Consolidate E2E tests and GH Actions (#6772)

.github/workflows/install.yaml

@@ -6,11 +6,14 @@ on:
       - "channel.yaml"
       - "install.sh"
       - "tests/install/**"
+      - ".github/workflows/install.yaml"
+
   pull_request:
     branches: [main, master]
     paths:
       - "install.sh"
       - "tests/install/**"
+      - ".github/workflows/install.yaml"
   workflow_dispatch: {}
 
 permissions:
@@ -18,21 +21,7 @@ permissions:
   
 jobs:
   build:
-    name: Build
-    runs-on: ubuntu-20.04
-    timeout-minutes: 20
-    steps:
-    - name: "Checkout"
-      uses: actions/checkout@v3
-      with:
-        fetch-depth: 1
-    - name: "Make"
-      run: DOCKER_BUILDKIT=1 SKIP_VALIDATE=1 make
-    - name: "Upload k3s binary"
-      uses: actions/upload-artifact@v3
-      with:
-        name: k3s
-        path: ./dist/artifacts/k3s
+    uses: ./.github/workflows/build-k3s.yaml
   test:
     name: "Smoke Test"
     needs: build
@@ -41,7 +30,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        vm: [centos-7, rocky-8, fedora, opensuse-leap, opensuse-microos, ubuntu-focal]
+        vm: [centos-7, rocky-8, fedora, opensuse-leap, ubuntu-focal]
       max-parallel: 2
     defaults:
       run:

.github/workflows/nightly-install.yaml

@@ -16,7 +16,7 @@ jobs:
       fail-fast: false
       matrix:
         channel: [stable]
-        vm: [centos-7, rocky-8, fedora, opensuse-leap, opensuse-microos, ubuntu-focal]
+        vm: [centos-7, rocky-8, fedora, opensuse-leap, ubuntu-focal]
         include:
           - {channel: latest, vm: rocky-8}
           - {channel: latest, vm: ubuntu-focal}

.github/workflows/snapshotter.yaml

@@ -24,22 +24,11 @@ permissions:
   contents: read
 
 jobs:
-  prep:
-    name: "Prepare"
-    runs-on: ubuntu-20.04
-    timeout-minutes: 40
-    steps:
-      - name: "Checkout"
-        uses: actions/checkout@v3
-        with: { fetch-depth: 1 }
-      - name: "Build"
-        run: DOCKER_BUILDKIT=1 SKIP_VALIDATE=1 make
-      - name: "Upload Binary"
-        uses: actions/upload-artifact@v3
-        with: { name: k3s, path: dist/artifacts/k3s }
+  build:
+    uses: ./.github/workflows/build-k3s.yaml
   test:
     name: "Smoke Test"
-    needs: prep
+    needs: build
     # nested virtualization is only available on macOS hosts
     runs-on: macos-12
     timeout-minutes: 40

.github/workflows/updatecli.yaml

@@ -8,13 +8,15 @@ on:
   workflow_dispatch:
 
 permissions:
-  contents: write
-  issues: write
-  pull-requests: write
+  contents: read
 
 jobs:
   updatecli:
     runs-on: ubuntu-latest
+    permissions:
+      contents: write
+      issues: write
+      pull-requests: write
     if: github.ref == 'refs/heads/master'
     steps:
       - name: Checkout

