add e2e startup test for rootless k3s (#8383)

tests/e2e/rootless/Vagrantfile

+ENV['VAGRANT_NO_PARALLEL'] = 'no'
+NODE_ROLES = (ENV['E2E_NODE_ROLES'] ||  ["server-0"])
+NODE_BOXES = (ENV['E2E_NODE_BOXES'] ||  ['generic/ubuntu2204'])
+GITHUB_BRANCH = (ENV['E2E_GITHUB_BRANCH'] || "master")
+RELEASE_VERSION = (ENV['E2E_RELEASE_VERSION'] || "")
+GOCOVER = (ENV['E2E_GOCOVER'] || "")
+NODE_CPUS = (ENV['E2E_NODE_CPUS'] || 2).to_i
+NODE_MEMORY = (ENV['E2E_NODE_MEMORY'] || 2048).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
+  vm.network "private_network", ip: "#{NETWORK_PREFIX}.#{100+node_num}", netmask: "255.255.255.0"
+
+  vagrant_defaults = '../vagrantdefaults.rb'
+  scripts_location = Dir.exists?("./scripts") ? "./scripts" : "../scripts"
+  load vagrant_defaults if File.exists?(vagrant_defaults)
+
+  defaultOSConfigure(vm)
+  addCoverageDir(vm, role, GOCOVER)
+
+  vm.provision "shell", inline: "ping -c 2 k3s.io"
+  vm.provision 'k3s-install', type: 'k3s', run: 'once' do |k3s|
+    k3s.args = "server "
+    k3s.env = %W[K3S_KUBECONFIG_MODE=0644 INSTALL_K3S_SKIP_START=true INSTALL_K3S_SKIP_ENABLE=true]
+    k3s.env_owner = "vagrant:vagrant"
+    k3s.config_mode = '0644' # side-step https://github.com/k3s-io/k3s/issues/4321
+    k3s.config_owner = "vagrant:vagrant"
+    k3s.config = <<~YAML
+      token: vagrant
+      node-external-ip: #{NETWORK_PREFIX}.100
+    YAML
+  end
+
+  vm.provision "Download k3s-rootless and enable cgroup v2", type: "shell", path: scripts_location + "/setup_rootless.sh"
+  vm.provision 'k3s-reload', type: 'reload', run: 'once'
+end
+
+Vagrant.configure("2") do |config|
+  config.vagrant.plugins = ["vagrant-k3s", "vagrant-reload"]
+  # 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
+  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|
+    role_num = name.split("-", -1).pop.to_i
+    config.vm.define name do |node|
+      provision(node.vm, name, role_num, i)
+    end
+  end
+end

