Move deps from _workspace/ to vendor/

godep restore
pushd $GOPATH/src/github.com/appc/spec
git co master
popd
go get go4.org/errorutil
rm -rf Godeps
godep save ./...
git add vendor
git add -f $(git ls-files --other vendor/)
git co -- Godeps/LICENSES Godeps/.license_file_state Godeps/OWNERS

vendor/github.com/coreos/etcd/etcdserver/api/cluster.go

@@ -0,0 +1,41 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package api
+
+import (
+	"github.com/coreos/etcd/etcdserver/membership"
+	"github.com/coreos/etcd/pkg/types"
+
+	"github.com/coreos/go-semver/semver"
+)
+
+// Cluster is an interface representing a collection of members in one etcd cluster.
+type Cluster interface {
+	// ID returns the cluster ID
+	ID() types.ID
+	// ClientURLs returns an aggregate set of all URLs on which this
+	// cluster is listening for client requests
+	ClientURLs() []string
+	// Members returns a slice of members sorted by their ID
+	Members() []*membership.Member
+	// Member retrieves a particular member based on ID, or nil if the
+	// member does not exist in the cluster
+	Member(id types.ID) *membership.Member
+	// IsIDRemoved checks whether the given ID has been removed from this
+	// cluster at some point in the past
+	IsIDRemoved(id types.ID) bool
+	// Version is the cluster-wide minimum major.minor version.
+	Version() *semver.Version
+}

vendor/github.com/coreos/etcd/etcdserver/api/v2http/capability.go

@@ -0,0 +1,101 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v2http
+
+import (
+	"fmt"
+	"net/http"
+	"sync"
+	"time"
+
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/api/v2http/httptypes"
+	"github.com/coreos/go-semver/semver"
+)
+
+type capability string
+
+const (
+	authCapability capability = "auth"
+)
+
+var (
+	// capabilityMaps is a static map of version to capability map.
+	// the base capabilities is the set of capability 2.0 supports.
+	capabilityMaps = map[string]map[capability]bool{
+		"2.1.0": {authCapability: true},
+		"2.2.0": {authCapability: true},
+		"2.3.0": {authCapability: true},
+	}
+
+	enableMapMu sync.Mutex
+	// enabledMap points to a map in capabilityMaps
+	enabledMap map[capability]bool
+)
+
+// capabilityLoop checks the cluster version every 500ms and updates
+// the enabledMap when the cluster version increased.
+// capabilityLoop MUST be ran in a goroutine before checking capability
+// or using capabilityHandler.
+func capabilityLoop(s *etcdserver.EtcdServer) {
+	stopped := s.StopNotify()
+
+	var pv *semver.Version
+	for {
+		if v := s.ClusterVersion(); v != pv {
+			if pv == nil {
+				pv = v
+			} else if v != nil && pv.LessThan(*v) {
+				pv = v
+			}
+			enableMapMu.Lock()
+			enabledMap = capabilityMaps[pv.String()]
+			enableMapMu.Unlock()
+			plog.Infof("enabled capabilities for version %s", pv)
+		}
+
+		select {
+		case <-stopped:
+			return
+		case <-time.After(500 * time.Millisecond):
+		}
+	}
+}
+
+func isCapabilityEnabled(c capability) bool {
+	enableMapMu.Lock()
+	defer enableMapMu.Unlock()
+	if enabledMap == nil {
+		return false
+	}
+	return enabledMap[c]
+}
+
+func capabilityHandler(c capability, fn func(http.ResponseWriter, *http.Request)) http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		if !isCapabilityEnabled(c) {
+			notCapable(w, r, c)
+			return
+		}
+		fn(w, r)
+	}
+}
+
+func notCapable(w http.ResponseWriter, r *http.Request, c capability) {
+	herr := httptypes.NewHTTPError(http.StatusInternalServerError, fmt.Sprintf("Not capable of accessing %s feature during rolling upgrades.", c))
