs/kind/issues/398

• For Minikube run:

  minikube addons enable metrics-server

Setup Seldon Core

Use the setup notebook to Setup Cluster with Ambassador Ingress.

!kubectl create namespace seldon

Error from server (AlreadyExists): namespaces "seldon" already exists

Create model with v2beta1 autoscaler

To create a model with an HorizontalPodAutoscaler there are three steps:

1. Ensure you have a resource request for the metric you want to scale on if it is a standard metric such as cpu or memory, e.g.:

          resources:
            requests:
              cpu: '0.5'
     

1. Add an v2beta1 HPA Spec referring to this Deployment, e.g.:

    - hpaSpec:
        maxReplicas: 3
        minReplicas: 1
        metrics:
        - resource:
            name: cpu
            targetAverageUtilization: 10
          type: Resource

The full SeldonDeployment spec is shown below.

!pygmentize model_with_hpa_v2beta1.yaml

apiVersion: machinelearning.seldon.io/v1
kind: SeldonDeployment
metadata:
  name: seldon-model
spec:
  name: test-deployment
  predictors:
  - componentSpecs:
    - hpaSpec:
        maxReplicas: 3
        metrics:
        - resource:
            name: cpu
            targetAverageUtilization: 10
          type: Resource
        minReplicas: 1
      spec:
        containers:
        - image: seldonio/mock_classifier:1.19.0-dev
          imagePullPolicy: IfNotPresent
          name: classifier
          resources:
            requests:
              cpu: '0.5'
        terminationGracePeriodSeconds: 1
    graph:
      children: []
      name: classifier
      type: MODEL
    name: example

!kubectl create -f model_with_hpa_v2beta1.yaml -n seldon

seldondeployment.machinelearning.seldon.io/seldon-model created

!kubectl wait sdep/seldon-model \
  --for=condition=ready \
  --timeout=120s \
  -n seldon

seldondeployment.machinelearning.seldon.io/seldon-model condition met

Create Load

We label some nodes for the loadtester. We attempt the first two as for Kind the first node shown will be the master.

!kubectl label nodes $(kubectl get nodes -o jsonpath='{.items[0].metadata.name}') role=locust

node/kind-control-plane not labeled

!helm install loadtester ../../../helm-charts/seldon-core-loadtesting -n seldon  \
    --set locust.host=http://seldon-model-example:8000 \
    --set oauth.enabled=false \
    --set locust.hatchRate=1 \
    --set locust.clients=1 \
    --set loadtest.sendFeedback=0 \
    --set locust.minWait=0 \
    --set locust.maxWait=0 \
    --set replicaCount=1

NAME: loadtester
LAST DEPLOYED: Thu Dec  4 11:32:40 2025
NAMESPACE: seldon
STATUS: deployed
REVISION: 1
TEST SUITE: None

After a few mins you should see the deployment my-dep scaled to 3 deployments

import json
import time


def getNumberPods():
    dp = !kubectl get deployment -n seldon seldon-model-example-0-classifier -o json
    dp = json.loads("".join(dp))
    return dp["status"]["replicas"]


scaled = False
for i in range(60):
    pods = getNumberPods()
    print(pods)
    if pods > 1:
        scaled = True
        break
    time.sleep(5)
assert scaled

1
1
1
1
1
1
1
1
1
3

!kubectl get pods,deployments,hpa -n seldon

NAME                                                     READY   STATUS    RESTARTS   AGE
pod/locust-master-1-92n86                                1/1     Running   0          75s
pod/locust-slave-1-v4cjz                                 1/1     Running   0          75s
pod/seldon-model-example-0-classifier-7b5659c8f9-pcd22   2/2     Running   0          25s
pod/seldon-model-example-0-classifier-7b5659c8f9-vpr2p   2/2     Running   0          115s
pod/seldon-model-example-0-classifier-7b5659c8f9-wpbfm   2/2     Running   0          25s

NAME                                                READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/seldon-model-example-0-classifier   3/3     3            3           116s

NAME                                                                    REFERENCE                                      TARGETS        MINPODS   MAXPODS   REPLICAS   AGE
horizontalpodautoscaler.autoscaling/seldon-model-example-0-classifier   Deployment/seldon-model-example-0-classifier   cpu: 51%/10%   1         3         3          116s

