Backport bootstrap release 1.21 (#4313)

go.mod

@@ -116,8 +116,8 @@ require (
 	go.etcd.io/etcd v0.5.0-alpha.5.0.20201208200253-50621aee4aea
 	golang.org/x/crypto v0.0.0-20210322153248-0c34fe9e7dc2
 	golang.org/x/net v0.0.0-20210226172049-e18ecbb05110
-	golang.org/x/sys v0.0.0-20210426230700-d19ff857e887
-	google.golang.org/grpc v1.37.0
+	golang.org/x/sys v0.0.0-20210603081109-ebe580a85c40
+	google.golang.org/grpc v1.38.0
 	gopkg.in/yaml.v2 v2.4.0
 	k8s.io/api v0.21.5
 	k8s.io/apimachinery v0.21.5

pkg/bootstrap/bootstrap.go

@@ -7,6 +7,7 @@ import (
 	"net/http"
 	"os"
 	"path/filepath"
+	"time"
 
 	"github.com/pkg/errors"
 	"github.com/rancher/k3s/pkg/daemons/config"
@@ -15,17 +16,19 @@ import (
 func Handler(bootstrap *config.ControlRuntimeBootstrap) http.Handler {
 	return http.HandlerFunc(func(rw http.ResponseWriter, req *http.Request) {
 		rw.Header().Set("Content-Type", "application/json")
-		Write(rw, bootstrap)
+		ReadFromDisk(rw, bootstrap)
 	})
 }
 
-func Write(w io.Writer, bootstrap *config.ControlRuntimeBootstrap) error {
-	paths, err := objToMap(bootstrap)
+// ReadFromDisk reads the bootstrap data from the files on disk and
+// writes their content in JSON form to the given io.Writer.
+func ReadFromDisk(w io.Writer, bootstrap *config.ControlRuntimeBootstrap) error {
+	paths, err := ObjToMap(bootstrap)
 	if err != nil {
 		return nil
 	}
 
-	dataMap := map[string][]byte{}
+	dataMap := make(map[string]File)
 	for pathKey, path := range paths {
 		if path == "" {
 			continue
@@ -35,24 +38,45 @@ func Write(w io.Writer, bootstrap *config.ControlRuntimeBootstrap) error {
 			return errors.Wrapf(err, "failed to read %s", path)
 		}
 
-		dataMap[pathKey] = data
+		info, err := os.Stat(path)
+		if err != nil {
+			return err
+		}
+
+		dataMap[pathKey] = File{
+			Timestamp: info.ModTime(),
+			Content:   data,
+		}
 	}
 
 	return json.NewEncoder(w).Encode(dataMap)
 }
 
-func Read(r io.Reader, bootstrap *config.ControlRuntimeBootstrap) error {
-	paths, err := objToMap(bootstrap)
+// File is a representation of a certificate
+// or key file within the bootstrap context that contains
+// the contents of the file as well as a timestamp from
+// when the file was last modified.
+type File struct {
+	Timestamp time.Time
+	Content   []byte
+}
+
+type PathsDataformat map[string]File
+
+// WriteToDiskFromStorage writes the contents of the given reader to the paths
+// derived from within the ControlRuntimeBootstrap.
+func WriteToDiskFromStorage(r io.Reader, bootstrap *config.ControlRuntimeBootstrap) error {
+	paths, err := ObjToMap(bootstrap)
 	if err != nil {
 		return err
 	}
 
-	files := map[string][]byte{}
+	files := make(PathsDataformat)
 	if err := json.NewDecoder(r).Decode(&files); err != nil {
 		return err
 	}
 
-	for pathKey, data := range files {
+	for pathKey, bsf := range files {
 		path, ok := paths[pathKey]
 		if !ok {
 			continue
@@ -61,8 +85,7 @@ func Read(r io.Reader, bootstrap *config.ControlRuntimeBootstrap) error {
 		if err := os.MkdirAll(filepath.Dir(path), 0700); err != nil {
 			return errors.Wrapf(err, "failed to mkdir %s", filepath.Dir(path))
 		}
-
-		if err := ioutil.WriteFile(path, data, 0600); err != nil {
+		if err := ioutil.WriteFile(path, bsf.Content, 0600); err != nil {
 			return errors.Wrapf(err, "failed to write to %s", path)
 		}
 	}
@@ -70,7 +93,7 @@ func Read(r io.Reader, bootstrap *config.ControlRuntimeBootstrap) error {
 	return nil
 }
 
-func objToMap(obj interface{}) (map[string]string, error) {
+func ObjToMap(obj interface{}) (map[string]string, error) {
 	bytes, err := json.Marshal(obj)
 	if err != nil {
 		return nil, err

pkg/bootstrap/bootstrap_test.go

+package bootstrap
+
+import (
+	"testing"
+
+	"github.com/rancher/k3s/pkg/daemons/config"
+)
+
+func TestObjToMap(t *testing.T) {
+	type args struct {
+		obj interface{}
+	}
+	tests := []struct {
+		name    string
+		args    args
+		want    map[string]string
+		wantErr bool
+	}{
+		{
+			name: "Minimal Valid",
+			args: args{
+				obj: &config.ControlRuntimeBootstrap{
+					ServerCA:    "/var/lib/rancher/k3s/server/tls/server-ca.crt",
+					ServerCAKey: "/var/lib/rancher/k3s/server/tls/server-ca.key",
+				},
+			},
+			wantErr: false,
+		},
+		{
+			name: "Minimal Invalid",
+			args: args{
+				obj: 1,
+			},
+			wantErr: true,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			_, err := ObjToMap(tt.args.obj)
+			if (err != nil) != tt.wantErr {
+				t.Errorf("ObjToMap() error = %v, wantErr %v", err, tt.wantErr)
+				return
+			}
+		})
+	}
+}