tests/e2e/clusterreset/Vagrantfile

-ENV['VAGRANT_NO_PARALLEL'] = 'no'
-NODE_ROLES = (ENV['E2E_NODE_ROLES'] ||
-  ["server-0", "server-1", "server-2", "agent-0", "agent-1"])
-NODE_BOXES = (ENV['E2E_NODE_BOXES'] ||
-  ['generic/ubuntu2004', 'generic/ubuntu2004', 'generic/ubuntu2004', 'generic/ubuntu2004', 'generic/ubuntu2004'])
-GITHUB_BRANCH = (ENV['E2E_GITHUB_BRANCH'] || "master")
-RELEASE_VERSION = (ENV['E2E_RELEASE_VERSION'] || "")
-EXTERNAL_DB = (ENV['E2E_EXTERNAL_DB'] || "etcd")
-HARDENED = (ENV['E2E_HARDENED'] || "")
-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 = ""
-hardened_arg = ""
-
-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"
-
-  scripts_location = Dir.exists?("./scripts") ? "./scripts" : "../scripts" 
-  vagrant_defaults = File.exists?("./vagrantdefaults.rb") ? "./vagrantdefaults.rb" : "../vagrantdefaults.rb"
-  load vagrant_defaults
-
-  defaultOSConfigure(vm)
-  install_type = getInstallType(vm, RELEASE_VERSION, GITHUB_BRANCH)
-  
-  vm.provision "shell", inline: "ping -c 2 k3s.io"
-  
-  db_type = getDBType(role, role_num, vm)
-
-  if !HARDENED.empty?
-    vm.provision "Set kernel parameters", type: "shell", path: scripts_location + "/harden.sh"
-    hardened_arg = "protect-kernel-defaults: true\nkube-apiserver-arg: \"enable-admission-plugins=NodeRestriction,PodSecurityPolicy,ServiceAccount\""
-  end
-
-  if role.include?("server") && role_num == 0
-    vm.provision 'k3s-primary-server', type: 'k3s', run: 'once' do |k3s|
-      k3s.args = "server "
-      k3s.config = <<~YAML
-        token: vagrant
-        node-external-ip: #{NETWORK_PREFIX}.100
-        flannel-iface: eth1
-        tls-san: #{NETWORK_PREFIX}.100.nip.io
-        #{db_type}
-        #{hardened_arg}
-      YAML
-      k3s.env = %W[K3S_KUBECONFIG_MODE=0644 #{install_type}]
-      k3s.config_mode = '0644' # side-step https://github.com/k3s-io/k3s/issues/4321
-    end
-  
-  elsif role.include?("server") && role_num != 0
-    vm.provision 'k3s-secondary-server', type: 'k3s', run: 'once' do |k3s|
-      k3s.args = "server"
-      k3s.config = <<~YAML
-        server: "https://#{NETWORK_PREFIX}.100:6443"
-        token: vagrant
-        node-external-ip: #{node_ip}
-        flannel-iface: eth1
-        #{db_type}
-        #{hardened_arg}
-      YAML
-      k3s.env = %W[K3S_KUBECONFIG_MODE=0644 K3S_TOKEN=vagrant #{install_type}]
-      k3s.config_mode = '0644' # side-step https://github.com/k3s-io/k3s/issues/4321
-    end
-  end
-  
-  if role.include?("agent")
-    vm.provision 'k3s-agent', type: 'k3s', run: 'once' do |k3s|
-      k3s.args = "agent"
-      k3s.config = <<~YAML
-        server: "https://#{NETWORK_PREFIX}.100:6443"
-        token: vagrant
-        node-external-ip: #{node_ip}
-        flannel-iface: eth1
-        #{db_type}
-        #{hardened_arg}
-      YAML
-      k3s.env = %W[K3S_KUBECONFIG_MODE=0644 #{install_type}]
-      k3s.config_mode = '0644' # side-step https://github.com/k3s-io/k3s/issues/4321
-    end
-  end
-  if vm.box.to_s.include?("microos")
-    vm.provision 'k3s-reload', type: 'reload', run: 'once'
-    if !EXTERNAL_DB.empty?
-      vm.provision "shell", inline: "docker start #{EXTERNAL_DB}"
-    end
-  end
-end
-
-def getDBType(role, role_num, vm)
-  if ( EXTERNAL_DB == "" || EXTERNAL_DB == "etcd" )
-    if role.include?("server") && role_num == 0
-      return "cluster-init: true"
-    end
-  elsif ( EXTERNAL_DB == "none" )
-    # Use internal sqlite, only valid for single node clusters
-  else
-    puts "Unknown EXTERNAL_DB: " + EXTERNAL_DB
-    abort
-  end
-  return ""
-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
-
-  # Must iterate on the index, vagrant does not understand iterating 
-  # over the node roles themselves
-  NODE_ROLES.length.times do |i|
-    name = NODE_ROLES[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/clusterreset/clusterreset_test.go