+	if err := herr.WriteTo(w); err != nil {
+		plog.Debugf("error writing HTTPError (%v) to %s", err, r.RemoteAddr)
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/api/v2http/client.go

@@ -0,0 +1,830 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v2http
+
+import (
+	"encoding/json"
+	"errors"
+	"expvar"
+	"fmt"
+	"io/ioutil"
+	"net/http"
+	"net/http/pprof"
+	"net/url"
+	"path"
+	"strconv"
+	"strings"
+	"time"
+
+	etcdErr "github.com/coreos/etcd/error"
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/api"
+	"github.com/coreos/etcd/etcdserver/api/v2http/httptypes"
+	"github.com/coreos/etcd/etcdserver/auth"
+	"github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/etcdserver/membership"
+	"github.com/coreos/etcd/etcdserver/stats"
+	"github.com/coreos/etcd/pkg/types"
+	"github.com/coreos/etcd/raft"
+	"github.com/coreos/etcd/store"
+	"github.com/coreos/etcd/version"
+	"github.com/coreos/pkg/capnslog"
+	"github.com/jonboulle/clockwork"
+	"github.com/prometheus/client_golang/prometheus"
+	"golang.org/x/net/context"
+)
+
+const (
+	authPrefix               = "/v2/auth"
+	keysPrefix               = "/v2/keys"
+	deprecatedMachinesPrefix = "/v2/machines"
+	membersPrefix            = "/v2/members"
+	statsPrefix              = "/v2/stats"
+	varsPath                 = "/debug/vars"
+	metricsPath              = "/metrics"
+	healthPath               = "/health"
+	versionPath              = "/version"
+	configPath               = "/config"
+	pprofPrefix              = "/debug/pprof"
+)
+
+// NewClientHandler generates a muxed http.Handler with the given parameters to serve etcd client requests.
+func NewClientHandler(server *etcdserver.EtcdServer, timeout time.Duration) http.Handler {
+	go capabilityLoop(server)
+
+	sec := auth.NewStore(server, timeout)
+
+	kh := &keysHandler{
+		sec:     sec,
+		server:  server,
+		cluster: server.Cluster(),
+		timer:   server,
+		timeout: timeout,
+	}
+
+	sh := &statsHandler{
+		stats: server,
+	}
+
+	mh := &membersHandler{
+		sec:     sec,
+		server:  server,
+		cluster: server.Cluster(),
+		timeout: timeout,
+		clock:   clockwork.NewRealClock(),
+	}
+
+	dmh := &deprecatedMachinesHandler{
+		cluster: server.Cluster(),
+	}
+
+	sech := &authHandler{
+		sec:     sec,
+		cluster: server.Cluster(),
+	}
+
+	mux := http.NewServeMux()
+	mux.HandleFunc("/", http.NotFound)
+	mux.Handle(healthPath, healthHandler(server))
+	mux.HandleFunc(versionPath, versionHandler(server.Cluster(), serveVersion))
+	mux.Handle(keysPrefix, kh)
+	mux.Handle(keysPrefix+"/", kh)
+	mux.HandleFunc(statsPrefix+"/store", sh.serveStore)
+	mux.HandleFunc(statsPrefix+"/self", sh.serveSelf)
+	mux.HandleFunc(statsPrefix+"/leader", sh.serveLeader)
+	mux.HandleFunc(varsPath, serveVars)
+	mux.HandleFunc(configPath+"/local/log", logHandleFunc)
+	mux.Handle(metricsPath, prometheus.Handler())
+	mux.Handle(membersPrefix, mh)
+	mux.Handle(membersPrefix+"/", mh)
+	mux.Handle(deprecatedMachinesPrefix, dmh)
+	handleAuth(mux, sech)
+
+	if server.IsPprofEnabled() {
+		plog.Infof("pprof is enabled under %s", pprofPrefix)
+
+		mux.HandleFunc(pprofPrefix, pprof.Index)
+		mux.HandleFunc(pprofPrefix+"/profile", pprof.Profile)
+		mux.HandleFunc(pprofPrefix+"/symbol", pprof.Symbol)
+		mux.HandleFunc(pprofPrefix+"/cmdline", pprof.Cmdline)
+		// TODO: currently, we don't create an entry for pprof.Trace,
+		// because go 1.4 doesn't provide it. After support of go 1.4 is dropped,
+		// we should add the entry.
