Merge pull request #4481 from k8s-infra-cherrypick-robot/cherry-pick-4470-to-release-1.13

[release-1.13] overcome transient failures + retry when creating test cluster

Author: k8s-ci-robot

Makefile

@@ -301,14 +301,14 @@ create-management-cluster: $(KUSTOMIZE) $(ENVSUBST) $(KUBECTL) $(KIND) ## Create
 	./hack/create-custom-cloud-provider-config.sh
 
 	# Deploy CAPI
-	curl --retry $(CURL_RETRIES) -sSL https://github.com/kubernetes-sigs/cluster-api/releases/download/v1.6.1/cluster-api-components.yaml | $(ENVSUBST) | $(KUBECTL) apply -f -
+	timeout --foreground 300 bash -c "until curl --retry $(CURL_RETRIES) -sSL https://github.com/kubernetes-sigs/cluster-api/releases/download/v1.6.1/cluster-api-components.yaml | $(ENVSUBST) | $(KUBECTL) apply -f -; do sleep 5; done"
 
 	# Deploy CAAPH
-	curl --retry $(CURL_RETRIES) -sSL https://github.com/kubernetes-sigs/cluster-api-addon-provider-helm/releases/download/v0.1.0-alpha.10/addon-components.yaml | $(ENVSUBST) | $(KUBECTL) apply -f -
+	timeout --foreground 300 bash -c "until curl --retry $(CURL_RETRIES) -sSL https://github.com/kubernetes-sigs/cluster-api-addon-provider-helm/releases/download/v0.1.0-alpha.10/addon-components.yaml | $(ENVSUBST) | $(KUBECTL) apply -f -; do sleep 5; done"
 
 	# Deploy CAPZ
 	$(KIND) load docker-image $(CONTROLLER_IMG)-$(ARCH):$(TAG) --name=$(KIND_CLUSTER_NAME)
-	$(KUSTOMIZE) build config/default | $(ENVSUBST) | $(KUBECTL) apply -f - --server-side=true
+	timeout --foreground 300 bash -c "until $(KUSTOMIZE) build config/default | $(ENVSUBST) | $(KUBECTL) apply -f - --server-side=true; do sleep 5; done"
 
 	# Wait for CAPI deployments
 	$(KUBECTL) wait --for=condition=Available --timeout=5m -n capi-system deployment -l cluster.x-k8s.io/provider=cluster-api
@@ -322,8 +322,8 @@ create-management-cluster: $(KUSTOMIZE) $(ENVSUBST) $(KUBECTL) $(KIND) ## Create
 	timeout --foreground 300 bash -c "until $(KUBECTL) get clusterresourcesets -A; do sleep 3; done"
 
 	# install Windows Calico cluster resource set
-	$(KUBECTL) create configmap calico-windows-addon --from-file="$(ADDONS_DIR)/windows/calico" --dry-run=client -o yaml | kubectl apply -f -
-	$(KUBECTL) apply -f templates/addons/windows/calico-resource-set.yaml
+	timeout --foreground 300 bash -c "until $(KUBECTL) create configmap calico-windows-addon --from-file="$(ADDONS_DIR)/windows/calico" --dry-run=client -o yaml | kubectl apply -f -; do sleep 5; done"
+	timeout --foreground 300 bash -c "until $(KUBECTL) apply -f templates/addons/windows/calico-resource-set.yaml; do sleep 5; done"
 
 	# Wait for CAPZ deployments
 	$(KUBECTL) wait --for=condition=Available --timeout=5m -n capz-system deployment --all
@@ -339,11 +339,11 @@ create-management-cluster: $(KUSTOMIZE) $(ENVSUBST) $(KUBECTL) $(KIND) ## Create
 create-workload-cluster: $(ENVSUBST) $(KUBECTL) ## Create a workload cluster.
 	# Create workload Cluster.
 	@if [ -f "$(TEMPLATES_DIR)/$(CLUSTER_TEMPLATE)" ]; then \
-		$(ENVSUBST) < "$(TEMPLATES_DIR)/$(CLUSTER_TEMPLATE)" | $(KUBECTL) apply -f -; \
+		timeout --foreground 300 bash -c "until $(ENVSUBST) < $(TEMPLATES_DIR)/$(CLUSTER_TEMPLATE) | $(KUBECTL) apply -f -; do sleep 5; done"; \
 	elif [ -f "$(CLUSTER_TEMPLATE)" ]; then \
-		$(ENVSUBST) < "$(CLUSTER_TEMPLATE)" | $(KUBECTL) apply -f -; \
+		timeout --foreground 300 bash -c "until $(ENVSUBST) < "$(CLUSTER_TEMPLATE)" | $(KUBECTL) apply -f -; do sleep 5; done"; \
 	else \
-		curl --retry "$(CURL_RETRIES)" "$(CLUSTER_TEMPLATE)" | "$(ENVSUBST)" | $(KUBECTL) apply -f -; \
+		timeout --foreground 300 bash -c "until curl --retry "$(CURL_RETRIES)" "$(CLUSTER_TEMPLATE)" | "$(ENVSUBST)" | $(KUBECTL) apply -f -; do sleep 5; done"; \
 	fi
 
 	# Wait for the kubeconfig to become available.

hack/create-dev-cluster.sh

@@ -28,6 +28,8 @@ make envsubst
 
 export REGISTRY="${REGISTRY:-registry.local/fake}"
 
+export CLUSTER_CREATE_ATTEMPTS="${CLUSTER_CREATE_ATTEMPTS:-3}"
+
 # Cluster settings.
 export CLUSTER_NAME="${CLUSTER_NAME:-capz-test}"
 export AZURE_VNET_NAME=${CLUSTER_NAME}-vnet
@@ -62,14 +64,28 @@ capz::util::generate_ssh_key
 echo "================ DOCKER BUILD ==============="
 PULL_POLICY=IfNotPresent make modules docker-build
 
-echo "================ MAKE CLEAN ==============="
-make clean
+setup() {
+    echo "================ MAKE CLEAN ==============="
+    make clean
+
+    echo "================ KIND RESET ==============="
+    make kind-reset
+
+    echo "================ INSTALL TOOLS ==============="
+    make install-tools
+}
 
-echo "================ KIND RESET ==============="
-make kind-reset
+create_cluster() {
+    echo "================ CREATE CLUSTER ==============="
+    make create-cluster
+}
 
-echo "================ INSTALL TOOLS ==============="
-make install-tools
+setup
 
-echo "================ CREATE CLUSTER ==============="
-make create-cluster
+retries=$CLUSTER_CREATE_ATTEMPTS
+while ((retries > 0)); do
+    create_cluster && break
+    setup
+    sleep 5
+    ((retries --))
+done