tests/e2e/rootless/rootless_test.go

+package rootless
+
+import (
+	"flag"
+	"fmt"
+	"os"
+	"strings"
+	"testing"
+
+	"github.com/k3s-io/k3s/tests/e2e"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+// Rootless is only valid on a single node, but requires node/kernel configuration, requiring a E2E test environment.
+
+// Valid nodeOS: generic/ubuntu2004, opensuse/Leap-15.3.x86_64
+var nodeOS = flag.String("nodeOS", "generic/ubuntu2204", "VM operating system")
+var ci = flag.Bool("ci", false, "running on CI")
+var local = flag.Bool("local", false, "deploy a locally built K3s binary")
+
+// Environment Variables Info:
+// E2E_RELEASE_VERSION=v1.27.1+k3s2 or nil for latest commit from master
+
+func Test_E2ERootlessStartupValidation(t *testing.T) {
+	RegisterFailHandler(Fail)
+	flag.Parse()
+	suiteConfig, reporterConfig := GinkgoConfiguration()
+	RunSpecs(t, "Startup Test Suite", suiteConfig, reporterConfig)
+}
+
+var (
+	kubeConfigFile  string
+	serverNodeNames []string
+)
+
+func StartK3sCluster(nodes []string, serverYAML string) error {
+	for _, node := range nodes {
+
+		resetCmd := "head -n 3 /etc/rancher/k3s/config.yaml > /tmp/config.yaml && sudo mv /tmp/config.yaml /etc/rancher/k3s/config.yaml"
+		yamlCmd := fmt.Sprintf("echo '%s' >> /etc/rancher/k3s/config.yaml", serverYAML)
+		startCmd := "systemctl --user restart k3s-rootless"
+
+		if _, err := e2e.RunCmdOnNode(resetCmd, node); err != nil {
+			return err
+		}
+		if _, err := e2e.RunCmdOnNode(yamlCmd, node); err != nil {
+			return err
+		}
+		if _, err := RunCmdOnRootlesNode("systemctl --user daemon-reload", node); err != nil {
+			return err
+		}
+		if _, err := RunCmdOnRootlesNode(startCmd, node); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func KillK3sCluster(nodes []string) error {
+	for _, node := range nodes {
+		if _, err := RunCmdOnRootlesNode(`systemctl --user stop k3s-rootless`, node); err != nil {
+			return err
+		}
+		if _, err := RunCmdOnRootlesNode("k3s-killall.sh", node); err != nil {
+			return err
+		}
+		if _, err := RunCmdOnRootlesNode("rm -rf /home/vagrant/.rancher/k3s/server/db", node); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+var _ = ReportAfterEach(e2e.GenReport)
+
+var _ = BeforeSuite(func() {
+	var err error
+	if *local {
+		serverNodeNames, _, err = e2e.CreateLocalCluster(*nodeOS, 1, 0)
+	} else {
+		serverNodeNames, _, err = e2e.CreateCluster(*nodeOS, 1, 0)
+	}
+	Expect(err).NotTo(HaveOccurred(), e2e.GetVagrantLog(err))
+	//Checks if system is using cgroup v2
+	_, err = e2e.RunCmdOnNode("cat /sys/fs/cgroup/cgroup.controllers", serverNodeNames[0])
+	Expect(err).NotTo(HaveOccurred(), e2e.GetVagrantLog(err))
+
+})
+
+var _ = Describe("Various Startup Configurations", Ordered, func() {
+	Context("Verify standard startup :", func() {
+		It("Starts K3s with no issues", func() {
+			err := StartK3sCluster(serverNodeNames, "")
+			Expect(err).NotTo(HaveOccurred(), e2e.GetVagrantLog(err))
+
+			fmt.Println("CLUSTER CONFIG")
+			fmt.Println("OS:", *nodeOS)
+			fmt.Println("Server Nodes:", serverNodeNames)
+			kubeConfigFile, err = GenRootlessKubeConfigFile(serverNodeNames[0])
+			Expect(err).NotTo(HaveOccurred())
+		})
+
+		It("Checks node and pod status", func() {
+			fmt.Printf("\nFetching node status\n")
+			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"))
+				}
+			}, "360s", "5s").Should(Succeed())
+			_, _ = e2e.ParseNodes(kubeConfigFile, false)
+
+			fmt.Printf("\nFetching pods status\n")
+			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)
+					}
+				}
+			}, "360s", "5s").Should(Succeed())
+			_, _ = e2e.ParsePods(kubeConfigFile, false)
+		})
+
+		It("Returns pod metrics", func() {
+			cmd := "kubectl top pod -A"
+			Eventually(func() error {
+				_, err := e2e.RunCommand(cmd)
+				return err
+			}, "600s", "5s").Should(Succeed())
+		})
+
+		It("Returns node metrics", func() {
+			cmd := "kubectl top node"
+			_, err := e2e.RunCommand(cmd)
+			Expect(err).NotTo(HaveOccurred())
+		})
+
+		It("Runs an interactive command a pod", func() {
+			cmd := "kubectl run busybox --rm -it --restart=Never --image=rancher/mirrored-library-busybox:1.34.1 -- uname -a"
+			_, err := e2e.RunCommand(cmd)
+			Expect(err).NotTo(HaveOccurred())
+		})
+
+		It("Collects logs from a pod", func() {
+			cmd := "kubectl logs -n kube-system -l app.kubernetes.io/name=traefik -c traefik"
+			_, err := e2e.RunCommand(cmd)
+			Expect(err).NotTo(HaveOccurred())
+		})
+
+		It("Kills the cluster", func() {
+			err := KillK3sCluster(serverNodeNames)
+			Expect(err).NotTo(HaveOccurred())
+		})
+	})
+
+})
+
+var failed bool
+var _ = AfterEach(func() {
+	failed = failed || CurrentSpecReport().Failed()
+})
+
+var _ = AfterSuite(func() {
+	if failed && !*ci {
+		fmt.Println("FAILED!")
+	} else {
+		Expect(e2e.GetCoverageReport(serverNodeNames)).To(Succeed())
+		Expect(e2e.DestroyCluster()).To(Succeed())
+		Expect(os.Remove(kubeConfigFile)).To(Succeed())
+	}
+})