!helm delete loadtester -n seldon

release "loadtester" uninstalled

!kubectl delete -f model_with_hpa_v2beta1.yaml -n seldon

seldondeployment.machinelearning.seldon.io "seldon-model" deleted

Create model with v2 autoscaler

To create a model with an HorizontalPodAutoscaler there are three steps:

1. Ensure you have a resource request for the metric you want to scale on if it is a standard metric such as cpu or memory, e.g.:

          resources:
            requests:
              cpu: '0.5'
     

1. Add an v2beta1 HPA Spec referring to this Deployment, e.g.:

    - hpaSpec:
        maxReplicas: 3
        minReplicas: 1
        metricsv2:
        - resource:
            name: cpu
            target:
              type: Utilization
              averageUtilization: 10
          type: Resource

The full SeldonDeployment spec is shown below.

!pygmentize model_with_hpa_v2.yaml

apiVersion: machinelearning.seldon.io/v1
kind: SeldonDeployment
metadata:
  name: seldon-model
spec:
  name: test-deployment
  predictors:
  - componentSpecs:
    - hpaSpec:
        maxReplicas: 3
        metricsv2:
        - resource:
            name: cpu
            target:
              type: Utilization
              averageUtilization: 10
          type: Resource
        minReplicas: 1
      spec:
        containers:
        - image: seldonio/mock_classifier:1.19.0-dev
          imagePullPolicy: IfNotPresent
          name: classifier
          resources:
            requests:
              cpu: '0.5'
        terminationGracePeriodSeconds: 1
    graph:
      children: []
      name: classifier
      type: MODEL
    name: example

!kubectl create -f model_with_hpa_v2.yaml -n seldon

seldondeployment.machinelearning.seldon.io/seldon-model created

!kubectl wait sdep/seldon-model \
  --for=condition=ready \
  --timeout=120s \
  -n seldon

seldondeployment.machinelearning.seldon.io/seldon-model condition met

Create Load

We label some nodes for the loadtester. We attempt the first two as for Kind the first node shown will be the master.

!kubectl label nodes $(kubectl get nodes -o jsonpath='{.items[0].metadata.name}') role=locust

node/kind-control-plane not labeled

!helm install loadtester ../../../helm-charts/seldon-core-loadtesting -n seldon  \
    --set locust.host=http://seldon-model-example:8000 \
    --set oauth.enabled=false \
    --set locust.hatchRate=1 \
    --set locust.clients=1 \
    --set loadtest.sendFeedback=0 \
    --set locust.minWait=0 \
    --set locust.maxWait=0 \
    --set replicaCount=1

NAME: loadtester
LAST DEPLOYED: Thu Dec  4 11:35:11 2025
NAMESPACE: seldon
STATUS: deployed
REVISION: 1
TEST SUITE: None

After a few mins you should see the deployment my-dep scaled to 3 deployments

import json
import time


def getNumberPods():
    dp = !kubectl get deployment -n seldon seldon-model-example-0-classifier -o json
    dp = json.loads("".join(dp))
    return dp["status"]["replicas"]


scaled = False
for i in range(60):
    pods = getNumberPods()
    print(pods)
    if pods > 1:
        scaled = True
        break
    time.sleep(5)
assert scaled

1
1
1
1
1
1
1
3

!kubectl get pods,deployments,hpa -n seldon

NAME                                                     READY   STATUS    RESTARTS   AGE
pod/locust-master-1-vztf6                                1/1     Running   0          59s
pod/locust-slave-1-98phc                                 1/1     Running   0          59s
pod/seldon-model-example-0-classifier-789cd4b649-47b5j   0/2     Running   0          18s
pod/seldon-model-example-0-classifier-789cd4b649-rkmz2   2/2     Running   0          93s
pod/seldon-model-example-0-classifier-789cd4b649-wrhm9   0/2     Running   0          17s

NAME                                                READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/seldon-model-example-0-classifier   1/3     3            1           93s

NAME                                                                    REFERENCE                                      TARGETS        MINPODS   MAXPODS   REPLICAS   AGE
horizontalpodautoscaler.autoscaling/seldon-model-example-0-classifier   Deployment/seldon-model-example-0-classifier   cpu: 51%/10%   1         3         3          93s

!helm delete loadtester -n seldon

release "loadtester" uninstalled

!kubectl delete -f model_with_hpa_v2.yaml -n seldon

seldondeployment.machinelearning.seldon.io "seldon-model" deleted