pkg/cli/server/server.go

@@ -175,6 +175,17 @@ func run(app *cli.Context, cfg *cmds.Server, leaderControllers server.CustomCont
 		// delete local loadbalancers state for apiserver and supervisor servers
 		loadbalancer.ResetLoadBalancer(filepath.Join(cfg.DataDir, "agent"), loadbalancer.SupervisorServiceName)
 		loadbalancer.ResetLoadBalancer(filepath.Join(cfg.DataDir, "agent"), loadbalancer.APIServerServiceName)
+
+		// at this point we're doing a restore. Check to see if we've
+		// passed in a token and if not, check if the token file exists.
+		// If it doesn't, return an error indicating the token is necessary.
+		if cfg.Token == "" {
+			if _, err := os.Stat(filepath.Join(cfg.DataDir, "server/token")); err != nil {
+				if os.IsNotExist(err) {
+					return errors.New("")
+				}
+			}
+		}
 	}
 
 	serverConfig.ControlConfig.ClusterReset = cfg.ClusterReset

pkg/clientaccess/kubeconfig.go

@@ -9,7 +9,7 @@ import (
 )
 
 // WriteClientKubeConfig generates a kubeconfig at destFile that can be used to connect to a server at url with the given certs and keys
-func WriteClientKubeConfig(destFile string, url string, serverCAFile string, clientCertFile string, clientKeyFile string) error {
+func WriteClientKubeConfig(destFile, url, serverCAFile, clientCertFile, clientKeyFile string) error {
 	serverCA, err := ioutil.ReadFile(serverCAFile)
 	if err != nil {
 		return errors.Wrapf(err, "failed to read %s", serverCAFile)

pkg/clientaccess/token.go

@@ -16,9 +16,15 @@ import (
 	"github.com/sirupsen/logrus"
 )
 
-var (
+const (
+	tokenPrefix  = "K10"
+	tokenFormat  = "%s%s::%s:%s"
+	caHashLength = sha256.Size * 2
+
 	defaultClientTimeout = 10 * time.Second
+)
 
+var (
 	defaultClient = &http.Client{
 		Timeout: defaultClientTimeout,
 	}
@@ -32,12 +38,6 @@ var (
 	}
 )
 
-const (
-	tokenPrefix  = "K10"
-	tokenFormat  = "%s%s::%s:%s"
-	caHashLength = sha256.Size * 2
-)
-
 type OverrideURLCallback func(config []byte) (*url.URL, error)
 
 type Info struct {
@@ -49,8 +49,8 @@ type Info struct {
 }
 
 // String returns the token data, templated according to the token format
-func (info *Info) String() string {
-	return fmt.Sprintf(tokenFormat, tokenPrefix, hashCA(info.CACerts), info.Username, info.Password)
+func (i *Info) String() string {
+	return fmt.Sprintf(tokenFormat, tokenPrefix, hashCA(i.CACerts), i.Username, i.Password)
 }
 
 // ParseAndValidateToken parses a token, downloads and validates the server's CA bundle,
@@ -70,7 +70,7 @@ func ParseAndValidateToken(server string, token string) (*Info, error) {
 
 // ParseAndValidateToken parses a token with user override, downloads and
 // validates the server's CA bundle, and validates it according to the caHash from the token if set.
-func ParseAndValidateTokenForUser(server string, token string, username string) (*Info, error) {
+func ParseAndValidateTokenForUser(server, token, username string) (*Info, error) {
 	info, err := parseToken(token)
 	if err != nil {
 		return nil, err
@@ -86,11 +86,11 @@ func ParseAndValidateTokenForUser(server string, token string, username string)
 }
 
 // setAndValidateServer updates the remote server's cert info, and validates it against the provided hash
-func (info *Info) setAndValidateServer(server string) error {
-	if err := info.setServer(server); err != nil {
+func (i *Info) setAndValidateServer(server string) error {
+	if err := i.setServer(server); err != nil {
 		return err
 	}
-	return info.validateCAHash()
+	return i.validateCAHash()
 }
 
 // validateCACerts returns a boolean indicating whether or not a CA bundle matches the provided hash,
@@ -118,7 +118,7 @@ func ParseUsernamePassword(token string) (string, string, bool) {
 
 // parseToken parses a token into an Info struct
 func parseToken(token string) (*Info, error) {
-	var info = &Info{}
+	var info Info
 
 	if len(token) == 0 {
 		return nil, errors.New("token must not be empty")
@@ -150,7 +150,7 @@ func parseToken(token string) (*Info, error) {
 	info.Username = parts[0]
 	info.Password = parts[1]
 
-	return info, nil
+	return &info, nil
 }
 
 // GetHTTPClient returns a http client that validates TLS server certificates using the provided CA bundle.
@@ -177,25 +177,25 @@ func GetHTTPClient(cacerts []byte) *http.Client {
 }
 
 // Get makes a request to a subpath of info's BaseURL
-func (info *Info) Get(path string) ([]byte, error) {
-	u, err := url.Parse(info.BaseURL)
+func (i *Info) Get(path string) ([]byte, error) {
+	u, err := url.Parse(i.BaseURL)
 	if err != nil {
 		return nil, err
 	}
 	u.Path = path
-	return get(u.String(), GetHTTPClient(info.CACerts), info.Username, info.Password)
+	return get(u.String(), GetHTTPClient(i.CACerts), i.Username, i.Password)
 }
 
 // setServer sets the BaseURL and CACerts fields of the Info by connecting to the server
 // and storing the CA bundle.
-func (info *Info) setServer(server string) error {
+func (i *Info) setServer(server string) error {
 	url, err := url.Parse(server)
 	if err != nil {
 		return errors.Wrapf(err, "Invalid server url, failed to parse: %s", server)
 	}
 
 	if url.Scheme != "https" {
-		return fmt.Errorf("only https:// URLs are supported, invalid scheme: %s", server)
+		return errors.New("only https:// URLs are supported, invalid scheme: " + server)
 	}
 
 	for strings.HasSuffix(url.Path, "/") {
@@ -207,25 +207,25 @@ func (info *Info) setServer(server string) error {
 		return err
 	}
 