-package clusterreset
-
-import (
-	"flag"
-	"fmt"
-	"os"
-	"strings"
-	"testing"
-
-	"github.com/k3s-io/k3s/tests/e2e"
-	. "github.com/onsi/ginkgo/v2"
-	. "github.com/onsi/gomega"
-)
-
-// Valid nodeOS:
-// generic/ubuntu2004, generic/centos7, generic/rocky8,
-// opensuse/Leap-15.3.x86_64, dweomer/microos.amd64
-var nodeOS = flag.String("nodeOS", "generic/ubuntu2004", "VM operating system")
-var serverCount = flag.Int("serverCount", 3, "number of server nodes")
-var agentCount = flag.Int("agentCount", 1, "number of agent nodes")
-var hardened = flag.Bool("hardened", false, "true or false")
-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_EXTERNAL_DB: mysql, postgres, etcd (default: etcd)
-// E2E_RELEASE_VERSION=v1.23.1+k3s2 (default: latest commit from master)
-
-func Test_E2EClusterReset(t *testing.T) {
-	RegisterFailHandler(Fail)
-	flag.Parse()
-	suiteConfig, reporterConfig := GinkgoConfiguration()
-	RunSpecs(t, "ClusterReset Test Suite", suiteConfig, reporterConfig)
-}
-
-var (
-	kubeConfigFile  string
-	serverNodeNames []string
-	agentNodeNames  []string
-)
-
-var _ = ReportAfterEach(e2e.GenReport)
-
-var _ = Describe("Verify Create", Ordered, func() {
-	Context("Cluster :", func() {
-		It("Starts up with no issues", func() {
-			var err error
-			if *local {
-				serverNodeNames, agentNodeNames, err = e2e.CreateLocalCluster(*nodeOS, *serverCount, *agentCount)
-			} else {
-				serverNodeNames, agentNodeNames, err = e2e.CreateCluster(*nodeOS, *serverCount, *agentCount)
-			}
-			Expect(err).NotTo(HaveOccurred(), e2e.GetVagrantLog(err))
-			fmt.Println("CLUSTER CONFIG")
-			fmt.Println("OS:", *nodeOS)
-			fmt.Println("Server Nodes:", serverNodeNames)
-			fmt.Println("Agent Nodes:", agentNodeNames)
-			kubeConfigFile, err = e2e.GenKubeConfigFile(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"))
-				}
-			}, "620s", "5s").Should(Succeed())
-			_, _ = e2e.ParseNodes(kubeConfigFile, true)
-
-			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)
-					}
-				}
-			}, "620s", "5s").Should(Succeed())
-			_, _ = e2e.ParsePods(kubeConfigFile, true)
-		})
-
-		It("Verifies ClusterReset Functionality", func() {
-			Eventually(func(g Gomega) {
-				for _, nodeName := range serverNodeNames {
-					if nodeName != "server-0" {
-						cmd := "sudo systemctl stop k3s"
-						_, err := e2e.RunCmdOnNode(cmd, nodeName)
-						Expect(err).NotTo(HaveOccurred())
-					}
-				}
-
-				cmd := "sudo systemctl stop k3s"
-				_, err := e2e.RunCmdOnNode(cmd, "server-0")
-				Expect(err).NotTo(HaveOccurred())
-
-				cmd = "sudo k3s server --cluster-reset"
-				res, err := e2e.RunCmdOnNode(cmd, "server-0")
-				Expect(err).NotTo(HaveOccurred())
-				Expect(res).Should(ContainSubstring("Managed etcd cluster membership has been reset, restart without --cluster-reset flag now"))
-
-				cmd = "sudo systemctl start k3s"
-				_, err = e2e.RunCmdOnNode(cmd, "server-0")
-				Expect(err).NotTo(HaveOccurred())
-
-				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 {
-						if strings.Contains(node.Name, "server-0") || strings.Contains(node.Name, "agent-") {
-							g.Expect(node.Status).Should(Equal("Ready"))
-						} else {
-							g.Expect(node.Status).Should(Equal("NotReady"))
-						}
-					}
-				}, "480s", "5s").Should(Succeed())
-				_, _ = e2e.ParseNodes(kubeConfigFile, true)
-
-				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)
-						}
-					}
-				}, "420s", "5s").Should(Succeed())
-				_, _ = e2e.ParsePods(kubeConfigFile, true)
-				for _, nodeName := range serverNodeNames {
-					if nodeName != "server-0" {
-						cmd := "sudo rm -rf /var/lib/rancher/k3s/server/db"
-						_, err := e2e.RunCmdOnNode(cmd, nodeName)
-						Expect(err).NotTo(HaveOccurred())
-						cmd = "sudo systemctl restart k3s"
-						_, err = e2e.RunCmdOnNode(cmd, nodeName)
-						Expect(err).NotTo(HaveOccurred())
-					}
-				}
-				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())
-				_, _ = e2e.ParseNodes(kubeConfigFile, true)
-
-				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)
-						}
-					}
-				}, "420s", "5s").Should(Succeed())
-				_, _ = e2e.ParsePods(kubeConfigFile, true)
-			}, "240s", "5s").Should(Succeed())
-		})
-	})
-})
-
-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/docker/docker_test.go

