CI:update integration test with deployment check and push action enhancement

Signed-off-by: Sam Yuan <[email protected]>

Author: SamYuan1990

.github/workflows/image-base.yml

@@ -6,7 +6,7 @@ on:
       - 'base'
 
 jobs:
-  docker:
+  baseimagebuild:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout

.github/workflows/image.yml

@@ -0,0 +1,22 @@
+name: image
+
+on:
+  push:
+    branches: [ main ]
+
+jobs:
+  image_build:
+    runs-on: ubuntu-latest
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v3
+    - name: Login to Quay
+      uses: docker/login-action@v1
+      with:
+        registry: quay.io/sustainable_computing_io
+        username: ${{ secrets.BOT_NAME }}
+        password: ${{ secrets.BOT_TOKEN }}
+    - name: make container
+      run: make
+    - name: push to quay
+      run: podman push quay.io/sustainable_computing_io/kepler:latest

.github/workflows/integration_test.yml

@@ -2,8 +2,6 @@ name: Integration test
 
 on:
   pull_request:
-  push:
-    branches: [ main ]
 
 jobs:
   integration_test:
@@ -32,19 +30,3 @@ jobs:
         env:
             CLUSTER_PROVIDER: ${{matrix.kube_provider}}
             CTR_CMD: docker
-
-      - name: Login to Quay
-        if: github.event_name == 'push'
-        uses: docker/login-action@v1
-        with:
-            registry: quay.io/sustainable_computing_io
-            username: ${{ secrets.BOT_NAME }}
-            password: ${{ secrets.BOT_TOKEN }}
-
-      - name: make container
-        if: github.event_name == 'push'
-        run: make
-
-      - name: push to quay
-        if: github.event_name == 'push'
-        run: podman push quay.io/sustainable_computing_io/kepler:latest

hack/cluster-deploy.sh

@@ -31,10 +31,29 @@ function main() {
 
     # Ignore errors because some clusters might not have prometheus operator
     kubectl apply -f ${MANIFESTS_OUT_DIR} || true
-    kubectl rollout status daemonset kepler-exporter -n kepler --timeout 60s
-
-    echo "Check the logs of the kepler-exporter with:"
-    echo "kubectl -n kepler logs daemonset.apps/kepler-exporter"
+    
+    # round for 3 times and each for 60s
+    # check if the rollout status is running
+    deploy_status=1
+    for i in 1 2 3
+    do
+        echo "check deployment status for round $i"
+        kubectl rollout status daemonset kepler-exporter -n kepler --timeout 60s
+        #check rollout status
+        if [ $? -eq 0 ]
+        then
+            deploy_status=0
+            break
+        fi
+    done 
+    # if deployment in error
+    if test $[deploy_status] -eq 1
+    then
+        echo "Check the status of the kepler-exporter"
+        kubectl -n kepler describe daemonset.apps/kepler-exporter
+        echo "Check the logs of the kepler-exporter"
+        kubectl -n kepler logs daemonset.apps/kepler-exporter
+    fi
 }
 
 main "$@"