mirror of
https://github.com/edgelesssys/constellation.git
synced 2024-10-01 01:36:09 -04:00
edf424d415
Signed-off-by: Daniel Weiße <dw@edgeless.systems>
315 lines
12 KiB
Go
315 lines
12 KiB
Go
package k8sapi
|
|
|
|
import (
|
|
"context"
|
|
"errors"
|
|
"fmt"
|
|
"net/http"
|
|
"os"
|
|
"os/exec"
|
|
"regexp"
|
|
"strings"
|
|
"time"
|
|
|
|
"github.com/edgelesssys/constellation/bootstrapper/internal/kubernetes/k8sapi/resources"
|
|
"github.com/edgelesssys/constellation/internal/logger"
|
|
"go.uber.org/zap"
|
|
)
|
|
|
|
const (
|
|
// kubeConfig is the path to the Kubernetes admin config (used for authentication).
|
|
kubeConfig = "/etc/kubernetes/admin.conf"
|
|
// kubeletStartTimeout is the maximum time given to the kubelet service to (re)start.
|
|
kubeletStartTimeout = 10 * time.Minute
|
|
)
|
|
|
|
var providerIDRegex = regexp.MustCompile(`^azure:///subscriptions/([^/]+)/resourceGroups/([^/]+)/providers/Microsoft.Compute/virtualMachineScaleSets/([^/]+)/virtualMachines/([^/]+)$`)
|
|
|
|
// Client provides the functionality of `kubectl apply`.
|
|
type Client interface {
|
|
Apply(resources resources.Marshaler, forceConflicts bool) error
|
|
SetKubeconfig(kubeconfig []byte)
|
|
// TODO: add tolerations
|
|
}
|
|
|
|
// KubernetesUtil provides low level management of the kubernetes cluster.
|
|
type KubernetesUtil struct {
|
|
inst installer
|
|
}
|
|
|
|
// NewKubernetesUtil creates a new KubernetesUtil.
|
|
func NewKubernetesUtil() *KubernetesUtil {
|
|
return &KubernetesUtil{
|
|
inst: newOSInstaller(),
|
|
}
|
|
}
|
|
|
|
// InstallComponents installs kubernetes components in the version specified.
|
|
func (k *KubernetesUtil) InstallComponents(ctx context.Context, version string) error {
|
|
var versionConf kubernetesVersion
|
|
var ok bool
|
|
if versionConf, ok = versionConfigs[version]; !ok {
|
|
return fmt.Errorf("unsupported kubernetes version %q", version)
|
|
}
|
|
if err := versionConf.installK8sComponents(ctx, k.inst); err != nil {
|
|
return err
|
|
}
|
|
|
|
return enableSystemdUnit(ctx, kubeletServiceEtcPath)
|
|
}
|
|
|
|
func (k *KubernetesUtil) InitCluster(ctx context.Context, initConfig []byte, log *logger.Logger) error {
|
|
// TODO: audit policy should be user input
|
|
auditPolicy, err := resources.NewDefaultAuditPolicy().Marshal()
|
|
if err != nil {
|
|
return fmt.Errorf("generating default audit policy: %w", err)
|
|
}
|
|
if err := os.WriteFile(auditPolicyPath, auditPolicy, 0o644); err != nil {
|
|
return fmt.Errorf("writing default audit policy: %w", err)
|
|
}
|
|
|
|
initConfigFile, err := os.CreateTemp("", "kubeadm-init.*.yaml")
|
|
if err != nil {
|
|
return fmt.Errorf("creating init config file %v: %w", initConfigFile.Name(), err)
|
|
}
|
|
|
|
if _, err := initConfigFile.Write(initConfig); err != nil {
|
|
return fmt.Errorf("writing kubeadm init yaml config %v: %w", initConfigFile.Name(), err)
|
|
}
|
|
|
|
cmd := exec.CommandContext(ctx, kubeadmPath, "init", "-v=5", "--config", initConfigFile.Name())
|
|
out, err := cmd.CombinedOutput()
|
|
if err != nil {
|
|
var exitErr *exec.ExitError
|
|
if errors.As(err, &exitErr) {
|
|
return fmt.Errorf("kubeadm init failed (code %v) with: %s", exitErr.ExitCode(), out)
|
|
}
|
|
return fmt.Errorf("kubeadm init: %w", err)
|
|
}
|
|
log.With(zap.String("output", string(out))).Infof("kubeadm init succeeded")
|
|
return nil
|
|
}
|
|
|
|
type SetupPodNetworkInput struct {
|
|
CloudProvider string
|
|
NodeName string
|
|
FirstNodePodCIDR string
|
|
SubnetworkPodCIDR string
|
|
ProviderID string
|
|
}
|
|
|
|
// SetupPodNetwork sets up the cilium pod network.
