test: end-to-end demo with KServe ModelCar (#5)

Signed-off-by: tarilabs <[email protected]>

.github/workflows/e2e.yaml

@@ -37,6 +37,35 @@ jobs:
     - name: Run E2E tests
       run: |
         make test-e2e-skopeo
+    - name: Continue E2E by deploying KServe
+      run: |
+        kubectl apply -f https://github.com/cert-manager/cert-manager/releases/download/v1.16.1/cert-manager.yaml
+        e2e/repeat.sh kubectl apply --server-side -f https://github.com/kserve/kserve/releases/download/v0.14.0/kserve.yaml
+        e2e/repeat.sh kubectl apply --server-side -f https://github.com/kserve/kserve/releases/download/v0.14.0/kserve-cluster-resources.yaml
+        kubectl patch configmap/inferenceservice-config -n kserve --type=strategic -p '{"data": {"deploy": "{\"defaultDeploymentMode\": \"RawDeployment\"}"}}'
+        e2e/enable-modelcar.sh
+    - name: Load image in KinD for amd64
+      run: |
+        digest=$(skopeo inspect --tls-verify=false --raw docker://localhost:5001/nstestorg/modelcar | jq -r '.manifests[] | select(.platform.architecture == "amd64") | .digest')
+        skopeo copy --src-tls-verify=false docker://localhost:5001/nstestorg/modelcar@$digest docker-daemon:localhost:5001/nstestorg/modelcar:v1
+        kind load docker-image -n "kind" "localhost:5001/nstestorg/modelcar:v1"
+    - name: Apply Isvc using Modelcar # since the enable modelcar restart controller pod, better guard the kubectl apply
+      run: |
+        e2e/repeat.sh kubectl apply -f e2e/isvc-modelcar.yaml
+        kubectl wait --for=condition=Ready isvc/my-inference-service --timeout=240s
+    - name: Basic testing of Isvc that has Modelcar
+      run: |
+        echo "Starting port-forward..."
+        kubectl port-forward svc/my-inference-service-predictor 8080:80 &
+        PID=$!
+        sleep 2
+        echo "I have launched port-forward in background with: $PID."
+        echo "Check that OIP return the expected name"
+        curl -s http://localhost:8080/v2/models | jq -e '.models | index("my-inference-service") != null'
+        echo "Check that OIP produces an Inference Prediction"
+        curl -s -H "Content-Type: application/json" -d @e2e/data/input0.json http://localhost:8080/v2/models/my-inference-service/infer | jq
+        curl -s -H "Content-Type: application/json" -d @e2e/data/input1.json http://localhost:8080/v2/models/my-inference-service/infer | jq
+        curl -s -H "Content-Type: application/json" -d @e2e/data/input4.json http://localhost:8080/v2/models/my-inference-service/infer | jq
   e2e-oras:
     name: E2E using Oras CP
     runs-on: ubuntu-24.04

e2e/data/input0.json

@@ -0,0 +1,11 @@
+{
+    "inputs": [
+      {
+        "name": "input-0",
+        "shape": [1, 4],
+        "datatype": "FP32",
+        "data": [6.2, 3.4, 5.4, 2.3]
+      }
+    ]
+  }
+

e2e/data/input1.json

@@ -0,0 +1,11 @@
+{
+    "inputs": [
+      {
+        "name": "input-0",
+        "shape": [1, 4],
+        "datatype": "FP32",
+        "data": [5.4, 3.9, 1.7, 0.4]
+      }
+    ]
+  }
+

e2e/data/input4.json

@@ -0,0 +1,11 @@
+{
+    "inputs": [
+      {
+        "name": "input-0",
+        "shape": [1, 4],
+        "datatype": "FP32",
+        "data": [6.6, 3.0,  4.4, 1.4]
+      }
+    ]
+  }
+

e2e/enable-modelcar.sh

@@ -0,0 +1,24 @@
+#!/bin/bash
+
+# from https://kserve.github.io/website/latest/modelserving/storage/oci/#enabling-modelcars
+# Script to enable Modelcars
+# Fetch the current storageInitializer configuration
+config=$(kubectl get configmap inferenceservice-config -n kserve -o jsonpath='{.data.storageInitializer}')
+# Enable modelcars and set the UID for the containers to run (required for minikube)
+newValue=$(echo $config | jq -c '. + {"enableModelcar": true, "uidModelcar": 1010}')
+
+# Create a temporary directory for the patch file
+tmpdir=$(mktemp -d)
+cat <<EOT > $tmpdir/patch.txt
+[{
+  "op": "replace",
+  "path": "/data/storageInitializer",
+  "value": '$newValue'
+}]
+EOT
+
+# Apply the patch to the ConfigMap
+kubectl patch configmap -n kserve inferenceservice-config --type=json --patch-file=$tmpdir/patch.txt
+
+# Restart the KServe controller to apply changes
+kubectl delete pod -n kserve -l control-plane=kserve-controller-manager

e2e/isvc-modelcar.yaml

@@ -0,0 +1,10 @@
+apiVersion: serving.kserve.io/v1beta1
+kind: InferenceService
+metadata:
+  name: my-inference-service
+spec:
+  predictor:
+    model:
+      modelFormat:
+        name: sklearn
+      storageUri: oci://localhost:5001/nstestorg/modelcar:v1

e2e/repeat.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+set -e
+set -o xtrace
+
+while ! $@ ; do echo "Retrying..."; sleep 5; done