@@ -12,7 +12,7 @@ import (
 	. "github.com/onsi/gomega"
 )
 
-// Valid nodeOS: generic/ubuntu2004, opensuse/Leap-15.3.x86_64, dweomer/microos.amd64
+// Valid nodeOS: generic/ubuntu2004, opensuse/Leap-15.3.x86_64
 var nodeOS = flag.String("nodeOS", "generic/ubuntu2004", "VM operating system")
 var serverCount = flag.Int("serverCount", 1, "number of server nodes")
 var agentCount = flag.Int("agentCount", 1, "number of agent nodes")

tests/e2e/preferbundled/preferbundled_test.go

@@ -12,7 +12,7 @@ import (
 	. "github.com/onsi/gomega"
 )
 
-// Valid nodeOS: generic/ubuntu2004, opensuse/Leap-15.3.x86_64, dweomer/microos.amd64
+// Valid nodeOS: generic/ubuntu2004, opensuse/Leap-15.3.x86_64
 var nodeOS = flag.String("nodeOS", "generic/ubuntu2004", "VM operating system")
 var serverCount = flag.Int("serverCount", 1, "number of server nodes")
 var agentCount = flag.Int("agentCount", 1, "number of agent nodes")

tests/e2e/secretsencryption/secretsencryption_test.go

@@ -12,7 +12,7 @@ import (
 	. "github.com/onsi/gomega"
 )
 
-// Valid nodeOS: generic/ubuntu2004, opensuse/Leap-15.3.x86_64, dweomer/microos.amd64
+// Valid nodeOS: generic/ubuntu2004, opensuse/Leap-15.3.x86_64
 var nodeOS = flag.String("nodeOS", "generic/ubuntu2004", "VM operating system")
 var serverCount = flag.Int("serverCount", 3, "number of server nodes")
 var hardened = flag.Bool("hardened", false, "true or false")
@@ -36,7 +36,7 @@ var (
 var _ = ReportAfterEach(e2e.GenReport)
 
 var _ = Describe("Verify Secrets Encryption Rotation", Ordered, func() {
-	Context("Cluster :", func() {
+	Context("Secrets Keys are rotated:", func() {
 		It("Starts up with no issues", func() {
 			var err error
 			serverNodeNames, _, err = e2e.CreateCluster(*nodeOS, *serverCount, 0)
@@ -219,73 +219,78 @@ var _ = Describe("Verify Secrets Encryption Rotation", Ordered, func() {
 		})
 	})
 
-	It("Disables encryption", func() {
-		cmd := "sudo k3s secrets-encrypt disable"
-		res, err := e2e.RunCmdOnNode(cmd, serverNodeNames[0])
-		Expect(err).NotTo(HaveOccurred(), res)
+	Context("Disabling Secrets-Encryption", func() {
+		It("Disables encryption", func() {
+			cmd := "sudo k3s secrets-encrypt disable"
+			res, err := e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+			Expect(err).NotTo(HaveOccurred(), res)
 