-	info.BaseURL = url.String()
-	info.CACerts = cacerts
+	i.BaseURL = url.String()
+	i.CACerts = cacerts
 	return nil
 }
 
 // ValidateCAHash validates that info's caHash matches the CACerts hash.
-func (info *Info) validateCAHash() error {
-	if len(info.caHash) > 0 && len(info.CACerts) == 0 {
+func (i *Info) validateCAHash() error {
+	if len(i.caHash) > 0 && len(i.CACerts) == 0 {
 		// Warn if the user provided a CA hash but we're not going to validate because it's already trusted
 		logrus.Warn("Cluster CA certificate is trusted by the host CA bundle. " +
 			"Token CA hash will not be validated.")
-	} else if len(info.caHash) == 0 && len(info.CACerts) > 0 {
+	} else if len(i.caHash) == 0 && len(i.CACerts) > 0 {
 		// Warn if the CA is self-signed but the user didn't provide a hash to validate it against
 		logrus.Warn("Cluster CA certificate is not trusted by the host CA bundle, but the token does not include a CA hash. " +
 			"Use the full token from the server's node-token file to enable Cluster CA validation.")
-	} else if len(info.CACerts) > 0 && len(info.caHash) > 0 {
+	} else if len(i.CACerts) > 0 && len(i.caHash) > 0 {
 		// only verify CA hash if the server cert is not trusted by the OS CA bundle
-		if ok, serverHash := validateCACerts(info.CACerts, info.caHash); !ok {
-			return fmt.Errorf("token CA hash does not match the Cluster CA certificate hash: %s != %s", info.caHash, serverHash)
+		if ok, serverHash := validateCACerts(i.CACerts, i.caHash); !ok {
+			return fmt.Errorf("token CA hash does not match the Cluster CA certificate hash: %s != %s", i.caHash, serverHash)
 		}
 	}
 	return nil
@@ -288,18 +288,18 @@ func get(u string, client *http.Client, username, password string) ([]byte, erro
 	return ioutil.ReadAll(resp.Body)
 }
 
-func FormatToken(token string, certFile string) (string, error) {
+func FormatToken(token, certFile string) (string, error) {
 	if len(token) == 0 {
 		return token, nil
 	}
 
 	certHash := ""
 	if len(certFile) > 0 {
-		bytes, err := ioutil.ReadFile(certFile)
+		b, err := ioutil.ReadFile(certFile)
 		if err != nil {
 			return "", nil
 		}
-		digest := sha256.Sum256(bytes)
+		digest := sha256.Sum256(b)
 		certHash = tokenPrefix + hex.EncodeToString(digest[:]) + "::"
 	}
 	return certHash + token, nil

pkg/clientaccess/token_test.go

@@ -320,7 +320,7 @@ func newTLSServer(t *testing.T, username, password string, sendWrongCA bool) *ht
 			}
 			bootstrapData := &config.ControlRuntimeBootstrap{}
 			w.Header().Set("Content-Type", "application/json")
-			if err := bootstrap.Write(w, bootstrapData); err != nil {
+			if err := bootstrap.ReadFromDisk(w, bootstrapData); err != nil {
 				t.Errorf("failed to write bootstrap: %v", err)
 			}
 			return