tests/e2e/rootless/rootless_utils.go

+package rootless
+
+import (
+	"fmt"
+	"os"
+	"strings"
+
+	"github.com/k3s-io/k3s/tests/e2e"
+)
+
+// RunCmdOnRootlesNode executes a command from within the given node as user vagrant
+func RunCmdOnRootlesNode(cmd string, nodename string) (string, error) {
+	injectEnv := ""
+	if _, ok := os.LookupEnv("E2E_GOCOVER"); ok && strings.HasPrefix(cmd, "k3s") {
+		injectEnv = "GOCOVERDIR=/tmp/k3scov "
+	}
+	runcmd := "vagrant ssh " + nodename + " -c \"" + injectEnv + cmd + "\""
+	out, err := e2e.RunCommand(runcmd)
+	if err != nil {
+		return out, fmt.Errorf("failed to run command: %s on node %s: %s, %v", cmd, nodename, out, err)
+	}
+	return out, nil
+}
+
+func GenRootlessKubeConfigFile(serverName string) (string, error) {
+	kubeConfig, err := RunCmdOnRootlesNode("cat /home/vagrant/.kube/k3s.yaml", serverName)
+	if err != nil {
+		return "", err
+	}
+	nodeIP, err := e2e.FetchNodeExternalIP(serverName)
+	if err != nil {
+		return "", err
+	}
+	kubeConfig = strings.Replace(kubeConfig, "127.0.0.1", nodeIP, 1)
+	kubeConfigFile := fmt.Sprintf("kubeconfig-%s", serverName)
+	if err := os.WriteFile(kubeConfigFile, []byte(kubeConfig), 0644); err != nil {
+		return "", err
+	}
+	if err := os.Setenv("E2E_KUBECONFIG", kubeConfigFile); err != nil {
+		return "", err
+	}
+	return kubeConfigFile, nil
+}

tests/e2e/scripts/setup_rootless.sh

+#!/bin/sh
+
+# GitHub repository URL
+github_url="https://raw.githubusercontent.com/k3s-io/k3s/master/k3s-rootless.service"
+
+# Destination file path
+destination_path="/home/vagrant/.config/systemd/user/"
+
+# Download the file from GitHub using curl
+curl -LJO "$github_url"
+
+# Check if the download was successful
+if [ $? -eq 0 ]; then
+    # Move the downloaded file to the desired destination
+    mkdir -p "$destination_path"
+    mv "k3s-rootless.service" "$destination_path/k3s-rootless.service"
+    # Add EnvironmentFile=-/etc/k3s-rootless.env to the service file
+    sed -i 's/^\[Service\]$/\[Service\]\nEnvironmentFile=-\/etc\/rancher\/k3s\/k3s.env/' "$destination_path/k3s-rootless.service"
+    chown -R vagrant:vagrant /home/vagrant/.config/
+
+    echo "File downloaded and moved to $destination_path"
+else
+    echo "Failed to download the file from GitHub."
+fi
+
+
+# Enable IPv4 forwarding
+echo "net.ipv4.ip_forward=1" >> /etc/sysctl.conf
+sysctl --system
+
+
+# Check if the string is already in GRUB_CMDLINE_LINUX
+if grep -qxF "GRUB_CMDLINE_LINUX=\"systemd.unified_cgroup_hierarchy=1 \"" /etc/default/grub; then
+    echo "String is already in GRUB_CMDLINE_LINUX. No changes made."
+else
+    # Add the string to GRUB_CMDLINE_LINUX
+    sed -i "s/\(GRUB_CMDLINE_LINUX=\)\"\(.*\)\"/\1\"systemd.unified_cgroup_hierarchy=1 \2\"/" /etc/default/grub
+
+    # Update GRUB
+    update-grub
+
+    echo "String 'systemd.unified_cgroup_hierarchy=1' added to GRUB_CMDLINE_LINUX and GRUB updated successfully."
+fi
+
+mkdir -p /etc/systemd/system/user@.service.d
+echo "[Service]
+Delegate=cpu cpuset io memory pids
+">> /etc/systemd/system/user@.service.d/delegate.conf
+apt-get install -y uidmap
+
+systemctl daemon-reload
+loginctl enable-linger vagrant
+# We need to run this as vagrant user, because rootless k3s will be run as vagrant user
+su -c 'XDG_RUNTIME_DIR="/run/user/$UID" DBUS_SESSION_BUS_ADDRESS="unix:path=${XDG_RUNTIME_DIR}/bus" systemctl --user daemon-reload' vagrant
+su -c 'XDG_RUNTIME_DIR="/run/user/$UID" DBUS_SESSION_BUS_ADDRESS="unix:path=${XDG_RUNTIME_DIR}/bus" systemctl --user enable --now k3s-rootless' vagrant
+