-		cmd = "sudo k3s secrets-encrypt reencrypt -f --skip"
-		res, err = e2e.RunCmdOnNode(cmd, serverNodeNames[0])
-		Expect(err).NotTo(HaveOccurred(), res)
+			cmd = "sudo k3s secrets-encrypt reencrypt -f --skip"
+			res, err = e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+			Expect(err).NotTo(HaveOccurred(), res)
 
-		cmd = "sudo k3s secrets-encrypt status"
-		Eventually(func() (string, error) {
-			return e2e.RunCmdOnNode(cmd, serverNodeNames[0])
-		}, "180s", "5s").Should(ContainSubstring("Current Rotation Stage: reencrypt_finished"))
+			cmd = "sudo k3s secrets-encrypt status"
+			Eventually(func() (string, error) {
+				return e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+			}, "180s", "5s").Should(ContainSubstring("Current Rotation Stage: reencrypt_finished"))
 
-		for i, nodeName := range serverNodeNames {
-			Eventually(func(g Gomega) {
-				res, err := e2e.RunCmdOnNode(cmd, nodeName)
-				g.Expect(err).NotTo(HaveOccurred(), res)
-				if i == 0 {
-					g.Expect(res).Should(ContainSubstring("Encryption Status: Disabled"))
-				} else {
-					g.Expect(res).Should(ContainSubstring("Encryption Status: Enabled"))
-				}
-			}, "420s", "2s").Should(Succeed())
-		}
-	})
+			for i, nodeName := range serverNodeNames {
+				Eventually(func(g Gomega) {
+					res, err := e2e.RunCmdOnNode(cmd, nodeName)
+					g.Expect(err).NotTo(HaveOccurred(), res)
+					if i == 0 {
+						g.Expect(res).Should(ContainSubstring("Encryption Status: Disabled"))
+					} else {
+						g.Expect(res).Should(ContainSubstring("Encryption Status: Enabled"))
+					}
+				}, "420s", "2s").Should(Succeed())
+			}
+		})
 
-	It("Restarts K3s servers", func() {
-		Expect(e2e.RestartCluster(serverNodeNames)).To(Succeed())
-	})
+		It("Restarts K3s servers", func() {
+			Expect(e2e.RestartCluster(serverNodeNames)).To(Succeed())
+		})
+
+		It("Verifies encryption disabled on all nodes", func() {
+			cmd := "sudo k3s secrets-encrypt status"
+			for _, nodeName := range serverNodeNames {
+				Eventually(func(g Gomega) {
+					g.Expect(e2e.RunCmdOnNode(cmd, nodeName)).Should(ContainSubstring("Encryption Status: Disabled"))
+				}, "420s", "2s").Should(Succeed())
+			}
+		})
 
-	It("Verifies encryption disabled on all nodes", func() {
-		cmd := "sudo k3s secrets-encrypt status"
-		for _, nodeName := range serverNodeNames {
-			Eventually(func(g Gomega) {
-				g.Expect(e2e.RunCmdOnNode(cmd, nodeName)).Should(ContainSubstring("Encryption Status: Disabled"))
-			}, "420s", "2s").Should(Succeed())
-		}
 	})
 
-	It("Enables encryption", func() {
-		cmd := "sudo k3s secrets-encrypt enable"
-		res, err := e2e.RunCmdOnNode(cmd, serverNodeNames[0])
-		Expect(err).NotTo(HaveOccurred(), res)
+	Context("Enabling Secrets-Encryption", func() {
+		It("Enables encryption", func() {
+			cmd := "sudo k3s secrets-encrypt enable"
+			res, err := e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+			Expect(err).NotTo(HaveOccurred(), res)
 
-		cmd = "sudo k3s secrets-encrypt reencrypt -f --skip"
-		res, err = e2e.RunCmdOnNode(cmd, serverNodeNames[0])
-		Expect(err).NotTo(HaveOccurred(), res)
+			cmd = "sudo k3s secrets-encrypt reencrypt -f --skip"
+			res, err = e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+			Expect(err).NotTo(HaveOccurred(), res)
 
