add canary test entrypoint script

scripts/run-canary-test.sh

@@ -0,0 +1,122 @@
+#!/bin/bash
+
+# The script runs amazon-vpc-cni Canary tests on the default
+# addon version and then runs smoke test on the latest addon version.
+
+set -e
+
+SECONDS=0
+SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+INTEGRATION_TEST_DIR="$SCRIPT_DIR/../test/integration-new"
+VPC_CNI_ADDON_NAME="vpc-cni"
+
+echo "Running Canary tests for amazon-vpc-cni-k8s with the following variables
+KUBE_CONFIG_PATH:  $KUBE_CONFIG_PATH
+CLUSTER_NAME: $CLUSTER_NAME
+REGION: $REGION
+ENDPOINT: $ENDPOINT"
+
+if [[ -n "${ENDPOINT}" ]]; then
+  ENDPOINT_FLAG="--endpoint $ENDPOINT"
+fi
+
+function load_cluster_details() {
+  echo "loading cluster details $CLUSTER_NAME"
+  DESCRIBE_CLUSTER_OP=$(aws eks describe-cluster --name "$CLUSTER_NAME" --region "$REGION" $ENDPOINT_FLAG)
+  VPC_ID=$(echo "$DESCRIBE_CLUSTER_OP" | jq -r '.cluster.resourcesVpcConfig.vpcId')
+  K8S_VERSION=$(echo "$DESCRIBE_CLUSTER_OP" | jq .cluster.version -r)
+}
+
+function load_addon_details() {
+  echo "loading $VPC_CNI_ADDON_NAME addon details"
+  DESCRIBE_ADDON_VERSIONS=$(aws eks describe-addon-versions --addon-name $VPC_CNI_ADDON_NAME --kubernetes-version "$K8S_VERSION")
+
+  LATEST_ADDON_VERSION=$(echo "$DESCRIBE_ADDON_VERSIONS" | jq '.addons[0].addonVersions[0].addonVersion' -r)
+  DEFAULT_ADDON_VERSION=$(echo "$DESCRIBE_ADDON_VERSIONS" | jq -r '.addons[].addonVersions[] | select(.compatibilities[0].defaultVersion == true) | .addonVersion')
+}
+
+function wait_for_addon_status() {
+  local expected_status=$1
+
+  if [ "$expected_status" =  "DELETED" ]; then
+    while $(aws eks describe-addon $ENDPOINT_FLAG --cluster-name "$CLUSTER_NAME" --addon-name $VPC_CNI_ADDON_NAME); do
+      echo "addon is still not deleted"
+      sleep 5
+    done
+    echo "addon deleted"
+    return
+  fi
+
+  while true
+  do
+    STATUS=$(aws eks describe-addon $ENDPOINT_FLAG --cluster-name "$CLUSTER_NAME" --addon-name $VPC_CNI_ADDON_NAME | jq -r '.addon.status')
+    if [ "$STATUS" = "$expected_status" ]; then
+      echo "addon status matches expected status"
+      return
+    fi
+    echo "addon status is not equal to $expected_status"
+    sleep 5
+  done
+}
+
+function install_add_on() {
+  local new_addon_version=$1
+
+  if DESCRIBE_ADDON=$(aws eks describe-addon $ENDPOINT_FLAG --cluster-name "$CLUSTER_NAME" --addon-name $VPC_CNI_ADDON_NAME); then
+    local current_addon_version=$(echo "$DESCRIBE_ADDON" | jq '.addon.addonVersion' -r)
+    if [ "$new_addon_version" != "$current_addon_version" ]; then
+      echo "deleting the $current_addon_version to install $new_addon_version"
+      aws eks delete-addon $ENDPOINT_FLAG --cluster-name "$CLUSTER_NAME" --addon-name "$VPC_CNI_ADDON_NAME"
+      wait_for_addon_status "DELETED"
+    else
+      echo "addon version $current_addon_version already installed"
+      patch_aws_node_maxunavialable
+      return
+    fi
+  fi
+
+  echo "installing addon $new_addon_version"
+  aws eks create-addon $ENDPOINT_FLAG --cluster-name "$CLUSTER_NAME" --addon-name $VPC_CNI_ADDON_NAME --resolve-conflicts OVERWRITE --addon-version $new_addon_version
+  patch_aws_node_maxunavialable
+  wait_for_addon_status "ACTIVE"
+}
+
+function patch_aws_node_maxunavialable() {
+   # Patch the aws-node, so any update in aws-node happens parallely for faster overall test execution
+   kubectl patch ds -n kube-system aws-node -p '{"spec":{"updateStrategy":{"rollingUpdate":{"maxUnavailable": "100%"}}}}'
+}
+
+function run_ginkgo_test() {
+  local focus=$1
+  echo "Running ginkgo tests with focus: $focus"
+  (cd "$INTEGRATION_TEST_DIR/cni" && ginkgo --focus="$focus" -v --timeout 20m --failOnPending -- --cluster-kubeconfig="$KUBE_CONFIG_PATH" --cluster-name="$CLUSTER_NAME" --aws-region="$REGION" --aws-vpc-id="$VPC_ID" --ng-name-label-key="kubernetes.io/os" --ng-name-label-val="linux")
+  (cd "$INTEGRATION_TEST_DIR/ipamd" && ginkgo --focus="$focus" -v --timeout 10m --failOnPending -- --cluster-kubeconfig="$KUBE_CONFIG_PATH" --cluster-name="$CLUSTER_NAME" --aws-region="$REGION" --aws-vpc-id="$VPC_ID" --ng-name-label-key="kubernetes.io/os" --ng-name-label-val="linux")
+}
+
+load_cluster_details
+load_addon_details
+
+# TODO: v1.7.5 restarts continiously if IMDS goes out of sync, the issue is mitigated
+# from v.1.8.0 onwards, once the default addon is updated to v1.8.0+ we should uncomment
+# the following code. See: https://github.com/aws/amazon-vpc-cni-k8s/issues/1340
+
+# Run more comprehensive test on the default addon version. CANARY focused tests
+# tests basic functionlity plus test that could detect issues with dependencies
+# early on.
+#echo "Running Canary tests on the default addon version"
+#install_add_on "$DEFAULT_ADDON_VERSION"
+#run_ginkgo_test "CANARY"
+
+# Run smoke test on the latest addon version. Smoke tests contains a subset of test
+# used in Canary tests.
+#echo "Running Smoke tests on the latest addon version"
+#install_add_on "$LATEST_ADDON_VERSION"
+#run_ginkgo_test "SMOKE"
+
+# TODO: Remove the following code once the v1.8.0+ is made the default addon version
+echo "Running Canary tests on the latest addon version"
+install_add_on "$LATEST_ADDON_VERSION"
+run_ginkgo_test "CANARY"
+
+
+echo "all tests ran successfully in $(($SECONDS / 60)) minutes and $(($SECONDS % 60)) seconds"

