Netpol (#27)

* update e2e tests with netpol tests, and fix minor issues

* fix lint

* turn tuning parameters into constants

* add changelog

* fix first version name

Author: g-bgg

CHANGELOG.md

@@ -1,14 +1,34 @@
-## Version (0.0.0)
-### Releshed/Unreleased
-### Date
+# Changelog
 
-Changes:
-- List of changes
+All notable changes to this project will be documented in this file.
 
-Improvements:
-- List of improvements 
+## Table of Contents
 
-Bug Fixes:
-- NA
+- [2.0.2](#202)
+- [2.0.1](#201)
+- [2.0.0](#200)
+- [0.1.0](#010)
 
----
+---
+
+## `2.0.2`
+
+- integrate e2e tests with network policies
+- fix a bug in udp testing
+
+## `2.0.1`
+
+- fix release naming
+
+## `2.0.0`
+
+- complete rewrite of the tool in Go, with unit and integration tests
+- leverages the ephemeral container support in Kubernetes > v1.25
+- test case(s) are written in YAML
+- support for Pods, StatefulSets, DaemonSets and Deployments which are directly referred through their names in the test suites
+- artifacts are available for download
+
+## `0.1.0`
+
+- initial release
+- no artifacts available

e2e/clusters/gke-dataplanev2/main.tf

@@ -13,7 +13,7 @@ provider "google" {
 
 resource "google_container_cluster" "e2etest" {
   name               = var.cluster_name
-  initial_node_count = 3
+  initial_node_count = 4
   datapath_provider  = var.use_dataplanev2 ? "ADVANCED_DATAPATH" : null
   ip_allocation_policy {}
   node_config {

e2e/clusters/gke-vpc/main.tf

@@ -13,11 +13,18 @@ provider "google" {
 
 resource "google_container_cluster" "e2etest" {
   name               = var.cluster_name
-  initial_node_count = 3
-  datapath_provider  = var.use_dataplanev2 ? "ADVANCED_DATAPATH" : null
+  initial_node_count = 4
+  addons_config {
+    network_policy_config {
+      disabled = false
+    }
+  }
+  network_policy {
+    enabled = true
+  }
   ip_allocation_policy {}
   node_config {
-    machine_type = "e2-medium"
+    machine_type = "e2-standard-2"
   }
 
   release_channel {

e2e/e2e_test.go

@@ -177,8 +177,33 @@ func createTestDestroy(t *testing.T, gc helpers.GenericCluster) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	// run the tests
 
+	// run the tests without network policies
+	runTests(ctx, t, svc, netAssertTestCases)
+
+	if gc.SkipNetPolTests() {
+		return
+	}
+
+	// create the network policies
+	k8s.KubectlApply(t, options, "./manifests/networkpolicies.yaml")
+
+	// read the tests again for a fresh start
+	netAssertTestCases, err = data.ReadTestsFromFile(testCasesFile)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// set the exit to 1 since this time the network policies will block the traffic
+	for _, tc := range netAssertTestCases {
+		tc.ExitCode = 1
+	}
+
+	// run the tests with network policies
+	runTests(ctx, t, svc, netAssertTestCases)
+}
+
+func runTests(ctx context.Context, t *testing.T, svc *kubeops.Service, netAssertTestCases data.Tests) {
 	lg := logger.NewHCLogger("INFO", "netassertv2-e2e", os.Stdout)
 	testRunner := engine.New(svc, lg)
 

e2e/helpers/common.go

@@ -14,4 +14,5 @@ type GenericCluster interface {
 	Create(t *testing.T)
 	Destroy(t *testing.T)
 	KubeConfigGet() string
+	SkipNetPolTests() bool
 }

e2e/helpers/eks.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"os"
 	"testing"
+	"time"
 
 	"github.com/controlplaneio/netassert/v2/internal/kubeops"
 	"github.com/controlplaneio/netassert/v2/internal/logger"
@@ -110,6 +111,9 @@ func (g *EKSCluster) installCalico(t *testing.T) {
 	if _, err := terraform.InitAndApplyE(t, newTFOptions); err != nil {
 		t.Fatalf("failed to run terraform init and apply: %s", err)
 	}
+
+	svc.Log.Info("Sleeping 20 minutes so connectivity from the cluster to the Internet is restored")
+	time.Sleep(20 * time.Minute)
 }
 
 func (g *EKSCluster) Destroy(t *testing.T) {
@@ -121,3 +125,7 @@ func (g *EKSCluster) Destroy(t *testing.T) {
 func (g *EKSCluster) KubeConfigGet() string {
 	return g.kubeConfigPath
 }
+
+func (g *EKSCluster) SkipNetPolTests() bool {
+	return g.networkMode != Calico
+}