-		cmd = "sudo k3s secrets-encrypt status"
-		Eventually(func() (string, error) {
-			return e2e.RunCmdOnNode(cmd, serverNodeNames[0])
-		}, "180s", "5s").Should(ContainSubstring("Current Rotation Stage: reencrypt_finished"))
-	})
+			cmd = "sudo k3s secrets-encrypt status"
+			Eventually(func() (string, error) {
+				return e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+			}, "180s", "5s").Should(ContainSubstring("Current Rotation Stage: reencrypt_finished"))
+		})
 
-	It("Restarts K3s servers", func() {
-		Expect(e2e.RestartCluster(serverNodeNames)).To(Succeed())
-	})
+		It("Restarts K3s servers", func() {
+			Expect(e2e.RestartCluster(serverNodeNames)).To(Succeed())
+		})
 
-	It("Verifies encryption enabled on all nodes", func() {
-		cmd := "sudo k3s secrets-encrypt status"
-		for _, nodeName := range serverNodeNames {
-			Eventually(func(g Gomega) {
-				g.Expect(e2e.RunCmdOnNode(cmd, nodeName)).Should(ContainSubstring("Encryption Status: Enabled"))
-			}, "420s", "2s").Should(Succeed())
-		}
+		It("Verifies encryption enabled on all nodes", func() {
+			cmd := "sudo k3s secrets-encrypt status"
+			for _, nodeName := range serverNodeNames {
+				Eventually(func(g Gomega) {
+					g.Expect(e2e.RunCmdOnNode(cmd, nodeName)).Should(ContainSubstring("Encryption Status: Enabled"))
+				}, "420s", "2s").Should(Succeed())
+			}
 
+		})
 	})
 
 })

tests/e2e/snapshotrestore/snapshotrestore_test.go

@@ -14,7 +14,7 @@ import (
 
 // Valid nodeOS:
 // generic/ubuntu2004, generic/centos7, generic/rocky8,
-// opensuse/Leap-15.3.x86_64, dweomer/microos.amd64
+// opensuse/Leap-15.3.x86_64
 
 var nodeOS = flag.String("nodeOS", "generic/ubuntu2004", "VM operating system")
 var serverCount = flag.Int("serverCount", 3, "number of server nodes")
@@ -43,7 +43,7 @@ var (
 
 var _ = ReportAfterEach(e2e.GenReport)
 
-var _ = Describe("Verify Create", Ordered, func() {
+var _ = Describe("Verify snapshots and cluster restores work", Ordered, func() {
 	Context("Cluster :", func() {
 		It("Starts up with no issues", func() {
 			var err error
@@ -123,26 +123,139 @@ var _ = Describe("Verify Create", Ordered, func() {
 			}, "240s", "5s").Should(Succeed())
 		})
 
