Add mixedos BGP e2e test

Signed-off-by: Manuel Buil <[email protected]>

tests/e2e/mixedosbgp/Vagrantfile

@@ -0,0 +1,144 @@
+ENV['VAGRANT_NO_PARALLEL'] = 'no'
+NODE_ROLES = (ENV['E2E_NODE_ROLES'] ||
+["server-0", "linux-agent-0", "windows-agent-0" ])
+NODE_BOXES = (ENV['E2E_NODE_BOXES'] ||
+['generic/ubuntu2004', 'generic/ubuntu2004', 'jborean93/WindowsServer2022'])
+GITHUB_BRANCH = (ENV['E2E_GITHUB_BRANCH'] || "master")
+RELEASE_VERSION = (ENV['E2E_RELEASE_VERSION'] || "")
+NODE_CPUS = (ENV['E2E_NODE_CPUS'] || 2).to_i
+NODE_MEMORY = (ENV['E2E_NODE_MEMORY'] || 3072).to_i
+# Virtualbox >= 6.1.28 require `/etc/vbox/network.conf` for expanded private networks 
+NETWORK_PREFIX = "10.10.10"
+install_type = ""
+
+def provision(vm, role, role_num, node_num)
+  vm.box = NODE_BOXES[node_num]
+  vm.hostname = role
+  # An expanded netmask is required to allow VM<-->VM communication, virtualbox defaults to /32
+  node_ip = "#{NETWORK_PREFIX}.#{100+node_num}"
+  vm.network "private_network", ip: node_ip, netmask: "255.255.255.0"
+
+  vagrant_defaults = '../vagrantdefaults.rb'
+  load vagrant_defaults if File.exists?(vagrant_defaults)
+
+  defaultOSConfigure(vm)
+
+  if role.include?("windows")
+    vm.provision "shell", path: "../scripts/install-bgp.ps1"
+    if !RELEASE_VERSION.empty?
+      install_type = "Version=#{RELEASE_VERSION}"
+    else 
+      vm.provision "shell", path: "../scripts/latest_commit.ps1", args: [GITHUB_BRANCH, "./rke2_commits.txt"]
+      install_type = "Commit=(Get-Content -TotalCount 1 ./rke2_commits.txt)"
+    end
+  else
+    if !RELEASE_VERSION.empty?
+      install_type = "INSTALL_RKE2_VERSION=#{RELEASE_VERSION}"
+    else 
+      # Grabs the last 5 commit SHA's from the given branch, then purges any commits that do not have a passing CI build
+      vm.provision "shell", path: "../scripts/latest_commit.sh", args: [GITHUB_BRANCH, "/tmp/rke2_commits"]
+      install_type = "INSTALL_RKE2_COMMIT=$(head\ -n\ 1\ /tmp/rke2_commits)"
+    end
+    vm.provision "shell", inline: "ping -c 2 rke2.io"
+    vm.provision "Create Calico Manifest", type: "shell", path: "../scripts/calico_manifestbgp.sh", args: [ "#{NETWORK_PREFIX}.1" ]
+  end
+
+  if role.include?("server") && role_num == 0
+    vm.provision :rke2, run: 'once' do |rke2|
+      rke2.env = %W[INSTALL_RKE2_TYPE=server #{install_type}]
+      rke2.config_mode = '0644' # side-step https://github.com/k3s-io/k3s/issues/4321
+      rke2.config = <<~YAML
+        write-kubeconfig-mode: '0644'
+        node-external-ip: #{NETWORK_PREFIX}.100
+        node-ip: #{NETWORK_PREFIX}.100
+        token: vagrant-rke2
+        cni: calico
+      YAML
+    end
+    vm.provision "Install sonobuoy", type: "shell", path: "../scripts/install_sonobuoy.sh"
+  elsif role.include?("server") && role_num != 0
+    vm.provision :rke2, run: 'once' do |rke2|