+
+		mux.Handle(pprofPrefix+"/heap", pprof.Handler("heap"))
+		mux.Handle(pprofPrefix+"/goroutine", pprof.Handler("goroutine"))
+		mux.Handle(pprofPrefix+"/threadcreate", pprof.Handler("threadcreate"))
+		mux.Handle(pprofPrefix+"/block", pprof.Handler("block"))
+	}
+
+	return requestLogger(mux)
+}
+
+type keysHandler struct {
+	sec     auth.Store
+	server  etcdserver.Server
+	cluster api.Cluster
+	timer   etcdserver.RaftTimer
+	timeout time.Duration
+}
+
+func (h *keysHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "HEAD", "GET", "PUT", "POST", "DELETE") {
+		return
+	}
+
+	w.Header().Set("X-Etcd-Cluster-ID", h.cluster.ID().String())
+
+	ctx, cancel := context.WithTimeout(context.Background(), h.timeout)
+	defer cancel()
+	clock := clockwork.NewRealClock()
+	startTime := clock.Now()
+	rr, err := parseKeyRequest(r, clock)
+	if err != nil {
+		writeKeyError(w, err)
+		return
+	}
+	// The path must be valid at this point (we've parsed the request successfully).
+	if !hasKeyPrefixAccess(h.sec, r, r.URL.Path[len(keysPrefix):], rr.Recursive) {
+		writeKeyNoAuth(w)
+		return
+	}
+	if !rr.Wait {
+		reportRequestReceived(rr)
+	}
+	resp, err := h.server.Do(ctx, rr)
+	if err != nil {
+		err = trimErrorPrefix(err, etcdserver.StoreKeysPrefix)
+		writeKeyError(w, err)
+		reportRequestFailed(rr, err)
+		return
+	}
+	switch {
+	case resp.Event != nil:
+		if err := writeKeyEvent(w, resp.Event, h.timer); err != nil {
+			// Should never be reached
+			plog.Errorf("error writing event (%v)", err)
+		}
+		reportRequestCompleted(rr, resp, startTime)
+	case resp.Watcher != nil:
+		ctx, cancel := context.WithTimeout(context.Background(), defaultWatchTimeout)
+		defer cancel()
+		handleKeyWatch(ctx, w, resp.Watcher, rr.Stream, h.timer)
+	default:
+		writeKeyError(w, errors.New("received response with no Event/Watcher!"))
+	}
+}
+
+type deprecatedMachinesHandler struct {
+	cluster api.Cluster
+}
+
+func (h *deprecatedMachinesHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "GET", "HEAD") {
+		return
+	}
+	endpoints := h.cluster.ClientURLs()
+	w.Write([]byte(strings.Join(endpoints, ", ")))
+}
+
+type membersHandler struct {
+	sec     auth.Store
+	server  etcdserver.Server
+	cluster api.Cluster
+	timeout time.Duration
+	clock   clockwork.Clock
+}
+
+func (h *membersHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "GET", "POST", "DELETE", "PUT") {
+		return
+	}
+	if !hasWriteRootAccess(h.sec, r) {
+		writeNoAuth(w, r)
+		return
+	}
+	w.Header().Set("X-Etcd-Cluster-ID", h.cluster.ID().String())
+
+	ctx, cancel := context.WithTimeout(context.Background(), h.timeout)
+	defer cancel()
+
+	switch r.Method {
+	case "GET":
+		switch trimPrefix(r.URL.Path, membersPrefix) {
+		case "":
+			mc := newMemberCollection(h.cluster.Members())
+			w.Header().Set("Content-Type", "application/json")
+			if err := json.NewEncoder(w).Encode(mc); err != nil {
+				plog.Warningf("failed to encode members response (%v)", err)
+			}
+		case "leader":
+			id := h.server.Leader()
+			if id == 0 {
+				writeError(w, r, httptypes.NewHTTPError(http.StatusServiceUnavailable, "During election"))
+				return
+			}
+			m := newMember(h.cluster.Member(id))
+			w.Header().Set("Content-Type", "application/json")
+			if err := json.NewEncoder(w).Encode(m); err != nil {
+				plog.Warningf("failed to encode members response (%v)", err)
+			}
+		default:
+			writeError(w, r, httptypes.NewHTTPError(http.StatusNotFound, "Not found"))