pkg/cluster/bootstrap_test.go

+package cluster
+
+import (
+	"bytes"
+	"context"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/k3s-io/kine/pkg/endpoint"
+	"github.com/rancher/k3s/pkg/bootstrap"
+	"github.com/rancher/k3s/pkg/clientaccess"
+	"github.com/rancher/k3s/pkg/cluster/managed"
+	"github.com/rancher/k3s/pkg/daemons/config"
+)
+
+func Test_isDirEmpty(t *testing.T) {
+	const tmpDir = "test_dir"
+
+	type args struct {
+		name string
+	}
+	tests := []struct {
+		name     string
+		args     args
+		setup    func() error
+		teardown func() error
+		want     bool
+		wantErr  bool
+	}{
+		{
+			name: "is empty",
+			args: args{
+				name: tmpDir,
+			},
+			setup: func() error {
+				return os.Mkdir(tmpDir, 0700)
+			},
+			teardown: func() error {
+				return os.RemoveAll(tmpDir)
+			},
+			want:    true,
+			wantErr: false,
+		},
+		{
+			name: "is not empty",
+			args: args{
+				name: tmpDir,
+			},
+			setup: func() error {
+				os.Mkdir(tmpDir, 0700)
+				_, _ = os.Create(filepath.Join(filepath.Dir(tmpDir), tmpDir, "test_file"))
+				return nil
+			},
+			teardown: func() error {
+				return os.RemoveAll(tmpDir)
+			},
+			want:    false,
+			wantErr: false,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			defer tt.teardown()
+			if err := tt.setup(); err != nil {
+				t.Errorf("Setup for isDirEmpty() failed = %v", err)
+				return
+			}
+			got, err := isDirEmpty(tt.args.name)
+			if (err != nil) != tt.wantErr {
+				t.Errorf("isDirEmpty() error = %v, wantErr %v", err, tt.wantErr)
+				return
+			}
+			if got != tt.want {
+				t.Errorf("isDirEmpty() = %+v\nWant = %+v", got, tt.want)
+			}
+		})
+	}
+}
+
+func TestCluster_certDirsExist(t *testing.T) {
+	const testDataDir = "/tmp/k3s/"
+
+	testTLSDir := filepath.Join(testDataDir, "server", "tls")
+	testTLSEtcdDir := filepath.Join(testDataDir, "server", "tls", "etcd")
+
+	type fields struct {
+		clientAccessInfo *clientaccess.Info
+		config           *config.Control
+		runtime          *config.ControlRuntime
+		managedDB        managed.Driver
+		etcdConfig       endpoint.ETCDConfig
+		shouldBootstrap  bool
+		storageStarted   bool
+		saveBootstrap    bool
+	}
+	tests := []struct {
+		name     string
+		fields   fields
+		setup    func() error
+		teardown func() error
+		wantErr  bool
+	}{
+		{
+			name: "exists",
+			fields: fields{
+				config: &config.Control{
+					DataDir: filepath.Join(testDataDir, "server"),
+				},
+			},
+			setup: func() error {
+				os.MkdirAll(testTLSEtcdDir, 0700)
+
+				_, _ = os.Create(filepath.Join(testTLSDir, "test_file"))
+				_, _ = os.Create(filepath.Join(testTLSEtcdDir, "test_file"))
+
+				return nil
+			},
+			teardown: func() error {
+				return os.RemoveAll(testDataDir)
+			},
+			wantErr: false,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			c := &Cluster{
+				clientAccessInfo: tt.fields.clientAccessInfo,
+				config:           tt.fields.config,
+				runtime:          tt.fields.runtime,
+				managedDB:        tt.fields.managedDB,
+				etcdConfig:       tt.fields.etcdConfig,
+				storageStarted:   tt.fields.storageStarted,
+				saveBootstrap:    tt.fields.saveBootstrap,
+			}
+			defer tt.teardown()
+			if err := tt.setup(); err != nil {
+				t.Errorf("Setup for Cluster.certDirsExist() failed = %v", err)
+				return
+			}
+			if err := c.certDirsExist(); (err != nil) != tt.wantErr {
+				t.Errorf("Cluster.certDirsExist() error = %v, wantErr %v", err, tt.wantErr)
+			}
+		})
+	}
+}
+
+func TestCluster_migrateBootstrapData(t *testing.T) {
+	type fields struct {
+		clientAccessInfo *clientaccess.Info
+		config           *config.Control
+		runtime          *config.ControlRuntime
+		managedDB        managed.Driver
+		etcdConfig       endpoint.ETCDConfig
+		joining          bool
+		storageStarted   bool
+		saveBootstrap    bool
+		shouldBootstrap  bool
+	}
+	type args struct {
+		ctx   context.Context
+		data  *bytes.Buffer
+		files bootstrap.PathsDataformat
+	}
+	tests := []struct {
+		name     string
+		args     args
+		setup    func() error // Optional, delete if unused
+		teardown func() error // Optional, delete if unused
+		wantErr  bool
+	}{
+		{
+			name: "Success",
+			args: args{
+				ctx:   context.Background(),
+				data:  bytes.NewBuffer([]byte(`{"ServerCA": "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"}`)),
+				files: make(bootstrap.PathsDataformat),
+			},
+			wantErr: false,
+		},
+		{
+			name: "Invalid Old Format",
+			args: args{
+				ctx:  context.Background(),
+				data: &bytes.Buffer{},
+				files: bootstrap.PathsDataformat{
+					"ServerCA": bootstrap.File{
+						Timestamp: time.Now(),
+					},
+				},
+			},
+			wantErr: true,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if err := migrateBootstrapData(tt.args.ctx, tt.args.data, tt.args.files); (err != nil) != tt.wantErr {
+				t.Errorf("Cluster.migrateBootstrapData() error = %v, wantErr %v", err, tt.wantErr)
+			}
+		})
+	}
+}
+
+func TestCluster_Snapshot(t *testing.T) {
+	type fields struct {
+		clientAccessInfo *clientaccess.Info
+		config           *config.Control
+		runtime          *config.ControlRuntime
+		managedDB        managed.Driver
+		etcdConfig       endpoint.ETCDConfig
+		joining          bool
+		storageStarted   bool
+		saveBootstrap    bool
+		shouldBootstrap  bool
+	}
+	type args struct {
+		ctx    context.Context
+		config *config.Control
+	}
+	tests := []struct {
+		name    string
+		fields  fields
+		args    args
+		wantErr bool
+	}{
+		{
+			name:   "Fail on non etcd cluster",
+			fields: fields{},
+			args: args{
+				ctx: context.Background(),
+			},
+			wantErr: true,
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			c := &Cluster{
+				clientAccessInfo: tt.fields.clientAccessInfo,
+				config:           tt.fields.config,
+				runtime:          tt.fields.runtime,
+				managedDB:        tt.fields.managedDB,
+				etcdConfig:       tt.fields.etcdConfig,
+				joining:          tt.fields.joining,
+				storageStarted:   tt.fields.storageStarted,
+				saveBootstrap:    tt.fields.saveBootstrap,
+				shouldBootstrap:  tt.fields.shouldBootstrap,
+			}
+			if err := c.Snapshot(tt.args.ctx, tt.args.config); (err != nil) != tt.wantErr {
+				t.Errorf("Cluster.Snapshot() error = %v, wantErr %v", err, tt.wantErr)
+			}
+		})
+	}
+}

pkg/cluster/cluster.go

@@ -20,11 +20,11 @@ type Cluster struct {
 	config           *config.Control
 	runtime          *config.ControlRuntime
 	managedDB        managed.Driver
-	shouldBootstrap  bool
-	storageStarted   bool
 	etcdConfig       endpoint.ETCDConfig
 	joining          bool
+	storageStarted   bool
 	saveBootstrap    bool
+	shouldBootstrap  bool
 }
 
 // Start creates the dynamic tls listener, http request handler,
@@ -81,14 +81,7 @@ func (c *Cluster) Start(ctx context.Context) (<-chan struct{}, error) {
 
 	// if necessary, store bootstrap data to datastore
 	if c.saveBootstrap {
-		if err := c.save(ctx); err != nil {
-			return nil, err
-		}
-	}
-
-	// if necessary, record successful bootstrap
-	if c.shouldBootstrap {
-		if err := c.bootstrapped(); err != nil {
+		if err := c.save(ctx, false); err != nil {
 			return nil, err
 		}
 	}
@@ -106,7 +99,7 @@ func (c *Cluster) Start(ctx context.Context) (<-chan struct{}, error) {
 			for {
 				select {
 				case <-ready:
-					if err := c.save(ctx); err != nil {
+					if err := c.save(ctx, false); err != nil {
 						panic(err)
 					}
 