+      rke2.env = %W[INSTALL_RKE2_TYPE=server #{install_type}]
+      rke2.config_mode = '0644' # side-step https://github.com/k3s-io/k3s/issues/4321
+      rke2.config = <<~YAML
+        write-kubeconfig-mode: '0644'
+        node-external-ip: #{node_ip}
+        node-ip: #{node_ip}
+        server: https://#{NETWORK_PREFIX}.100:9345
+        token: vagrant-rke2
+        cni: calico
+      YAML
+    end
+  end
+  if role.include?("linux-agent")
+    vm.provision :rke2, run: 'once' do |rke2|
+      rke2.env = %W[INSTALL_RKE2_TYPE=agent #{install_type}]
+      rke2.config_mode = '0644' # side-step https://github.com/k3s-io/k3s/issues/4321
+      rke2.install_path = false
+      rke2.config = <<~YAML
+        write-kubeconfig-mode: '0644'
+        node-external-ip: #{node_ip}
+        node-ip: #{node_ip}
+        server: https://#{NETWORK_PREFIX}.100:9345
+        token: vagrant-rke2
+      YAML
+    end
+  end
+  if role.include?("windows-agent")
+    if !vm.box.match?(/Windows.*2022/) && !vm.box.match?(/Windows.*2019/)
+      puts "invalid box: " + vm.box + " found for windows agent"
+      abort
+    end
+
+    # For Windows GUI on virtualbox
+    # vm.provider "virtualbox" do |v|
+      # v.gui = true
+      # v.customize ["modifyvm", :id, "--vram", 128]
+      # v.customize ["modifyvm", :id, "--clipboard", "bidirectional"]
+      # v.customize ["modifyvm", :id, "--accelerate3d", "on"]
+      # v.customize ["modifyvm", :id, "--accelerate2dvideo", "on"]
+    # end
+    # If using libvirt, use virt-viewer for GUI after bring up the node
+    vm.provision :rke2, run: 'once' do |rke2|
+      rke2.env = [install_type]
+      rke2.config_mode = '0644' # side-step https://github.com/k3s-io/k3s/issues/4321
+      # rke2.skip_start = true
+      rke2.config = <<~YAML
+        node-external-ip: #{node_ip}
+        node-ip: #{node_ip}
+        server: https://#{NETWORK_PREFIX}.100:9345
+        token: vagrant-rke2
+      YAML
+    end
+  end
+end
+
+Vagrant.configure("2") do |config|
+  config.vagrant.plugins = ["vagrant-rke2", "vagrant-reload"]
+  # For windows, just use the password not the private key
+  config.ssh.password    = "vagrant"
+  # Default provider is libvirt, virtualbox is only provided as a backup
+  config.vm.provider "libvirt" do |v|
+    v.cpus = NODE_CPUS
+    v.memory = NODE_MEMORY
+  end
+  config.vm.provider "virtualbox" do |v|
+    v.cpus = NODE_CPUS
+    v.memory = NODE_MEMORY
+    v.linked_clone = false
+  end
+
+  if NODE_ROLES.kind_of?(String)
+    NODE_ROLES = NODE_ROLES.split(" ", -1)
+  end
+  if NODE_BOXES.kind_of?(String)
+    NODE_BOXES = NODE_BOXES.split(" ", -1)
+  end
+
+  NODE_ROLES.each_with_index do |name, i|
+    config.vm.define name do |node|
+      role_num = name.split("-", -1).pop.to_i
+      provision(node.vm, name, role_num, i)
+    end
+  end
+end