+		}
+	case "POST":
+		req := httptypes.MemberCreateRequest{}
+		if ok := unmarshalRequest(r, &req, w); !ok {
+			return
+		}
+		now := h.clock.Now()
+		m := membership.NewMember("", req.PeerURLs, "", &now)
+		err := h.server.AddMember(ctx, *m)
+		switch {
+		case err == membership.ErrIDExists || err == membership.ErrPeerURLexists:
+			writeError(w, r, httptypes.NewHTTPError(http.StatusConflict, err.Error()))
+			return
+		case err != nil:
+			plog.Errorf("error adding member %s (%v)", m.ID, err)
+			writeError(w, r, err)
+			return
+		}
+		res := newMember(m)
+		w.Header().Set("Content-Type", "application/json")
+		w.WriteHeader(http.StatusCreated)
+		if err := json.NewEncoder(w).Encode(res); err != nil {
+			plog.Warningf("failed to encode members response (%v)", err)
+		}
+	case "DELETE":
+		id, ok := getID(r.URL.Path, w)
+		if !ok {
+			return
+		}
+		err := h.server.RemoveMember(ctx, uint64(id))
+		switch {
+		case err == membership.ErrIDRemoved:
+			writeError(w, r, httptypes.NewHTTPError(http.StatusGone, fmt.Sprintf("Member permanently removed: %s", id)))
+		case err == membership.ErrIDNotFound:
+			writeError(w, r, httptypes.NewHTTPError(http.StatusNotFound, fmt.Sprintf("No such member: %s", id)))
+		case err != nil:
+			plog.Errorf("error removing member %s (%v)", id, err)
+			writeError(w, r, err)
+		default:
+			w.WriteHeader(http.StatusNoContent)
+		}
+	case "PUT":
+		id, ok := getID(r.URL.Path, w)
+		if !ok {
+			return
+		}
+		req := httptypes.MemberUpdateRequest{}
+		if ok := unmarshalRequest(r, &req, w); !ok {
+			return
+		}
+		m := membership.Member{
+			ID:             id,
+			RaftAttributes: membership.RaftAttributes{PeerURLs: req.PeerURLs.StringSlice()},
+		}
+		err := h.server.UpdateMember(ctx, m)
+		switch {
+		case err == membership.ErrPeerURLexists:
+			writeError(w, r, httptypes.NewHTTPError(http.StatusConflict, err.Error()))
+		case err == membership.ErrIDNotFound:
+			writeError(w, r, httptypes.NewHTTPError(http.StatusNotFound, fmt.Sprintf("No such member: %s", id)))
+		case err != nil:
+			plog.Errorf("error updating member %s (%v)", m.ID, err)
+			writeError(w, r, err)
+		default:
+			w.WriteHeader(http.StatusNoContent)
+		}
+	}
+}
+
+type statsHandler struct {
+	stats stats.Stats
+}
+
+func (h *statsHandler) serveStore(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "GET") {
+		return
+	}
+	w.Header().Set("Content-Type", "application/json")
+	w.Write(h.stats.StoreStats())
+}
+
+func (h *statsHandler) serveSelf(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "GET") {
+		return
+	}
+	w.Header().Set("Content-Type", "application/json")
+	w.Write(h.stats.SelfStats())
+}
+
+func (h *statsHandler) serveLeader(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "GET") {
+		return
+	}
+	stats := h.stats.LeaderStats()
+	if stats == nil {
+		writeError(w, r, httptypes.NewHTTPError(http.StatusForbidden, "not current leader"))
+		return
+	}
+	w.Header().Set("Content-Type", "application/json")
+	w.Write(stats)
+}
+
+func serveVars(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "GET") {
+		return
+	}
+
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
+	fmt.Fprintf(w, "{\n")
+	first := true
+	expvar.Do(func(kv expvar.KeyValue) {
+		if !first {
+			fmt.Fprintf(w, ",\n")
+		}
+		first = false
+		fmt.Fprintf(w, "%q: %s", kv.Key, kv.Value)
+	})
+	fmt.Fprintf(w, "\n}\n")
+}
+
+// TODO: change etcdserver to raft interface when we have it.
