refactor: replace kubectl operation with k8s apiclient call

pkg/runtime/kubernetes/init.go

@@ -32,19 +32,16 @@ import (
 )
 
 const (
-	RemoteCmdCopyStatic            = "mkdir -p %s && cp -f %s %s"
-	RemoteApplyYaml                = `echo '%s' | kubectl apply -f -`
-	RemoteCmdGetNetworkInterface   = "ls /sys/class/net"
-	RemoteCmdExistNetworkInterface = "ip addr show %s | egrep \"%s\" || true"
-	WriteKubeadmConfigCmd          = `cd %s && echo '%s' > etc/kubeadm.yml`
-	DefaultVIP                     = "10.103.97.2"
-	DefaultAPIserverDomain         = "apiserver.cluster.local"
-	DefaultRegistryPort            = 5000
-	DockerCertDir                  = "/etc/docker/certs.d"
+	RemoteCmdCopyStatic    = "mkdir -p %s && cp -f %s %s"
+	WriteKubeadmConfigCmd  = `cd %s && echo '%s' > etc/kubeadm.yml`
+	DefaultVIP             = "10.103.97.2"
+	DefaultAPIserverDomain = "apiserver.cluster.local"
+	DefaultRegistryPort    = 5000
+	DockerCertDir          = "/etc/docker/certs.d"
 )
 
 func (k *Runtime) ConfigKubeadmOnMaster0() error {
-	if err := k.LoadFromClusterfile(k.Config.ClusterFileKubeConfig); err != nil {
+	if err := k.LoadFromClusterfile(k.KubeadmConfigFromClusterfile); err != nil {
 		return fmt.Errorf("failed to load kubeadm config from clusterfile: %v", err)
 	}
 	// TODO handle the kubeadm config, like kubeproxy config

pkg/runtime/kubernetes/join_masters.go

@@ -23,7 +23,6 @@ import (
 	"strings"
 	"sync"
 
-	"github.com/pkg/errors"
 	"github.com/sealerio/sealer/common"
 	"github.com/sealerio/sealer/pkg/clustercert"
 	"github.com/sealerio/sealer/pkg/ipvs"
@@ -32,6 +31,8 @@ import (
 	"github.com/sealerio/sealer/utils/ssh"
 	versionUtils "github.com/sealerio/sealer/utils/version"
 	"github.com/sealerio/sealer/utils/yaml"
+
+	"github.com/pkg/errors"
 	"github.com/sirupsen/logrus"
 	"golang.org/x/sync/errgroup"
 )
@@ -71,7 +72,6 @@ rm -rf /var/lib/etcd && rm -rf /var/etcd
 	RemoteRemoveRegistryCerts    = "rm -rf " + DockerCertDir + "/%s*"
 	RemoveLvscareStaticPod       = "rm -rf  /etc/kubernetes/manifests/kube-sealyun-lvscare*"
 	CreateLvscareStaticPod       = "mkdir -p /etc/kubernetes/manifests && echo '%s' > /etc/kubernetes/manifests/kube-sealyun-lvscare.yaml"
-	KubeDeleteNode               = "kubectl delete node %s"
 	// TODO check kubernetes certs
 	RemoteCheckCerts = "kubeadm alpha certs check-expiration"
 )
@@ -389,30 +389,32 @@ func (k *Runtime) deleteMasters(masters []net.IP) error {
 	return eg.Wait()
 }
 
-func (k *Runtime) isHostName(master, host net.IP) (string, error) {
-	hostString, err := k.CmdToString(master, "kubectl get nodes | grep -v NAME  | awk '{print $1}'", ",")
+// isHostInExistingCluster checks if the input host is contained in the existing cluster.
+// It gets all nodes within existing cluster,
+//  judges each node name with the hostname of input host node
+//  and finially get the comparison result.
+func (k *Runtime) isHostInExistingCluster(host net.IP) (string, error) {
+	nodes, err := k.ListNodes()
 	if err != nil {
 		return "", err
 	}
+	if len(nodes.Items) == 0 {
+		return "", fmt.Errorf("no node gotten by k8s client")
+	}
+
 	hostName, err := k.CmdToString(host, "hostname", "")
 	if err != nil {
 		return "", err
 	}
-	hosts := strings.Split(hostString, ",")
-	var name string
-	for _, h := range hosts {
-		if strings.TrimSpace(h) == "" {
-			continue
-		} else {
-			hh := strings.ToLower(h)
-			fromH := strings.ToLower(hostName)
-			if hh == fromH {
-				name = h
-				break
-			}
+
+	for _, nodeItem := range nodes.Items {
+		hh := strings.ToLower(nodeItem.Name)
+		fromH := strings.ToLower(hostName)
+		if hh == fromH {
+			return hh, nil
 		}
 	}
-	return name, nil
+	return "", fmt.Errorf("failed to get node: no node's hostname matches hostname of host(%s)", host)
 }
 