@@ -153,7 +146,7 @@ func (c *Cluster) startStorage(ctx context.Context) error {
 	return nil
 }
 
-// New creates an initial cluster using the provided configuration
+// New creates an initial cluster using the provided configuration.
 func New(config *config.Control) *Cluster {
 	return &Cluster{
 		config:  config,

pkg/cluster/https.go

@@ -79,7 +79,9 @@ func (c *Cluster) initClusterAndHTTPS(ctx context.Context) error {
 
 	// Create a HTTP server with the registered request handlers, using logrus for logging
 	server := http.Server{
-		Handler: handler}
+		Handler: handler,
+	}
+
 	if logrus.IsLevelEnabled(logrus.DebugLevel) {
 		server.ErrorLog = log.New(logrus.StandardLogger().Writer(), "Cluster-Http-Server ", log.LstdFlags)
 	} else {

pkg/cluster/storage.go

@@ -19,9 +19,9 @@ import (
 // snapshot of the cluster's CA certs and keys, encryption passphrases, etc - encrypted with the join token.
 // This is used when bootstrapping a cluster from a managed database or external etcd cluster.
 // This is NOT used with embedded etcd, which bootstraps over HTTP.
-func (c *Cluster) save(ctx context.Context) error {
+func (c *Cluster) save(ctx context.Context, override bool) error {
 	buf := &bytes.Buffer{}
-	if err := bootstrap.Write(buf, &c.runtime.ControlRuntimeBootstrap); err != nil {
+	if err := bootstrap.ReadFromDisk(buf, &c.runtime.ControlRuntimeBootstrap); err != nil {
 		return err
 	}
 	token := c.config.Token
@@ -47,14 +47,20 @@ func (c *Cluster) save(ctx context.Context) error {
 		return err
 	}
 
-	_, err = c.getBootstrapKeyFromStorage(ctx, storageClient, normalizedToken, token)
-	if err != nil {
+	if _, err := c.getBootstrapKeyFromStorage(ctx, storageClient, normalizedToken, token); err != nil {
 		return err
 	}
 
 	if err := storageClient.Create(ctx, storageKey(normalizedToken), data); err != nil {
 		if err.Error() == "key exists" {
-			logrus.Warnln("bootstrap key already exists")
+			logrus.Warn("bootstrap key already exists")
+			if override {
+				bsd, err := c.bootstrapKeyData(ctx, storageClient)
+				if err != nil {
+					return err
+				}
+				return storageClient.Update(ctx, storageKey(normalizedToken), bsd.Modified, data)
+			}
 			return nil
 		} else if strings.Contains(err.Error(), "not supported for learner") {
 			logrus.Debug("skipping bootstrap data save on learner")
@@ -66,9 +72,25 @@ func (c *Cluster) save(ctx context.Context) error {
 	return nil
 }
 
+// bootstrapKeyData lists keys stored in the datastore with the prefix "/bootstrap", and
+// will return the first such key. It will return an error if not exactly one key is found.
+func (c *Cluster) bootstrapKeyData(ctx context.Context, storageClient client.Client) (*client.Value, error) {
+	bootstrapList, err := storageClient.List(ctx, "/bootstrap", 0)
+	if err != nil {
+		return nil, err
+	}
+	if len(bootstrapList) == 0 {
+		return nil, errors.New("no bootstrap data found")
+	}
+	if len(bootstrapList) > 1 {
+		return nil, errors.New("found multiple bootstrap keys in storage")
+	}
+	return &bootstrapList[0], nil
+}
+
 // storageBootstrap loads data from the datastore into the ControlRuntimeBootstrap struct.
 // The storage key and encryption passphrase are both derived from the join token.
-// token is either passed
+// token is either passed.
 func (c *Cluster) storageBootstrap(ctx context.Context) error {
 	if err := c.startStorage(ctx); err != nil {
 		return err
@@ -110,7 +132,7 @@ func (c *Cluster) storageBootstrap(ctx context.Context) error {
 		return err
 	}
 
-	return bootstrap.Read(bytes.NewBuffer(data), &c.runtime.ControlRuntimeBootstrap)
+	return c.ReconcileBootstrapData(ctx, bytes.NewReader(data), &c.config.Runtime.ControlRuntimeBootstrap)
 }
 
 // getBootstrapKeyFromStorage will list all keys that has prefix /bootstrap and will check for key that is
@@ -157,6 +179,7 @@ func (c *Cluster) getBootstrapKeyFromStorage(ctx context.Context, storageClient
 // found then it will still strip the token from any additional info
 func readTokenFromFile(serverToken, certs, dataDir string) (string, error) {
 	tokenFile := filepath.Join(dataDir, "token")
+
 	b, err := ioutil.ReadFile(tokenFile)
 	if err != nil {
 		if os.IsNotExist(err) {
@@ -168,6 +191,7 @@ func readTokenFromFile(serverToken, certs, dataDir string) (string, error) {
 		}
 		return "", err
 	}
+
 	// strip the token from any new line if its read from file
 	return string(bytes.TrimRight(b, "\n")), nil
 }
@@ -178,6 +202,7 @@ func normalizeToken(token string) (string, error) {
 	if !ok {
 		return password, errors.New("failed to normalize token")
 	}
+
 	return password, nil
 }
 