test/README.md

@@ -0,0 +1,26 @@
+## amazon-vpc-cni-k8s Test Framework
+The test framework consists of integration and e2e tests using the Ginkgo framework invoked manually and using collection of bash scripts using GitHub Workflow and Prow(not publicly available).
+
+### Types of Tests
+
+#### Pull Request Tests
+Runs on each Pull Request, verifies the new code changes don't introduce any regression. Given the entire test suite may span for long duration, run only the integration tests.
+
+#### Nightly Integration Tests
+Runs the entire test suite every night using the current GitHub build to catch regression.
+
+#### Canary Tests
+Canary tests run frequently, multiple times in a day on live production environment. Given all integration tests run spans for hours we can only run a limited set of tests of most important features along with tests that have dependencies like Load Balancer Service Creation. These test runs are not publicly accessible at this moment.
+
+Ginkgo Focus: [CANARY]
+
+### Smoke Tests
+Smoke test provide fail early mechanism by failing the test if basic functionality doesn't work. This can be used as a pre-requisite for the running the much longer Integration tests.
+
+Ginkgo Focus: [SMOKE]
+
+
+#### Work In Progress
+- Run Upstream Conformance tests as part of Nightly Integration tests.
+- Run All Integration/e2e tests as part of Nightly Integration tests.
+- Run all Integration tests on each Pull Request.