|
|
func (k *KubernetesUtil) SetupPodNetwork(ctx context.Context, in SetupPodNetworkInput) error {
|
|
switch in.CloudProvider {
|
|
case "gcp":
|
|
return k.setupGCPPodNetwork(ctx, in.NodeName, in.FirstNodePodCIDR, in.SubnetworkPodCIDR)
|
|
case "azure":
|
|
return k.setupAzurePodNetwork(ctx, in.ProviderID, in.SubnetworkPodCIDR)
|
|
case "qemu":
|
|
return k.setupQemuPodNetwork(ctx, in.SubnetworkPodCIDR)
|
|
default:
|
|
return fmt.Errorf("unsupported cloud provider %q", in.CloudProvider)
|
|
}
|
|
}
|
|
|
|
func (k *KubernetesUtil) setupAzurePodNetwork(ctx context.Context, providerID, subnetworkPodCIDR string) error {
|
|
matches := providerIDRegex.FindStringSubmatch(providerID)
|
|
if len(matches) != 5 {
|
|
return fmt.Errorf("error splitting providerID %q", providerID)
|
|
}
|
|
|
|
ciliumInstall := exec.CommandContext(ctx, "cilium", "install", "--azure-resource-group", matches[2], "--ipam", "azure",
|
|
"--helm-set",
|
|
"tunnel=disabled,enableIPv4Masquerade=true,azure.enabled=true,debug.enabled=true,ipv4NativeRoutingCIDR="+subnetworkPodCIDR+
|
|
",endpointRoutes.enabled=true,encryption.enabled=true,encryption.type=wireguard,l7Proxy=false,egressMasqueradeInterfaces=eth0")
|
|
ciliumInstall.Env = append(os.Environ(), "KUBECONFIG="+kubeConfig)
|
|
out, err := ciliumInstall.CombinedOutput()
|
|
if err != nil {
|
|
err = errors.New(string(out))
|
|
return err
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func (k *KubernetesUtil) setupGCPPodNetwork(ctx context.Context, nodeName, nodePodCIDR, subnetworkPodCIDR string) error {
|
|
out, err := exec.CommandContext(ctx, kubectlPath, "--kubeconfig", kubeConfig, "patch", "node", nodeName, "-p", "{\"spec\":{\"podCIDR\": \""+nodePodCIDR+"\"}}").CombinedOutput()
|
|
if err != nil {
|
|
err = errors.New(string(out))
|
|
return err
|
|
}
|
|
|
|
// allow coredns to run on uninitialized nodes (required by cloud-controller-manager)
|
|
err = exec.CommandContext(ctx, kubectlPath, "--kubeconfig", kubeConfig, "-n", "kube-system", "patch", "deployment", "coredns", "--type", "json", "-p", "[{\"op\":\"add\",\"path\":\"/spec/template/spec/tolerations/-\",\"value\":{\"key\":\"node.cloudprovider.kubernetes.io/uninitialized\",\"value\":\"true\",\"effect\":\"NoSchedule\"}},{\"op\":\"add\",\"path\":\"/spec/template/spec/nodeSelector\",\"value\":{\"node-role.kubernetes.io/control-plane\":\"\"}}]").Run()
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
ciliumInstall := exec.CommandContext(ctx, "cilium", "install", "--ipam", "kubernetes", "--ipv4-native-routing-cidr", subnetworkPodCIDR,
|
|
"--helm-set", "endpointRoutes.enabled=true,tunnel=disabled,encryption.enabled=true,encryption.type=wireguard,l7Proxy=false")
|
|
ciliumInstall.Env = append(os.Environ(), "KUBECONFIG="+kubeConfig)
|
|
out, err = ciliumInstall.CombinedOutput()
|
|
if err != nil {
|
|
err = errors.New(string(out))
|
|
return err
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
// FixCilium fixes https://github.com/cilium/cilium/issues/19958 but instead of a rollout restart of
|
|
// the cilium daemonset, it only restarts the local cilium pod.