tests/e2e/mixedosbgp/mixedosbgp_test.go

@@ -0,0 +1,185 @@
+package mixedosbgp
+
+import (
+	"flag"
+	"fmt"
+	"os"
+	"strconv"
+	"strings"
+	"testing"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+	"github.com/rancher/rke2/tests/e2e"
+)
+
+// Valid nodeOS: generic/ubuntu2004, opensuse/Leap-15.3.x86_64
+var nodeOS = flag.String("nodeOS", "generic/ubuntu2004", "operating system for linux nodes")
+var serverCount = flag.Int("serverCount", 1, "number of server nodes")
+var linuxAgentCount = flag.Int("linuxAgentCount", 1, "number of linux agent nodes")
+var windowsAgentCount = flag.Int("windowsAgentCount", 1, "number of windows agent nodes")
+var ci = flag.Bool("ci", false, "running on CI")
+
+const defaultWindowsOS = "jborean93/WindowsServer2022"
+
+func Test_E2EMixedOSBGPValidation(t *testing.T) {
+	flag.Parse()
+	RegisterFailHandler(Fail)
+	suiteConfig, reporterConfig := GinkgoConfiguration()
+	RunSpecs(t, "Validate MixedOS BGP Test Suite", suiteConfig, reporterConfig)
+}
+
+func createMixedCluster(nodeOS string, serverCount, linuxAgentCount, windowsAgentCount int) ([]string, []string, []string, error) {
+	serverNodeNames := []string{}
+	for i := 0; i < serverCount; i++ {
+		serverNodeNames = append(serverNodeNames, "server-"+strconv.Itoa(i))
+	}
+	linuxAgentNames := []string{}
+	for i := 0; i < linuxAgentCount; i++ {
+		linuxAgentNames = append(linuxAgentNames, "linux-agent-"+strconv.Itoa(i))
+	}
+	windowsAgentNames := []string{}
+	for i := 0; i < linuxAgentCount; i++ {
+		windowsAgentNames = append(windowsAgentNames, "linux-agent-"+strconv.Itoa(i))
+	}
+	nodeRoles := strings.Join(serverNodeNames, " ") + " " + strings.Join(linuxAgentNames, " ") + " " + strings.Join(windowsAgentNames, " ")
+	nodeRoles = strings.TrimSpace(nodeRoles)
+	nodeBoxes := strings.Repeat(nodeOS+" ", serverCount+linuxAgentCount)
+	nodeBoxes += strings.Repeat(defaultWindowsOS+" ", windowsAgentCount)
+	nodeBoxes = strings.TrimSpace(nodeBoxes)
+
+	var testOptions string
+	for _, env := range os.Environ() {
+		if strings.HasPrefix(env, "E2E_") {
+			testOptions += " " + env
+		}
+	}
+
+	cmd := fmt.Sprintf("NODE_ROLES=\"%s\" NODE_BOXES=\"%s\" %s vagrant up &> vagrant.log", nodeRoles, nodeBoxes, testOptions)
+	fmt.Println(cmd)
+	if _, err := e2e.RunCommand(cmd); err != nil {
+		fmt.Println("Error Creating Cluster", err)
+		return nil, nil, nil, err
+	}
+	return serverNodeNames, linuxAgentNames, windowsAgentNames, nil
+}
+
+var (
+	kubeConfigFile    string
+	serverNodeNames   []string
+	linuxAgentNames   []string
+	windowsAgentNames []string
+)
+
+var _ = ReportAfterEach(e2e.GenReport)
+var _ = Describe("Verify Basic Cluster Creation", Ordered, func() {
+
+	It("Starts up with no issues", func() {
+		var err error
+		serverNodeNames, linuxAgentNames, windowsAgentNames, err = createMixedCluster(*nodeOS, *serverCount, *linuxAgentCount, *windowsAgentCount)
+		Expect(err).NotTo(HaveOccurred(), e2e.GetVagrantLog(err))
+		fmt.Println("CLUSTER CONFIG")
+		fmt.Println("OS:", *nodeOS)
+		fmt.Println("Server Nodes:", serverNodeNames)
+		fmt.Println("Agent Nodes:", linuxAgentNames)
+		kubeConfigFile, err = e2e.GenKubeConfigFile(serverNodeNames[0])
+		Expect(err).NotTo(HaveOccurred())
+	})
+
+	It("Checks Node Status", func() {
+		Eventually(func(g Gomega) {
+			nodes, err := e2e.ParseNodes(kubeConfigFile, false)
+			g.Expect(err).NotTo(HaveOccurred())
+			for _, node := range nodes {
+				g.Expect(node.Status).Should(Equal("Ready"))
+			}
+		}, "420s", "5s").Should(Succeed())
+		_, err := e2e.ParseNodes(kubeConfigFile, true)
+		Expect(err).NotTo(HaveOccurred())
+	})
+
+	It("Checks Pod Status", func() {
+		Eventually(func(g Gomega) {
+			pods, err := e2e.ParsePods(kubeConfigFile, false)
+			g.Expect(err).NotTo(HaveOccurred())
+			for _, pod := range pods {
+				if strings.Contains(pod.Name, "helm-install") {
+					g.Expect(pod.Status).Should(Equal("Completed"), pod.Name)
+				} else {
