mirror of
https://github.com/k3s-io/k3s.git
synced 2024-06-07 19:41:36 +00:00
86e36225f5
* Consolidate cluster reset and snapshot E2E tests * Add more context to secrets-encryption test * Reuse build workflow * Convert updatecli to job level permissions * Remove dweomer microos from E2E and install testing Signed-off-by: Derek Nola <derek.nola@suse.com>
308 lines
11 KiB
Go
308 lines
11 KiB
Go
package snapshotrestore
|
|
|
|
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
|
|
|
|
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_E2ESnapshotRestore(t *testing.T) {
|
|
RegisterFailHandler(Fail)
|
|
flag.Parse()
|
|
suiteConfig, reporterConfig := GinkgoConfiguration()
|
|
RunSpecs(t, "SnapshotRestore Test Suite", suiteConfig, reporterConfig)
|
|
}
|
|
|
|
var (
|
|
kubeConfigFile string
|
|
serverNodeNames []string
|
|
agentNodeNames []string
|
|
snapshotname string
|
|
)
|
|
|
|
var _ = ReportAfterEach(e2e.GenReport)
|
|
|
|
var _ = Describe("Verify snapshots and cluster restores work", 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 test workload before snapshot is created", func() {
|
|
res, err := e2e.DeployWorkload("clusterip.yaml", kubeConfigFile, *hardened)
|
|
Expect(err).NotTo(HaveOccurred(), "Cluster IP manifest not deployed: "+res)
|
|
|
|
Eventually(func(g Gomega) {
|
|
cmd := "kubectl get pods -o=name -l k8s-app=nginx-app-clusterip --field-selector=status.phase=Running --kubeconfig=" + kubeConfigFile
|
|
res, err := e2e.RunCommand(cmd)
|
|
g.Expect(err).NotTo(HaveOccurred())
|
|
g.Expect(res).Should((ContainSubstring("test-clusterip")), "failed cmd: "+cmd+" result: "+res)
|
|
}, "240s", "5s").Should(Succeed())
|
|
})
|
|
|
|
It("Verifies Snapshot is created", func() {
|
|
Eventually(func(g Gomega) {
|
|
cmd := "sudo k3s etcd-snapshot save"
|
|
_, err := e2e.RunCmdOnNode(cmd, "server-0")
|
|
g.Expect(err).NotTo(HaveOccurred())
|
|
cmd = "sudo ls /var/lib/rancher/k3s/server/db/snapshots/"
|
|
snapshotname, err = e2e.RunCmdOnNode(cmd, "server-0")
|
|
g.Expect(err).NotTo(HaveOccurred())
|
|
fmt.Println("Snapshot Name", snapshotname)
|
|
g.Expect(snapshotname).Should(ContainSubstring("on-demand-server-0"))
|
|
}, "420s", "10s").Should(Succeed())
|
|
})
|
|
|
|
It("Verifies another test workload after snapshot is created", func() {
|
|
_, err := e2e.DeployWorkload("nodeport.yaml", kubeConfigFile, *hardened)
|
|
Expect(err).NotTo(HaveOccurred(), "NodePort manifest not deployed")
|
|
Eventually(func(g Gomega) {
|
|
cmd := "kubectl get pods -o=name -l k8s-app=nginx-app-nodeport --field-selector=status.phase=Running --kubeconfig=" + kubeConfigFile
|
|
res, err := e2e.RunCommand(cmd)
|
|
g.Expect(err).NotTo(HaveOccurred())
|
|
g.Expect(res).Should(ContainSubstring("test-nodeport"), "nodeport pod was not created")
|
|
}, "240s", "5s").Should(Succeed())
|
|
})
|
|
|
|
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"
|
|
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
|
|
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) {
|
|
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)
|
|
}
|
|
}
|
|
}, "620s", "5s").Should(Succeed())
|
|
_, _ = e2e.ParsePods(kubeConfigFile, true)
|
|
})
|
|
|
|
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())
|
|
Expect(res).Should(ContainSubstring("test-clusterip"))
|
|
Expect(res).ShouldNot(ContainSubstring("test-nodeport"))
|
|
})
|
|
})
|
|
})
|
|
|
|
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())
|
|
}
|
|
})
|