|
|
func (k *KubernetesUtil) FixCilium(nodeNameK8s string) {
|
|
// wait for cilium pod to be healthy
|
|
for {
|
|
time.Sleep(5 * time.Second)
|
|
resp, err := http.Get("http://127.0.0.1:9876/healthz")
|
|
if err != nil {
|
|
fmt.Printf("waiting for local cilium daemonset pod not healthy: %v\n", err)
|
|
continue
|
|
}
|
|
resp.Body.Close()
|
|
if resp.StatusCode == 200 {
|
|
break
|
|
}
|
|
}
|
|
|
|
// get cilium pod name
|
|
out, err := exec.CommandContext(context.Background(), "/bin/bash", "-c", "/run/state/bin/crictl ps -o json | jq -r '.containers[] | select(.metadata.name == \"cilium-agent\") | .podSandboxId'").CombinedOutput()
|
|
if err != nil {
|
|
fmt.Printf("getting pod id failed: %v: %v\n", err, string(out))
|
|
return
|
|
}
|
|
outLines := strings.Split(string(out), "\n")
|
|
fmt.Println(outLines)
|
|
podID := outLines[len(outLines)-2]
|
|
|
|
// stop and delete pod
|
|
out, err = exec.CommandContext(context.Background(), "/run/state/bin/crictl", "stopp", podID).CombinedOutput()
|
|
if err != nil {
|
|
fmt.Printf("stopping cilium agent pod failed: %v: %v\n", err, string(out))
|
|
return
|
|
}
|
|
out, err = exec.CommandContext(context.Background(), "/run/state/bin/crictl", "rmp", podID).CombinedOutput()
|
|
if err != nil {
|
|
fmt.Printf("removing cilium agent pod failed: %v: %v\n", err, string(out))
|
|
}
|
|
}
|
|
|
|
func (k *KubernetesUtil) setupQemuPodNetwork(ctx context.Context, subnetworkPodCIDR string) error {
|
|
ciliumInstall := exec.CommandContext(ctx, "cilium", "install", "--encryption", "wireguard", "--helm-set", "ipam.operator.clusterPoolIPv4PodCIDRList="+subnetworkPodCIDR+",endpointRoutes.enabled=true")
|
|
ciliumInstall.Env = append(os.Environ(), "KUBECONFIG="+kubeConfig)
|
|
out, err := ciliumInstall.CombinedOutput()
|
|
if err != nil {
|
|
err = errors.New(string(out))
|
|
return err
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// SetupAutoscaling deploys the k8s cluster autoscaler.
|
|
func (k *KubernetesUtil) SetupAutoscaling(kubectl Client, clusterAutoscalerConfiguration resources.Marshaler, secrets resources.Marshaler) error {
|
|
if err := kubectl.Apply(secrets, true); err != nil {
|
|
return fmt.Errorf("applying cluster-autoscaler Secrets: %w", err)
|
|
}
|
|
return kubectl.Apply(clusterAutoscalerConfiguration, true)
|
|
}
|
|
|
|
// SetupJoinService deploys the Constellation node join service.
|
|
func (k *KubernetesUtil) SetupJoinService(kubectl Client, joinServiceConfiguration resources.Marshaler) error {
|
|
return kubectl.Apply(joinServiceConfiguration, true)
|
|
}
|
|
|
|
// SetupGCPGuestAgent deploys the GCP guest agent daemon set.
|
|
func (k *KubernetesUtil) SetupGCPGuestAgent(kubectl Client, guestAgentDaemonset resources.Marshaler) error {
|
|
return kubectl.Apply(guestAgentDaemonset, true)
|
|
}
|
|
|
|
// SetupCloudControllerManager deploys the k8s cloud-controller-manager.
|
|
func (k *KubernetesUtil) SetupCloudControllerManager(kubectl Client, cloudControllerManagerConfiguration resources.Marshaler, configMaps resources.Marshaler, secrets resources.Marshaler) error {
|
|
if err := kubectl.Apply(configMaps, true); err != nil {
|
|
return fmt.Errorf("applying ccm ConfigMaps: %w", err)
|
|
}
|
|
if err := kubectl.Apply(secrets, true); err != nil {
|
|
return fmt.Errorf("applying ccm Secrets: %w", err)
|
|
}
|
|
if err := kubectl.Apply(cloudControllerManagerConfiguration, true); err != nil {
|
|
return fmt.Errorf("applying ccm: %w", err)
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// SetupCloudNodeManager deploys the k8s cloud-node-manager.