+					g.Expect(pod.Status).Should(Equal("Running"), pod.Name)
+				}
+			}
+		}, "420s", "5s").Should(Succeed())
+		_, err := e2e.ParsePods(kubeConfigFile, true)
+		Expect(err).NotTo(HaveOccurred())
+	})
+	It("Verifies internode connectivity over BGP", func() {
+		_, err := e2e.DeployWorkload("pod_client.yaml", kubeConfigFile)
+		Expect(err).NotTo(HaveOccurred())
+
+		_, err = e2e.DeployWorkload("windows_app_deployment.yaml", kubeConfigFile)
+		Expect(err).NotTo(HaveOccurred())
+
+		// Wait for the pod_client pods to have an IP
+		Eventually(func() string {
+			ips, _ := e2e.PodIPsUsingLabel(kubeConfigFile, "app=client")
+			return ips[0]
+		}, "120s", "10s").Should(ContainSubstring("10.42"), "failed getClientIPs")
+
+		// Wait for the windows_app_deployment pods to have an IP (We must wait 250s because it takes time)
+		Eventually(func() string {
+			ips, _ := e2e.PodIPsUsingLabel(kubeConfigFile, "app=windows-app")
+			return ips[0]
+		}, "620s", "10s").Should(ContainSubstring("10.42"), "failed getClientIPs")
+
+		// Verify there are BGP routes
+		cmdRoute := "ip route"
+		for _, node := range append(serverNodeNames, linuxAgentNames...) {
+			output, err := e2e.RunCmdOnNode(cmdRoute, node)
+			Expect(err).NotTo(HaveOccurred())
+			Expect(output).Should(ContainSubstring("bird"))
+		}
+x
+		// Test Linux -> Windows communication
+		fmt.Println("Testing Linux -> Windows communication")
+		cmd := "kubectl exec svc/client-curl --kubeconfig=" + kubeConfigFile + " -- curl -m7 windows-app-svc:3000"
+		Eventually(func() (string, error) {
+			return e2e.RunCommand(cmd)
+		}, "120s", "3s").Should(ContainSubstring("Welcome to PSTools for K8s Debugging"), "failed cmd: "+cmd)
+
+		// Test Windows -> Linux communication
+		fmt.Println("Testing Windows -> Linux communication")
+		cmd = "kubectl exec svc/windows-app-svc --kubeconfig=" + kubeConfigFile + " -- curl -m7 client-curl:8080"
+		Eventually(func() (string, error) {
+			return e2e.RunCommand(cmd)
+		}, "20s", "3s").Should(ContainSubstring("Welcome to nginx!"), "failed cmd: "+cmd)
+
+	})
+	It("Runs the mixed os sonobuoy plugin", func() {
+		cmd := "sonobuoy run --kubeconfig=/etc/rancher/rke2/rke2.yaml --plugin my-sonobuoy-plugins/mixed-workload-e2e/mixed-workload-e2e.yaml --aggregator-node-selector kubernetes.io/os:linux --wait"
+		res, err := e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+		Expect(err).NotTo(HaveOccurred(), "failed output:"+res)
+		cmd = "sonobuoy retrieve --kubeconfig=/etc/rancher/rke2/rke2.yaml"
+		testResultTar, err := e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+		Expect(err).NotTo(HaveOccurred(), "failed cmd: "+cmd)
+		cmd = "sonobuoy results " + testResultTar
+		res, err = e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+		Expect(err).NotTo(HaveOccurred(), "failed cmd: "+cmd)
+		Expect(res).Should(ContainSubstring("Plugin: mixed-workload-e2e\nStatus: passed\n"))
+	})
+
+})
+
+var failed bool
+var _ = AfterEach(func() {
+	failed = failed || CurrentSpecReport().Failed()
+})
+
+var _ = AfterSuite(func() {
+	if failed && !*ci {
+		fmt.Println("FAILED!")
+	} else {
+		Expect(e2e.DestroyCluster()).To(Succeed())
+		Expect(os.Remove(kubeConfigFile)).To(Succeed())
+	}
+})

tests/e2e/scripts/calico_manifestbgp.sh

@@ -0,0 +1,23 @@
+#!/bin/bash
+ip4_addr=$1
+
+# Override default calico and specify the interface for windows-agent
+# by default, the windows-agent use a different interface name than the linux-agent
+mkdir -p /var/lib/rancher/rke2/server/manifests
+
+echo "Creating calico chart"
+echo "apiVersion: helm.cattle.io/v1
+kind: HelmChartConfig
+metadata:
+  name: rke2-calico
+  namespace: kube-system
+spec:
+  valuesContent: |-
+    installation:
+      calicoNetwork:
+        bgp: Enabled
+        ipPools:
+        - cidr: 10.42.0.0/16
+          encapsulation: None
+        nodeAddressAutodetectionV4:
+          canReach: $ip4_addr" >> /var/lib/rancher/rke2/server/manifests/e2e-calico.yaml