 func (k *Runtime) deleteMaster(master net.IP) error {
@@ -449,19 +451,15 @@ func (k *Runtime) deleteMaster(master net.IP) error {
 	}
 
 	if len(masterIPs) > 0 {
-		hostname, err := k.isHostName(k.cluster.GetMaster0IP(), master)
+		hostname, err := k.isHostInExistingCluster(master)
 		if err != nil {
 			return err
 		}
-		master0SSH, err := k.getHostSSHClient(k.cluster.GetMaster0IP())
-		if err != nil {
-			return fmt.Errorf("failed to get master0 ssh client: %v", err)
-		}
-
-		if err := master0SSH.CmdAsync(k.cluster.GetMaster0IP(), fmt.Sprintf(KubeDeleteNode, strings.TrimSpace(hostname))); err != nil {
-			return fmt.Errorf("failed to delete node %s: %v", hostname, err)
+		if err := k.DeleteNode(strings.TrimSpace(hostname)); err != nil {
+			return err
 		}
 	}
+
 	lvsImage := k.RegConfig.Repo() + "/fanux/lvscare:latest"
 	yaml := ipvs.LvsStaticPodYaml(k.getVIP(), masterIPs, lvsImage)
 	eg, _ := errgroup.WithContext(context.Background())

pkg/runtime/kubernetes/join_nodes.go

@@ -166,21 +166,17 @@ func (k *Runtime) deleteNode(node net.IP) error {
 	if err := ssh.CmdAsync(node, remoteCleanCmds...); err != nil {
 		return err
 	}
-	//remove node
+
+	// remove node
 	if len(k.cluster.GetMasterIPList()) > 0 {
-		hostname, err := k.isHostName(k.cluster.GetMaster0IP(), node)
+		hostname, err := k.isHostInExistingCluster(node)
 		if err != nil {
 			return err
 		}
-		ssh, err := k.getHostSSHClient(k.cluster.GetMaster0IP())
-		if err != nil {
-			return fmt.Errorf("failed to get master0 ssh client(%s): %v", k.cluster.GetMaster0IP(), err)
-		}
-		if err := ssh.CmdAsync(k.cluster.GetMaster0IP(), fmt.Sprintf(KubeDeleteNode, strings.TrimSpace(hostname))); err != nil {
-			return fmt.Errorf("failed to delete node %s: %v", hostname, err)
+		if err := k.DeleteNode(strings.TrimSpace(hostname)); err != nil {
+			return err
 		}
 	}
-
 	return nil
 }
 

pkg/runtime/kubernetes/runtime.go

@@ -17,45 +17,53 @@ package kubernetes
 import (
 	"context"
 	"fmt"
-	"strings"
-	"sync"
-
-	"github.com/sealerio/sealer/pkg/registry"
-	"github.com/sealerio/sealer/pkg/runtime/kubernetes/kubeadm"
-	"github.com/sealerio/sealer/utils"
-	versionUtils "github.com/sealerio/sealer/utils/version"
-
 	"net"
 	"path/filepath"
+	"strings"
+	"sync"
 	"time"
 
 	"github.com/sealerio/sealer/common"
+	"github.com/sealerio/sealer/pkg/client/k8s"
+	"github.com/sealerio/sealer/pkg/registry"
 	"github.com/sealerio/sealer/pkg/runtime"
+	"github.com/sealerio/sealer/pkg/runtime/kubernetes/kubeadm"
 	v2 "github.com/sealerio/sealer/types/api/v2"
+	"github.com/sealerio/sealer/utils"
 	"github.com/sealerio/sealer/utils/platform"
 	"github.com/sealerio/sealer/utils/ssh"
 	strUtils "github.com/sealerio/sealer/utils/strings"
-	"k8s.io/kubernetes/cmd/kubeadm/app/apis/kubeadm/v1beta2"
+	versionUtils "github.com/sealerio/sealer/utils/version"
 