e2e/helpers/gke.go

@@ -60,3 +60,7 @@ func (g *GKECluster) Destroy(t *testing.T) {
 func (g *GKECluster) KubeConfigGet() string {
 	return g.kubeConfigPath
 }
+
+func (g *GKECluster) SkipNetPolTests() bool {
+	return false // network policies are supported by all gke cluster configurations
+}

e2e/manifests/networkpolicies.yaml

@@ -0,0 +1,70 @@
+---
+kind: NetworkPolicy
+apiVersion: networking.k8s.io/v1
+metadata:
+  namespace: fluentd
+  name: fluentd-elasticsearch
+spec:
+  podSelector:
+    matchLabels:
+      name: fluentd-elasticsearch
+  ingress:
+---
+kind: NetworkPolicy
+apiVersion: networking.k8s.io/v1
+metadata:
+  namespace: echoserver
+  name: echoserver
+spec:
+  podSelector:
+    matchLabels:
+      app: echoserver
+  ingress:
+---
+kind: NetworkPolicy
+apiVersion: networking.k8s.io/v1
+metadata:
+  namespace: busybox
+  name: busybox
+spec:
+  podSelector:
+    matchLabels:
+      app: busybox
+  policyTypes:
+  - Ingress
+  - Egress
+  ingress:
+  egress:
+---
+kind: NetworkPolicy
+apiVersion: networking.k8s.io/v1
+metadata:
+  namespace: pod1
+  name: pod1
+spec:
+  podSelector:
+    matchLabels:
+      name: pod1
+  ingress:
+---
+kind: NetworkPolicy
+apiVersion: networking.k8s.io/v1
+metadata:
+  namespace: pod2
+  name: pod2
+spec:
+  podSelector:
+    matchLabels:
+      name: pod2
+  ingress:
+---
+kind: NetworkPolicy
+apiVersion: networking.k8s.io/v1
+metadata:
+  namespace: web
+  name: nginx
+spec:
+  podSelector:
+    matchLabels:
+      app: nginx
+  ingress:

e2e/manifests/pod1-pod2.yaml

@@ -15,6 +15,8 @@ kind: Pod
 metadata:
   name: pod2
   namespace: pod2
+  labels:
+    name: pod2
 spec:
   containers:
     - name: webserver
@@ -27,6 +29,8 @@ kind: Pod
 metadata:
   name: pod1
   namespace: pod1
+  labels:
+    name: pod1
 spec:
   containers:
     - name: busybox

internal/engine/run_udp.go

@@ -12,8 +12,10 @@ import (
 )
 
 const (
-	defaultNetInt  = `eth0` // default network interface
-	defaultSnapLen = 1024   // default size of the packet snap length
+	defaultNetInt                   = `eth0` // default network interface
+	defaultSnapLen                  = 1024   // default size of the packet snap length
+	ephemeralContainersExtraSeconds = 23     // fixed extra time given for the ephemeral containers to come online
+	attemptsMultiplier              = 3      // increase the attempts to ensure that we send three times the packets
 )
 
 // RunUDPTest - runs a UDP test
@@ -114,7 +116,7 @@ func (e *Engine) RunUDPTest(
 		string(te.Protocol),
 		te.Attempts,
 		networkInterface,
-		te.TimeoutSeconds+5, // add 5 seconds for the Container to come online
+		te.TimeoutSeconds,
 	)
 	if err != nil {
 		return fmt.Errorf("failed to build sniffer ephemeral container for test %s: %w", te.Name, err)
@@ -128,7 +130,7 @@ func (e *Engine) RunUDPTest(
 		strconv.Itoa(te.TargetPort),
 		string(te.Protocol),
 		msg,
-		te.Attempts*3, // increase the attempts to ensure that we send three times the packets
+		te.Attempts*attemptsMultiplier,
 	)
 	if err != nil {
 		return fmt.Errorf("unable to build ephemeral scanner container for test %s: %w", te.Name, err)
@@ -152,7 +154,7 @@ func (e *Engine) RunUDPTest(
 	exitCodeSnifferCtr, err := e.Service.GetExitStatusOfEphemeralContainer(
 		ctx,
 		snifferContainerName,
-		time.Duration(te.TimeoutSeconds)*time.Second,
+		time.Duration(te.TimeoutSeconds+ephemeralContainersExtraSeconds)*time.Second,
 		dstPod.Name,
 		dstPod.Namespace,
 	)
@@ -174,7 +176,7 @@ func (e *Engine) RunUDPTest(
 	// get the exit status of the scanner container
 	exitCodeScanner, err := e.Service.GetExitStatusOfEphemeralContainer(
 		ctx, scannerContainerName,
-		time.Duration(te.TimeoutSeconds+10)*time.Second,
+		time.Duration(te.TimeoutSeconds+ephemeralContainersExtraSeconds)*time.Second,
 		srcPod.Name,
 		srcPod.Namespace,
 	)