@@ -186,6 +211,7 @@ func normalizeToken(token string) (string, error) {
 // then migrate those and resave only with the normalized token
 func (c *Cluster) migrateOldTokens(ctx context.Context, bootstrapList []client.Value, storageClient client.Client, emptyStringKey, tokenKey, token, oldToken string) error {
 	oldTokenKey := storageKey(oldToken)
+
 	for _, bootstrapKV := range bootstrapList {
 		// checking for empty string bootstrap key
 		if string(bootstrapKV.Key) == emptyStringKey {
@@ -200,6 +226,7 @@ func (c *Cluster) migrateOldTokens(ctx context.Context, bootstrapList []client.V
 			}
 		}
 	}
+
 	return nil
 }
 
@@ -209,10 +236,12 @@ func doMigrateToken(ctx context.Context, storageClient client.Client, keyValue c
 	if err != nil {
 		return err
 	}
+
 	encryptedData, err := encrypt(newToken, data)
 	if err != nil {
 		return err
 	}
+
 	// saving the new encrypted data with the right token key
 	if err := storageClient.Create(ctx, newTokenKey, encryptedData); err != nil {
 		if err.Error() == "key exists" {
@@ -224,10 +253,12 @@ func doMigrateToken(ctx context.Context, storageClient client.Client, keyValue c
 			return err
 		}
 	}
+
 	logrus.Infof("created bootstrap key %s", newTokenKey)
 	// deleting the old key
 	if err := storageClient.Delete(ctx, oldTokenKey, keyValue.Modified); err != nil {
 		logrus.Warnf("failed to delete old bootstrap key %s", oldTokenKey)
 	}
+
 	return nil
 }

pkg/daemons/control/deps/deps.go

@@ -89,6 +89,67 @@ func KubeConfig(dest, url, caCert, clientCert, clientKey string) error {
 	return kubeconfigTemplate.Execute(output, &data)
 }
 
+// CreateRuntimeCertFiles is responsible for filling out all the
+// .crt and .key filenames for a ControlRuntime.
+func CreateRuntimeCertFiles(config *config.Control, runtime *config.ControlRuntime) {
+	runtime.ClientCA = filepath.Join(config.DataDir, "tls", "client-ca.crt")
+	runtime.ClientCAKey = filepath.Join(config.DataDir, "tls", "client-ca.key")
+	runtime.ServerCA = filepath.Join(config.DataDir, "tls", "server-ca.crt")
+	runtime.ServerCAKey = filepath.Join(config.DataDir, "tls", "server-ca.key")
+	runtime.RequestHeaderCA = filepath.Join(config.DataDir, "tls", "request-header-ca.crt")
+	runtime.RequestHeaderCAKey = filepath.Join(config.DataDir, "tls", "request-header-ca.key")
+	runtime.IPSECKey = filepath.Join(config.DataDir, "cred", "ipsec.psk")
+
+	runtime.ServiceKey = filepath.Join(config.DataDir, "tls", "service.key")
+	runtime.PasswdFile = filepath.Join(config.DataDir, "cred", "passwd")
+	runtime.NodePasswdFile = filepath.Join(config.DataDir, "cred", "node-passwd")
+
+	runtime.KubeConfigAdmin = filepath.Join(config.DataDir, "cred", "admin.kubeconfig")
+	runtime.KubeConfigController = filepath.Join(config.DataDir, "cred", "controller.kubeconfig")
+	runtime.KubeConfigScheduler = filepath.Join(config.DataDir, "cred", "scheduler.kubeconfig")
+	runtime.KubeConfigAPIServer = filepath.Join(config.DataDir, "cred", "api-server.kubeconfig")
+	runtime.KubeConfigCloudController = filepath.Join(config.DataDir, "cred", "cloud-controller.kubeconfig")
+
+	runtime.ClientAdminCert = filepath.Join(config.DataDir, "tls", "client-admin.crt")
+	runtime.ClientAdminKey = filepath.Join(config.DataDir, "tls", "client-admin.key")
+	runtime.ClientControllerCert = filepath.Join(config.DataDir, "tls", "client-controller.crt")
+	runtime.ClientControllerKey = filepath.Join(config.DataDir, "tls", "client-controller.key")
+	runtime.ClientCloudControllerCert = filepath.Join(config.DataDir, "tls", "client-"+version.Program+"-cloud-controller.crt")
+	runtime.ClientCloudControllerKey = filepath.Join(config.DataDir, "tls", "client-"+version.Program+"-cloud-controller.key")
+	runtime.ClientSchedulerCert = filepath.Join(config.DataDir, "tls", "client-scheduler.crt")
+	runtime.ClientSchedulerKey = filepath.Join(config.DataDir, "tls", "client-scheduler.key")
+	runtime.ClientKubeAPICert = filepath.Join(config.DataDir, "tls", "client-kube-apiserver.crt")
+	runtime.ClientKubeAPIKey = filepath.Join(config.DataDir, "tls", "client-kube-apiserver.key")
+	runtime.ClientKubeProxyCert = filepath.Join(config.DataDir, "tls", "client-kube-proxy.crt")
+	runtime.ClientKubeProxyKey = filepath.Join(config.DataDir, "tls", "client-kube-proxy.key")
+	runtime.ClientK3sControllerCert = filepath.Join(config.DataDir, "tls", "client-"+version.Program+"-controller.crt")
+	runtime.ClientK3sControllerKey = filepath.Join(config.DataDir, "tls", "client-"+version.Program+"-controller.key")
+
+	runtime.ServingKubeAPICert = filepath.Join(config.DataDir, "tls", "serving-kube-apiserver.crt")
+	runtime.ServingKubeAPIKey = filepath.Join(config.DataDir, "tls", "serving-kube-apiserver.key")
+
+	runtime.ClientKubeletKey = filepath.Join(config.DataDir, "tls", "client-kubelet.key")
+	runtime.ServingKubeletKey = filepath.Join(config.DataDir, "tls", "serving-kubelet.key")
+
+	runtime.ClientAuthProxyCert = filepath.Join(config.DataDir, "tls", "client-auth-proxy.crt")
+	runtime.ClientAuthProxyKey = filepath.Join(config.DataDir, "tls", "client-auth-proxy.key")
+
+	runtime.ETCDServerCA = filepath.Join(config.DataDir, "tls", "etcd", "server-ca.crt")
+	runtime.ETCDServerCAKey = filepath.Join(config.DataDir, "tls", "etcd", "server-ca.key")
+	runtime.ETCDPeerCA = filepath.Join(config.DataDir, "tls", "etcd", "peer-ca.crt")
+	runtime.ETCDPeerCAKey = filepath.Join(config.DataDir, "tls", "etcd", "peer-ca.key")
+	runtime.ServerETCDCert = filepath.Join(config.DataDir, "tls", "etcd", "server-client.crt")
+	runtime.ServerETCDKey = filepath.Join(config.DataDir, "tls", "etcd", "server-client.key")
+	runtime.PeerServerClientETCDCert = filepath.Join(config.DataDir, "tls", "etcd", "peer-server-client.crt")
+	runtime.PeerServerClientETCDKey = filepath.Join(config.DataDir, "tls", "etcd", "peer-server-client.key")
+	runtime.ClientETCDCert = filepath.Join(config.DataDir, "tls", "etcd", "client.crt")
+	runtime.ClientETCDKey = filepath.Join(config.DataDir, "tls", "etcd", "client.key")
+
+	if config.EncryptSecrets {
+		runtime.EncryptionConfig = filepath.Join(config.DataDir, "cred", "encryption-config.json")
+	}
+}
+
 // GenServerDeps is responsible for generating the cluster dependencies
 // needed to successfully bootstrap a cluster.
 func GenServerDeps(config *config.Control, runtime *config.ControlRuntime) error {