|
|
func (k *KubernetesUtil) SetupCloudNodeManager(kubectl Client, cloudNodeManagerConfiguration resources.Marshaler) error {
|
|
return kubectl.Apply(cloudNodeManagerConfiguration, true)
|
|
}
|
|
|
|
// SetupAccessManager deploys the constellation-access-manager for deploying SSH keys on control-plane & worker nodes.
|
|
func (k *KubernetesUtil) SetupAccessManager(kubectl Client, accessManagerConfiguration resources.Marshaler) error {
|
|
return kubectl.Apply(accessManagerConfiguration, true)
|
|
}
|
|
|
|
// SetupKMS deploys the KMS deployment.
|
|
func (k *KubernetesUtil) SetupKMS(kubectl Client, kmsConfiguration resources.Marshaler) error {
|
|
if err := kubectl.Apply(kmsConfiguration, true); err != nil {
|
|
return fmt.Errorf("applying KMS configuration: %w", err)
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// SetupVerificationService deploys the verification service.
|
|
func (k *KubernetesUtil) SetupVerificationService(kubectl Client, verificationServiceConfiguration resources.Marshaler) error {
|
|
return kubectl.Apply(verificationServiceConfiguration, true)
|
|
}
|
|
|
|
// JoinCluster joins existing Kubernetes cluster using kubeadm join.
|
|
func (k *KubernetesUtil) JoinCluster(ctx context.Context, joinConfig []byte, log *logger.Logger) error {
|
|
// TODO: audit policy should be user input
|
|
auditPolicy, err := resources.NewDefaultAuditPolicy().Marshal()
|
|
if err != nil {
|
|
return fmt.Errorf("generating default audit policy: %w", err)
|
|
}
|
|
if err := os.WriteFile(auditPolicyPath, auditPolicy, 0o644); err != nil {
|
|
return fmt.Errorf("writing default audit policy: %w", err)
|
|
}
|
|
|
|
joinConfigFile, err := os.CreateTemp("", "kubeadm-join.*.yaml")
|
|
if err != nil {
|
|
return fmt.Errorf("creating join config file %v: %w", joinConfigFile.Name(), err)
|
|
}
|
|
|
|
if _, err := joinConfigFile.Write(joinConfig); err != nil {
|
|
return fmt.Errorf("writing kubeadm init yaml config %v: %w", joinConfigFile.Name(), err)
|
|
}
|
|
|
|
// run `kubeadm join` to join a worker node to an existing Kubernetes cluster
|
|
cmd := exec.CommandContext(ctx, kubeadmPath, "join", "-v=5", "--config", joinConfigFile.Name())
|
|
out, err := cmd.CombinedOutput()
|
|
if err != nil {
|
|
var exitErr *exec.ExitError
|
|
if errors.As(err, &exitErr) {
|
|
return fmt.Errorf("kubeadm join failed (code %v) with: %s (full err: %s)", exitErr.ExitCode(), out, err)
|
|
}
|
|
return fmt.Errorf("kubeadm join: %w", err)
|
|
}
|
|
log.With(zap.String("output", string(out))).Infof("kubeadm join succeeded")
|
|
|
|
return nil
|
|
}
|
|
|
|
// StartKubelet enables and starts the kubelet systemd unit.
|
|
func (k *KubernetesUtil) StartKubelet() error {
|
|
ctx, cancel := context.WithTimeout(context.TODO(), kubeletStartTimeout)
|
|
defer cancel()
|
|
if err := enableSystemdUnit(ctx, kubeletServiceEtcPath); err != nil {
|
|
return fmt.Errorf("enabling kubelet systemd unit: %w", err)
|
|
}
|
|
return startSystemdUnit(ctx, "kubelet.service")
|
|
}
|
|
|
|
// RestartKubelet restarts a kubelet.
|
|
func (k *KubernetesUtil) RestartKubelet() error {
|
|
ctx, cancel := context.WithTimeout(context.TODO(), kubeletStartTimeout)
|
|
defer cancel()
|
|
return restartSystemdUnit(ctx, "kubelet.service")
|
|
}
|