Only display running pod logs in CI

Signed-off-by: Sivanantham Chinnaiyan <[email protected]>

.github/workflows/scheduled-image-scan.yml

@@ -45,7 +45,7 @@ jobs:
 
       - name: Upload sarif file to Github Code Scanning
         if: always()
-        uses: github/codeql-action/upload-sarif@v2
+        uses: github/codeql-action/upload-sarif@v3
         with:
           sarif_file: application/${{ matrix.image.name }}/docker.snyk.sarif
 
@@ -81,7 +81,7 @@ jobs:
 
       - name: Upload sarif file to Github Code Scanning
         if: always()
-        uses: github/codeql-action/upload-sarif@v2
+        uses: github/codeql-action/upload-sarif@v3
         with:
           sarif_file: application/${{ matrix.image.name }}/docker.snyk.sarif
 
@@ -115,6 +115,6 @@ jobs:
 
       - name: Upload sarif file to Github Code Scanning
         if: always()
-        uses: github/codeql-action/upload-sarif@v2
+        uses: github/codeql-action/upload-sarif@v3
         with:
           sarif_file: application/${{ matrix.image.name }}/docker.snyk.sarif

pkg/controller/v1beta1/inferenceservice/reconcilers/ingress/ingress_reconciler.go

@@ -483,7 +483,7 @@ func probeIngress(url string) (bool, error) {
 	defer cancel()
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, target, nil)
 	if err != nil {
-		return isReady, errors.Wrapf(err, "failed to probe ingress %s", target)
+		return isReady, errors.Wrapf(err, "failed to create ingress probe request %s", target)
 	}
 	// ProbeKey is the name of a header that can be added to requests to probe the ingress.
 	// Requests with this header will not be passed to the user container or included in request metrics.
@@ -497,6 +497,8 @@ func probeIngress(url string) (bool, error) {
 	}
 	if resp.StatusCode == http.StatusOK {
 		isReady = true
+	} else {
+		log.V(1).Error(fmt.Errorf("failed to probe ingress"), "Failed to probe ingress", "url", target)
 	}
 	return isReady, nil
 }

test/scripts/gh-actions/setup-deps.sh

@@ -27,10 +27,14 @@ DEPLOYMENT_MODE="${1:-'serverless'}"
 ISTIO_VERSION="1.19.4"
 CERT_MANAGER_VERSION="v1.5.0"
 YQ_VERSION="v4.28.1"
+JQ_VERSION="1.7.1"
 
 echo "Installing yq ..."
 wget https://github.com/mikefarah/yq/releases/download/${YQ_VERSION}/yq_linux_amd64 -O /usr/local/bin/yq && chmod +x /usr/local/bin/yq
 
+echo "Installing jq ..."
+wget https://github.com/jqlang/jq/releases/download/jq-${JQ_VERSION}/jq-linux-amd64 -O /usr/local/bin/jq && chmod +x /usr/local/bin/jq
+
 echo "Installing Istio ..."
 mkdir istio_tmp
 pushd istio_tmp >/dev/null

test/scripts/gh-actions/status-check.sh

@@ -31,31 +31,39 @@ kubectl logs -l control-plane=kserve-controller-manager -n kserve -c manager --t
 echo "::endgroup::"
 
 echo "::group::Predictor Pod logs"
-for pod in $(kubectl get pods -l 'component in (predictor)' -o jsonpath='{.items[*].metadata.name}' -n kserve-ci-e2e-test); do
+# Get only pods that are not being deleted.
+predictor_pods=$(kubectl get pods -l 'component in (predictor)' -o json -n kserve-ci-e2e-test | jq -r '.items[] | select(.metadata.deletionTimestamp == null) | .metadata.name')
+for pod in $predictor_pods; do
     echo "=====================================  Logs for Predictor Pod: $pod  ========================================="
     kubectl logs "$pod" -c kserve-container -n kserve-ci-e2e-test --tail 500
     echo "================================================================================================================"
 done
 echo "::endgroup::"
 
 echo "::group::Transformer Pod logs"
-for pod in $(kubectl get pods -l 'component in (transformer)' -o jsonpath='{.items[*].metadata.name}' -n kserve-ci-e2e-test); do
+# Get only pods that are not being deleted.
+transformer_pods=$(kubectl get pods -l 'component in (transformer)' -o json -n kserve-ci-e2e-test | jq -r '.items[] | select(.metadata.deletionTimestamp == null) | .metadata.name')
+for pod in $transformer_pods; do
     echo "=====================================  Logs for Transformer Pod: $pod  ======================================="
     kubectl logs "$pod" -c kserve-container -n kserve-ci-e2e-test --tail 500
     echo "================================================================================================================"
 done
 echo "::endgroup::"
 
 echo "::group::Explainer Pod logs"
-for pod in $(kubectl get pods -l 'component in (explainer)' -o jsonpath='{.items[*].metadata.name}' -n kserve-ci-e2e-test); do
+# Get only pods that are not being deleted.
+explainer_pods=$(kubectl get pods -l 'component in (explainer)' -o json -n kserve-ci-e2e-test | jq -r '.items[] | select(.metadata.deletionTimestamp == null) | .metadata.name')
+for pod in $explainer_pods; do
     echo "=====================================  Logs for Explainer Pod: $pod  ========================================="
     kubectl logs "$pod" -c kserve-container -n kserve-ci-e2e-test --tail 500
     echo "================================================================================================================"
 done
 echo "::endgroup::"
 
 echo "::group::InferenceGraph Pod logs"
-for pod in $(kubectl get pods -l 'serving.kserve.io/inferencegraph=model-chainer' -o jsonpath='{.items[*].metadata.name}' -n kserve-ci-e2e-test); do
+# Get only pods that are not being deleted.
+graph_pods=$(kubectl get pods -l 'serving.kserve.io/inferencegraph=model-chainer' -o json -n kserve-ci-e2e-test | jq -r '.items[] | select(.metadata.deletionTimestamp == null) | .metadata.name')
+for pod in $graph_pods; do
     echo "=====================================  Logs for Graph Pod: $pod  ========================================="
     kubectl logs "$pod" -c user-container -n kserve-ci-e2e-test --tail 500
     echo "================================================================================================================"