pkg/daemons/control/server.go

@@ -239,62 +239,7 @@ func prepare(ctx context.Context, config *config.Control, runtime *config.Contro
 	os.MkdirAll(filepath.Join(config.DataDir, "tls"), 0700)
 	os.MkdirAll(filepath.Join(config.DataDir, "cred"), 0700)
 
-	runtime.ClientCA = filepath.Join(config.DataDir, "tls", "client-ca.crt")
-	runtime.ClientCAKey = filepath.Join(config.DataDir, "tls", "client-ca.key")
-	runtime.ServerCA = filepath.Join(config.DataDir, "tls", "server-ca.crt")
-	runtime.ServerCAKey = filepath.Join(config.DataDir, "tls", "server-ca.key")
-	runtime.RequestHeaderCA = filepath.Join(config.DataDir, "tls", "request-header-ca.crt")
-	runtime.RequestHeaderCAKey = filepath.Join(config.DataDir, "tls", "request-header-ca.key")
-	runtime.IPSECKey = filepath.Join(config.DataDir, "cred", "ipsec.psk")
-
-	runtime.ServiceKey = filepath.Join(config.DataDir, "tls", "service.key")
-	runtime.PasswdFile = filepath.Join(config.DataDir, "cred", "passwd")
-	runtime.NodePasswdFile = filepath.Join(config.DataDir, "cred", "node-passwd")
-
-	runtime.KubeConfigAdmin = filepath.Join(config.DataDir, "cred", "admin.kubeconfig")
-	runtime.KubeConfigController = filepath.Join(config.DataDir, "cred", "controller.kubeconfig")
-	runtime.KubeConfigScheduler = filepath.Join(config.DataDir, "cred", "scheduler.kubeconfig")
-	runtime.KubeConfigAPIServer = filepath.Join(config.DataDir, "cred", "api-server.kubeconfig")
-	runtime.KubeConfigCloudController = filepath.Join(config.DataDir, "cred", "cloud-controller.kubeconfig")
-
-	runtime.ClientAdminCert = filepath.Join(config.DataDir, "tls", "client-admin.crt")
-	runtime.ClientAdminKey = filepath.Join(config.DataDir, "tls", "client-admin.key")
-	runtime.ClientControllerCert = filepath.Join(config.DataDir, "tls", "client-controller.crt")
-	runtime.ClientControllerKey = filepath.Join(config.DataDir, "tls", "client-controller.key")
-	runtime.ClientCloudControllerCert = filepath.Join(config.DataDir, "tls", "client-"+version.Program+"-cloud-controller.crt")
-	runtime.ClientCloudControllerKey = filepath.Join(config.DataDir, "tls", "client-"+version.Program+"-cloud-controller.key")
-	runtime.ClientSchedulerCert = filepath.Join(config.DataDir, "tls", "client-scheduler.crt")
-	runtime.ClientSchedulerKey = filepath.Join(config.DataDir, "tls", "client-scheduler.key")
-	runtime.ClientKubeAPICert = filepath.Join(config.DataDir, "tls", "client-kube-apiserver.crt")
-	runtime.ClientKubeAPIKey = filepath.Join(config.DataDir, "tls", "client-kube-apiserver.key")
-	runtime.ClientKubeProxyCert = filepath.Join(config.DataDir, "tls", "client-kube-proxy.crt")
-	runtime.ClientKubeProxyKey = filepath.Join(config.DataDir, "tls", "client-kube-proxy.key")
-	runtime.ClientK3sControllerCert = filepath.Join(config.DataDir, "tls", "client-"+version.Program+"-controller.crt")
-	runtime.ClientK3sControllerKey = filepath.Join(config.DataDir, "tls", "client-"+version.Program+"-controller.key")
-
-	runtime.ServingKubeAPICert = filepath.Join(config.DataDir, "tls", "serving-kube-apiserver.crt")
-	runtime.ServingKubeAPIKey = filepath.Join(config.DataDir, "tls", "serving-kube-apiserver.key")
-
-	runtime.ClientKubeletKey = filepath.Join(config.DataDir, "tls", "client-kubelet.key")
-	runtime.ServingKubeletKey = filepath.Join(config.DataDir, "tls", "serving-kubelet.key")
-
-	runtime.ClientAuthProxyCert = filepath.Join(config.DataDir, "tls", "client-auth-proxy.crt")
-	runtime.ClientAuthProxyKey = filepath.Join(config.DataDir, "tls", "client-auth-proxy.key")
-
-	runtime.ETCDServerCA = filepath.Join(config.DataDir, "tls", "etcd", "server-ca.crt")
-	runtime.ETCDServerCAKey = filepath.Join(config.DataDir, "tls", "etcd", "server-ca.key")
-	runtime.ETCDPeerCA = filepath.Join(config.DataDir, "tls", "etcd", "peer-ca.crt")
-	runtime.ETCDPeerCAKey = filepath.Join(config.DataDir, "tls", "etcd", "peer-ca.key")
-	runtime.ServerETCDCert = filepath.Join(config.DataDir, "tls", "etcd", "server-client.crt")
-	runtime.ServerETCDKey = filepath.Join(config.DataDir, "tls", "etcd", "server-client.key")
-	runtime.PeerServerClientETCDCert = filepath.Join(config.DataDir, "tls", "etcd", "peer-server-client.crt")
-	runtime.PeerServerClientETCDKey = filepath.Join(config.DataDir, "tls", "etcd", "peer-server-client.key")
-	runtime.ClientETCDCert = filepath.Join(config.DataDir, "tls", "etcd", "client.crt")
-	runtime.ClientETCDKey = filepath.Join(config.DataDir, "tls", "etcd", "client.key")
-
-	if config.EncryptSecrets {
-		runtime.EncryptionConfig = filepath.Join(config.DataDir, "cred", "encryption-config.json")
-	}
+	deps.CreateRuntimeCertFiles(config, runtime)
 
