Keep cluster alive?

.gitea/workflows/e2e.yml

@@ -17,90 +17,56 @@ jobs:
           # Install kubectl
           curl -LO "https://dl.k8s.io/release/$(curl -L -s https://dl.k8s.io/release/stable.txt)/bin/linux/amd64/kubectl"
           install kubectl /usr/local/bin/
-      
-      - name: Debug All
-        run: |
-          docker ps -adocker network rm kind || true
-          docker network ls
-          docker volume ls
-
-      - name: Delete Kind Cluster
-        run: |
-          kind delete cluster --name test-cluster-2
-          
-      - name: Sleep 5s
-        run: sleep 5
-        
-      - name: Delete kind network
-        run: |
-          docker network rm kind || true
-      
-      - name: Debug All
-        run: |
-          docker ps -a
-          docker network ls
-          docker volume ls
 
       - name: Create kind cluster
         run: |
-          cat > cluster.yml << 'EOF'
-          # three node (two workers) cluster config
-          kind: Cluster
-          apiVersion: kind.x-k8s.io/v1alpha4
-          nodes:
-          - role: control-plane
-          EOF
-          kind create cluster --name test-cluster-2 --config cluster.yml
+          docker inspect "$KIND_CONTAINER" || kind create cluster --name test-cluster --wait 5m
 
-      - name: Attach kind control-plane to job network and wait
+      - name: Connect kind to the job container's network and fix kubeconfig
+        if: always()
         run: |
-          apt-get update && apt-get install -y jq docker.io
+          # 1. Find the Docker network the job container is on
+          #    The job container's hostname is the container ID
+          JOB_CONTAINER_ID=$(hostname)
+          NETWORK_NAME=$(docker inspect "$JOB_CONTAINER_ID" \
+            --format '{{range $k, $v := .NetworkSettings.Networks}}{{$k}}{{end}}')
+            
+          echo "NETWORK_NAME=$NETWORK_NAME" >> $GITHUB_ENV
 
-          JOB_CONTAINER_ID="$(hostname)"
-          NETWORK_NAME="$(docker inspect "$JOB_CONTAINER_ID" --format '{{range $k, $v := .NetworkSettings.Networks}}{{$k}}{{end}}')"
-          echo "Job network: $NETWORK_NAME"
+          echo "Job container network: $NETWORK_NAME"
 
-          CONTROL_PLANE="test-cluster-2-control-plane"
-          docker network connect "$NETWORK_NAME" "$CONTROL_PLANE" || true
+          # 2. Get the kind control-plane container name
+          KIND_CONTAINER="test-cluster-control-plane"
 
-          KIND_IP="$(docker inspect "$CONTROL_PLANE" --format "{{(index .NetworkSettings.Networks \"$NETWORK_NAME\").IPAddress}}")"
-          echo "Kind IP: $KIND_IP"
+          echo "KIND_CONTAINER=$KIND_CONTAINER" >> $GITHUB_ENV
 
-          # Point kubectl at the control-plane directly
-          kubectl config set-cluster kind-test-cluster-2 --server="https://${KIND_IP}:6443"
-          kubectl config set-cluster kind-test-cluster-2 --insecure-skip-tls-verify=true
+          # 3. Connect the kind container to the same network
+          docker network connect "$NETWORK_NAME" "$KIND_CONTAINER"
 
-          # Wait for API and nodes
-          for i in $(seq 1 120); do
-            if kubectl get nodes >/dev/null 2>&1; then
-              break
-            fi
-            echo "Waiting for kind API... ($i/120)"
-            sleep 2
-          done
-          kubectl get nodes
+          # 4. Get the kind container's IP on that network
+          KIND_IP=$(docker inspect "$KIND_CONTAINER" \
+            --format "{{(index .NetworkSettings.Networks \"$NETWORK_NAME\").IPAddress}}")
+
+          echo "Kind container IP on shared network: $KIND_IP"
+
+          # 5. Rewrite the kubeconfig to use the kind container's IP
+          #    kind's API server listens on port 6443 inside the container
+          kubectl config set-cluster kind-test-cluster \
+            --server="https://${KIND_IP}:6443"
+
+          # 6. Since the TLS cert won't match the new IP, use insecure mode
+          #    OR set insecure-skip-tls-verify
+          kubectl config set-cluster kind-test-cluster \
+            --insecure-skip-tls-verify=true
 
       - name: Verify cluster access
+        if: always()
         run: |
           kubectl cluster-info
           kubectl get nodes
           kubectl get pods -A
-
-      - name: Debug All
-        if: always()
-        run: |
-          docker ps -a
-          docker network ls
-          docker volume ls
           
-      - name: Delete Kind Cluster
+      - name: Disconnect Kind Network
         if: always()
         run: |
-          kind delete cluster --name test-cluster-2
-
-      - name: Debug All
-        if: always()
-        run: |
-          docker ps -a
-          docker network ls
-          docker volume ls
+          docker network connect "$NETWORK_NAME" "$KIND_CONTAINER"