 	"github.com/sirupsen/logrus"
 	"golang.org/x/sync/errgroup"
+	"k8s.io/kubernetes/cmd/kubeadm/app/apis/kubeadm/v1beta2"
 )
 
 type Config struct {
-	Vlog      int
-	VIP       string
-	RegConfig *registry.Config
 	// Clusterfile: the absolute path, we need to read kubeadm config from Clusterfile
 	ClusterFileKubeConfig *kubeadm.KubeadmConfig
-	APIServerDomain       string
 }
 
 // Runtime struct is the runtime interface for kubernetes
 type Runtime struct {
 	*sync.Mutex
+	// TODO: remove field cluster from runtime pkg
+	// just make runtime to use essential data from cluster, rather than the whole cluster scope.
 	cluster *v2.Cluster
+
+	// The KubeadmConfig used to setup the final cluster.
+	// Its data is from KubeadmConfig input from Clusterfile and KubeadmConfig in ClusterImage.
 	*kubeadm.KubeadmConfig
-	*Config
+	// *Config
+	KubeadmConfigFromClusterfile *kubeadm.KubeadmConfig
+	APIServerDomain              string
+	Vlog                         int
+	VIP                          string
+
+	// RegConfig contains the embedded registry configuration of cluster
+	RegConfig *registry.Config
+	*k8s.Client
 }
 
 // NewDefaultRuntime arg "clusterfileKubeConfig" is the Clusterfile path/name, runtime need read kubeadm config from it
@@ -67,13 +75,24 @@ func NewDefaultRuntime(cluster *v2.Cluster, clusterfileKubeConfig *kubeadm.Kubea
 func newKubernetesRuntime(cluster *v2.Cluster, clusterFileKubeConfig *kubeadm.KubeadmConfig) (runtime.Interface, error) {
 	k := &Runtime{
 		cluster: cluster,
-		Config: &Config{
+		/*Config: &Config{
 			ClusterFileKubeConfig: clusterFileKubeConfig,
-			APIServerDomain:       DefaultAPIserverDomain,
-		},
-		KubeadmConfig: &kubeadm.KubeadmConfig{},
+		},*/
+		KubeadmConfigFromClusterfile: clusterFileKubeConfig,
+		KubeadmConfig:                &kubeadm.KubeadmConfig{},
+		APIServerDomain:              DefaultAPIserverDomain,
 	}
-	k.Config.RegConfig = registry.GetConfig(k.getImageMountDir(), k.cluster.GetMaster0IP())
+
+	var err error
+	if k.Client, err = k8s.Newk8sClient(); err != nil {
+		// In current design, as runtime controls all cluster operations including run, join, delete
+		// and so on, then when executing run operation, it will definitely fail when creating k8s client
+		// since no k8s cluster is setup. While when join and delete operation, the cluster already exists,
+		// we can make it to create k8s client. Therefore just throw a warn log to move on.
+		logrus.Warnf("failed to create k8s client: %v", err)
+	}
+
+	k.RegConfig = registry.GetConfig(k.getImageMountDir(), k.cluster.GetMaster0IP())
 	k.setCertSANS(append(
 		[]string{"127.0.0.1", k.getAPIServerDomain(), k.getVIP().String()},
 		k.cluster.GetMasterIPStrList()...),
@@ -239,7 +258,7 @@ func (k *Runtime) getDNSDomain() string {
 }
 
 func (k *Runtime) getAPIServerDomain() string {
-	return k.Config.APIServerDomain
+	return k.APIServerDomain
 }
 
 func (k *Runtime) getKubeVersion() string {
@@ -369,8 +388,8 @@ func (k *Runtime) MergeKubeadmConfig() error {
 	if k.getKubeVersion() != "" {
 		return nil
 	}
-	if k.Config.ClusterFileKubeConfig != nil {
-		if err := k.LoadFromClusterfile(k.Config.ClusterFileKubeConfig); err != nil {
+	if k.KubeadmConfigFromClusterfile != nil {
+		if err := k.LoadFromClusterfile(k.KubeadmConfigFromClusterfile); err != nil {
 			return fmt.Errorf("failed to load kubeadm config from clusterfile: %v", err)
 		}
 	}