 	cluster := cluster.New(config)
 

pkg/etcd/etcd.go

@@ -178,7 +178,7 @@ func (e *ETCD) IsInitialized(ctx context.Context, config *config.Control) (bool,
 	} else if os.IsNotExist(err) {
 		return false, nil
 	} else {
-		return false, errors.Wrapf(err, "invalid state for wal directory %s", dir)
+		return false, errors.Wrap(err, "invalid state for wal directory "+dir)
 	}
 }
 

tests/util/runtime.go

+package util
+
+import (
+	"net"
+	"os"
+	"path/filepath"
+
+	"github.com/rancher/k3s/pkg/daemons/config"
+	"github.com/rancher/k3s/pkg/daemons/control/deps"
+)
+
+// GenerateDataDir creates a temporary directory at "/tmp/k3s/<RANDOM_STRING>/".
+// The latest directory created with this function is soft linked to "/tmp/k3s/latest/".
+// This allows tests to replicate the "/var/lib/rancher/k3s" directory structure.
+func GenerateDataDir(cnf *config.Control) error {
+	if err := os.MkdirAll(cnf.DataDir, 0700); err != nil {
+		return err
+	}
+	testDir, err := os.MkdirTemp(cnf.DataDir, "*")
+	if err != nil {
+		return err
+	}
+	// Remove old symlink and add new one
+	os.Remove(filepath.Join(cnf.DataDir, "latest"))
+	if err = os.Symlink(testDir, filepath.Join(cnf.DataDir, "latest")); err != nil {
+		return err
+	}
+	cnf.DataDir = testDir
+	cnf.DataDir, err = filepath.Abs(cnf.DataDir)
+	if err != nil {
+		return err
+	}
+	return nil
+}
+
+// CleanupDataDir removes the associated "/tmp/k3s/<RANDOM_STRING>"
+// directory.
+func CleanupDataDir(cnf *config.Control) {
+	os.RemoveAll(cnf.DataDir)
+}
+
+// GenerateRuntime creates a temporary data dir and configures
+// config.ControlRuntime with all the appropriate certificate keys.
+func GenerateRuntime(cnf *config.Control) error {
+	runtime := &config.ControlRuntime{}
+	if err := GenerateDataDir(cnf); err != nil {
+		return err
+	}
+	runtime.AgentReady = make(chan struct{})
+
+	os.MkdirAll(filepath.Join(cnf.DataDir, "tls"), 0700)
+	os.MkdirAll(filepath.Join(cnf.DataDir, "cred"), 0700)
+
+	deps.CreateRuntimeCertFiles(cnf, runtime)
+
+	if err := deps.GenServerDeps(cnf, runtime); err != nil {
+		return err
+	}
+	cnf.Runtime = runtime
+	return nil
+}
+
+func ClusterIPNet() *net.IPNet {
+	_, clusterIPNet, _ := net.ParseCIDR("10.42.0.0/16")
+	return clusterIPNet
+}
+
+func ServiceIPNet() *net.IPNet {
+	_, serviceIPNet, _ := net.ParseCIDR("10.43.0.0/16")
+	return serviceIPNet
+}

vendor/modules.txt

@@ -1251,7 +1251,7 @@ golang.org/x/oauth2/jwt
 golang.org/x/sync/errgroup
 golang.org/x/sync/semaphore
 golang.org/x/sync/singleflight
-# golang.org/x/sys v0.0.0-20210426230700-d19ff857e887 => golang.org/x/sys v0.0.0-20210225134936-a50acf3fe073
+# golang.org/x/sys v0.0.0-20210603081109-ebe580a85c40 => golang.org/x/sys v0.0.0-20210225134936-a50acf3fe073
 ## explicit
 golang.org/x/sys/cpu
 golang.org/x/sys/execabs
@@ -1357,7 +1357,7 @@ google.golang.org/appengine/urlfetch
 google.golang.org/genproto/googleapis/api/httpbody
 google.golang.org/genproto/googleapis/rpc/status
 google.golang.org/genproto/protobuf/field_mask
-# google.golang.org/grpc v1.37.0 => google.golang.org/grpc v1.27.1
+# google.golang.org/grpc v1.38.0 => google.golang.org/grpc v1.27.1
 ## explicit
 google.golang.org/grpc
 google.golang.org/grpc/attributes