test/framework/resources/k8s/manifest/deployment.go

@@ -43,7 +43,7 @@ func NewBusyBoxDeploymentBuilder() *DeploymentBuilder {
 		replicas:               10,
 		container:              NewBusyBoxContainerBuilder().Build(),
 		labels:                 map[string]string{"role": "test"},
-		nodeSelector:           map[string]string{},
+		nodeSelector:           map[string]string{"kubernetes.io/os": "linux"},
 		terminationGracePeriod: 0,
 	}
 }
@@ -53,7 +53,7 @@ func NewDefaultDeploymentBuilder() *DeploymentBuilder {
 		namespace:              utils.DefaultTestNamespace,
 		terminationGracePeriod: 0,
 		labels:                 map[string]string{"role": "test"},
-		nodeSelector:           map[string]string{},
+		nodeSelector:           map[string]string{"kubernetes.io/os": "linux"},
 	}
 }
 

test/framework/resources/k8s/manifest/job.go

@@ -30,6 +30,7 @@ type JobBuilder struct {
 	terminationGracePeriod int
 	nodeName               string
 	hostNetwork            bool
+	nodeSelector           map[string]string
 }
 
 func NewDefaultJobBuilder() *JobBuilder {
@@ -39,6 +40,7 @@ func NewDefaultJobBuilder() *JobBuilder {
 		parallelism:            1,
 		terminationGracePeriod: 0,
 		labels:                 map[string]string{},
+		nodeSelector:           map[string]string{"kubernetes.io/os": "linux"},
 	}
 }
 
@@ -47,6 +49,11 @@ func (j *JobBuilder) Name(name string) *JobBuilder {
 	return j
 }
 
+func (j *JobBuilder) NodeSelector(selectorKey string, selectorVal string) *JobBuilder {
+	j.nodeSelector[selectorKey] = selectorVal
+	return j
+}
+
 func (j *JobBuilder) Namespace(namespace string) *JobBuilder {
 	j.namespace = namespace
 	return j
@@ -100,6 +107,7 @@ func (j *JobBuilder) Build() *batchV1.Job {
 					TerminationGracePeriodSeconds: aws.Int64(int64(j.terminationGracePeriod)),
 					RestartPolicy:                 v1.RestartPolicyNever,
 					NodeName:                      j.nodeName,
+					NodeSelector:                  j.nodeSelector,
 				},
 			},
 		},

test/framework/resources/k8s/manifest/pod.go

@@ -40,7 +40,7 @@ func NewDefaultPodBuilder() *PodBuilder {
 		labels:                 map[string]string{},
 		terminationGracePeriod: 0,
 		restartPolicy:          v1.RestartPolicyNever,
-		nodeSelector:           map[string]string{},
+		nodeSelector:           map[string]string{"kubernetes.io/os": "linux"},
 	}
 }
 

test/integration-new/cni/pod_traffic_test.go

@@ -187,7 +187,7 @@ var _ = Describe("test pod networking", func() {
 		})
 	})
 
-	Context("when establishing UDP connection from tester to server", func() {
+	Context("[CANARY][SMOKE] when establishing UDP connection from tester to server", func() {
 		BeforeEach(func() {
 			serverPort = 2273
 			protocol = ec2.ProtocolUdp
@@ -225,7 +225,7 @@ var _ = Describe("test pod networking", func() {
 		})
 	})
 
-	Context("when establishing TCP connection from tester to server", func() {
+	Context("[CANARY][SMOKE] when establishing TCP connection from tester to server", func() {
 
 		BeforeEach(func() {
 			serverPort = 2273

test/integration-new/cni/service_connectivity_test.go

@@ -34,7 +34,7 @@ const (
 )
 
 // Verifies connectivity to deployment behind different service types
-var _ = Describe("test service connectivity", func() {
+var _ = Describe("[CANARY] test service connectivity", func() {
 	var err error
 
 	// Deployment running the http server

test/integration-new/ipamd/eni_ip_leak_test.go

@@ -17,7 +17,7 @@ const (
 	HOST_POD_LABEL_VAL = "host"
 )
 
-var _ = Describe("ENI/IP Leak Test", func() {
+var _ = Describe("[CANARY] ENI/IP Leak Test", func() {
 	Context("ENI/IP Released on Pod Deletion", func() {
 		BeforeEach(func() {
 			By("creating test namespace")