-		It("Verifies snapshot is restored successfully and validates only test workload1 is present", func() {
+		It("Resets the cluster", func() {
+			for _, nodeName := range serverNodeNames {
+				cmd := "sudo systemctl stop k3s"
+				Expect(e2e.RunCmdOnNode(cmd, nodeName)).Error().NotTo(HaveOccurred())
+				if nodeName != serverNodeNames[0] {
+					cmd = "k3s-killall.sh"
+					Expect(e2e.RunCmdOnNode(cmd, nodeName)).Error().NotTo(HaveOccurred())
+				}
+			}
+
+			cmd := "sudo k3s server --cluster-reset"
+			res, err := e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+			Expect(err).NotTo(HaveOccurred())
+			Expect(res).Should(ContainSubstring("Managed etcd cluster membership has been reset, restart without --cluster-reset flag now"))
+
+			cmd = "sudo systemctl start k3s"
+			Expect(e2e.RunCmdOnNode(cmd, serverNodeNames[0])).Error().NotTo(HaveOccurred())
+		})
+
+		It("Checks that other servers are not ready", 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 {
+					if strings.Contains(node.Name, serverNodeNames[0]) || strings.Contains(node.Name, "agent-") {
+						g.Expect(node.Status).Should(Equal("Ready"))
+					} else {
+						g.Expect(node.Status).Should(Equal("NotReady"))
+					}
+				}
+			}, "240s", "5s").Should(Succeed())
+			_, _ = e2e.ParseNodes(kubeConfigFile, true)
+		})
+
+		It("Rejoins other servers to cluster", func() {
+			// We must remove the db directory on the other servers before restarting k3s
+			// otherwise the nodes may join the old cluster
+			for _, nodeName := range serverNodeNames[1:] {
+				cmd := "sudo rm -rf /var/lib/rancher/k3s/server/db"
+				Expect(e2e.RunCmdOnNode(cmd, nodeName)).Error().NotTo(HaveOccurred())
+			}
+
+			for _, nodeName := range serverNodeNames[1:] {
+				cmd := "sudo systemctl start k3s"
+				Expect(e2e.RunCmdOnNode(cmd, nodeName)).Error().NotTo(HaveOccurred())
+			}
+		})
+
+		It("Checks that all nodes and pods are ready", 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())
+
+			_, _ = e2e.ParseNodes(kubeConfigFile, true)
+
+			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)
+					}
+				}
+			}, "420s", "5s").Should(Succeed())
+		})
+		It("Verifies that workload1 and workload1 exist", func() {
+			cmd := "kubectl get pods --kubeconfig=" + kubeConfigFile
+			res, err := e2e.RunCommand(cmd)
+			Expect(err).NotTo(HaveOccurred())
+			Expect(res).Should(ContainSubstring("test-clusterip"))
+			Expect(res).Should(ContainSubstring("test-nodeport"))
+		})
+
+		It("Restores the snapshot", func() {
 			//Stop k3s on all nodes
 			for _, nodeName := range serverNodeNames {
 				cmd := "sudo systemctl stop k3s"
-				_, err := e2e.RunCmdOnNode(cmd, nodeName)
-				Expect(err).NotTo(HaveOccurred())
+				Expect(e2e.RunCmdOnNode(cmd, nodeName)).Error().NotTo(HaveOccurred())
+				if nodeName != serverNodeNames[0] {
+					cmd = "k3s-killall.sh"
+					Expect(e2e.RunCmdOnNode(cmd, nodeName)).Error().NotTo(HaveOccurred())
+				}
 			}
 			//Restores from snapshot on server-0
-			for _, nodeName := range serverNodeNames {
-				if nodeName == "server-0" {
-					cmd := "sudo k3s server --cluster-init --cluster-reset --cluster-reset-restore-path=/var/lib/rancher/k3s/server/db/snapshots/" + snapshotname
-					res, err := e2e.RunCmdOnNode(cmd, nodeName)
-					Expect(err).NotTo(HaveOccurred())
-					Expect(res).Should(ContainSubstring("Managed etcd cluster membership has been reset, restart without --cluster-reset flag now"))
-
-					cmd = "sudo systemctl start k3s"
-					_, err = e2e.RunCmdOnNode(cmd, nodeName)
-					Expect(err).NotTo(HaveOccurred())
+			cmd := "sudo k3s server --cluster-init --cluster-reset --cluster-reset-restore-path=/var/lib/rancher/k3s/server/db/snapshots/" + snapshotname
+			res, err := e2e.RunCmdOnNode(cmd, serverNodeNames[0])
+			Expect(err).NotTo(HaveOccurred())
+			Expect(res).Should(ContainSubstring("Managed etcd cluster membership has been reset, restart without --cluster-reset flag now"))
+
+			cmd = "sudo systemctl start k3s"
+			Expect(e2e.RunCmdOnNode(cmd, serverNodeNames[0])).Error().NotTo(HaveOccurred())
+
+		})
+
+		It("Checks that other servers are not ready", 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 {
+					if strings.Contains(node.Name, serverNodeNames[0]) || strings.Contains(node.Name, "agent-") {
+						g.Expect(node.Status).Should(Equal("Ready"))
+					} else {
+						g.Expect(node.Status).Should(Equal("NotReady"))
+					}
 				}
+			}, "240s", "5s").Should(Succeed())
+			_, _ = e2e.ParseNodes(kubeConfigFile, true)
+		})
+
+		It("Rejoins other servers to cluster", func() {
+			// We must remove the db directory on the other servers before restarting k3s
+			// otherwise the nodes may join the old cluster
+			for _, nodeName := range serverNodeNames[1:] {
+				cmd := "sudo rm -rf /var/lib/rancher/k3s/server/db"
+				Expect(e2e.RunCmdOnNode(cmd, nodeName)).Error().NotTo(HaveOccurred())
+			}
+
+			for _, nodeName := range serverNodeNames[1:] {
+				cmd := "sudo systemctl start k3s"
+				Expect(e2e.RunCmdOnNode(cmd, nodeName)).Error().NotTo(HaveOccurred())
 			}
+		})
+
+		It("Checks that all nodes and pods are ready", func() {
 			//Verifies node is up and pods running
 			fmt.Printf("\nFetching node status\n")
 			Eventually(func(g Gomega) {
@@ -167,8 +280,9 @@ var _ = Describe("Verify Create", Ordered, func() {
 				}
 			}, "620s", "5s").Should(Succeed())
 			_, _ = e2e.ParsePods(kubeConfigFile, true)
-			//Verifies test workload1 is present
-			//Verifies test workload2 is not present
+		})
+
+		It("Verifies that workload1 exists and workload2 does not", func() {
 			cmd := "kubectl get pods --kubeconfig=" + kubeConfigFile
 			res, err := e2e.RunCommand(cmd)
 			Expect(err).NotTo(HaveOccurred())

tests/e2e/splitserver/splitserver_test.go

@@ -13,7 +13,7 @@ import (
 	. "github.com/onsi/gomega"
 )
 
-// Valid nodeOS: generic/ubuntu2004, opensuse/Leap-15.3.x86_64, dweomer/microos.amd64
+// Valid nodeOS: generic/ubuntu2004, opensuse/Leap-15.3.x86_64
 var nodeOS = flag.String("nodeOS", "generic/ubuntu2004", "VM operating system")
 var etcdCount = flag.Int("etcdCount", 1, "number of server nodes only deploying etcd")
 var controlPlaneCount = flag.Int("controlPlaneCount", 1, "number of server nodes acting as control plane")

tests/e2e/testutils.go

@@ -123,7 +123,11 @@ func CreateCluster(nodeOS string, serverCount, agentCount int) ([]string, []stri
 			return nil
 		})
 		// We must wait a bit between provisioning nodes to avoid too many learners attempting to join the cluster
-		time.Sleep(20 * time.Second)
+		if strings.Contains(node, "agent") {
+			time.Sleep(5 * time.Second)
+		} else {
+			time.Sleep(30 * time.Second)
+		}
 	}
 	if err := errg.Wait(); err != nil {
 		return nil, nil, err

tests/e2e/upgradecluster/upgradecluster_test.go

@@ -14,7 +14,7 @@ import (
 
 // Valid nodeOS:
 // generic/ubuntu2004, generic/centos7, generic/rocky8
-// opensuse/Leap-15.3.x86_64, dweomer/microos.amd64
+// opensuse/Leap-15.3.x86_64
 var nodeOS = flag.String("nodeOS", "generic/ubuntu2004", "VM operating system")
 var serverCount = flag.Int("serverCount", 3, "number of server nodes")
 var agentCount = flag.Int("agentCount", 2, "number of agent nodes")

tests/e2e/validatecluster/validatecluster_test.go

@@ -14,7 +14,7 @@ import (
 
 // Valid nodeOS:
 // generic/ubuntu2004, generic/centos7, generic/rocky8,
-// opensuse/Leap-15.3.x86_64, dweomer/microos.amd64
+// opensuse/Leap-15.3.x86_64
 var nodeOS = flag.String("nodeOS", "generic/ubuntu2004", "VM operating system")
 var serverCount = flag.Int("serverCount", 3, "number of server nodes")
 var agentCount = flag.Int("agentCount", 2, "number of agent nodes")