+//       add test for healthHandler when we have the interface ready.
+func healthHandler(server *etcdserver.EtcdServer) http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		if !allowMethod(w, r.Method, "GET") {
+			return
+		}
+
+		if uint64(server.Leader()) == raft.None {
+			http.Error(w, `{"health": "false"}`, http.StatusServiceUnavailable)
+			return
+		}
+
+		// wait for raft's progress
+		index := server.Index()
+		for i := 0; i < 3; i++ {
+			time.Sleep(250 * time.Millisecond)
+			if server.Index() > index {
+				w.WriteHeader(http.StatusOK)
+				w.Write([]byte(`{"health": "true"}`))
+				return
+			}
+		}
+
+		http.Error(w, `{"health": "false"}`, http.StatusServiceUnavailable)
+		return
+	}
+}
+
+func versionHandler(c api.Cluster, fn func(http.ResponseWriter, *http.Request, string)) http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		v := c.Version()
+		if v != nil {
+			fn(w, r, v.String())
+		} else {
+			fn(w, r, "not_decided")
+		}
+	}
+}
+
+func serveVersion(w http.ResponseWriter, r *http.Request, clusterV string) {
+	if !allowMethod(w, r.Method, "GET") {
+		return
+	}
+	vs := version.Versions{
+		Server:  version.Version,
+		Cluster: clusterV,
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	b, err := json.Marshal(&vs)
+	if err != nil {
+		plog.Panicf("cannot marshal versions to json (%v)", err)
+	}
+	w.Write(b)
+}
+
+func logHandleFunc(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "PUT") {
+		return
+	}
+
+	in := struct{ Level string }{}
+
+	d := json.NewDecoder(r.Body)
+	if err := d.Decode(&in); err != nil {
+		writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, "Invalid json body"))
+		return
+	}
+
+	logl, err := capnslog.ParseLevel(strings.ToUpper(in.Level))
+	if err != nil {
+		writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, "Invalid log level "+in.Level))
+		return
+	}
+
+	plog.Noticef("globalLogLevel set to %q", logl.String())
+	capnslog.SetGlobalLogLevel(logl)
+	w.WriteHeader(http.StatusNoContent)
+}
+
+// parseKeyRequest converts a received http.Request on keysPrefix to
+// a server Request, performing validation of supplied fields as appropriate.
+// If any validation fails, an empty Request and non-nil error is returned.
+func parseKeyRequest(r *http.Request, clock clockwork.Clock) (etcdserverpb.Request, error) {
+	emptyReq := etcdserverpb.Request{}
+
+	err := r.ParseForm()
+	if err != nil {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeInvalidForm,
+			err.Error(),
+		)
+	}
+
+	if !strings.HasPrefix(r.URL.Path, keysPrefix) {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeInvalidForm,
+			"incorrect key prefix",
+		)
+	}
+	p := path.Join(etcdserver.StoreKeysPrefix, r.URL.Path[len(keysPrefix):])
+
+	var pIdx, wIdx uint64
+	if pIdx, err = getUint64(r.Form, "prevIndex"); err != nil {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeIndexNaN,
+			`invalid value for "prevIndex"`,
+		)
+	}
+	if wIdx, err = getUint64(r.Form, "waitIndex"); err != nil {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeIndexNaN,
+			`invalid value for "waitIndex"`,
+		)
+	}
+
+	var rec, sort, wait, dir, quorum, stream bool
+	if rec, err = getBool(r.Form, "recursive"); err != nil {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeInvalidField,
+			`invalid value for "recursive"`,
+		)
+	}
+	if sort, err = getBool(r.Form, "sorted"); err != nil {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeInvalidField,
+			`invalid value for "sorted"`,
+		)
+	}
+	if wait, err = getBool(r.Form, "wait"); err != nil {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeInvalidField,
+			`invalid value for "wait"`,
+		)
+	}
+	// TODO(jonboulle): define what parameters dir is/isn't compatible with?
+	if dir, err = getBool(r.Form, "dir"); err != nil {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeInvalidField,
+			`invalid value for "dir"`,
+		)
+	}
+	if quorum, err = getBool(r.Form, "quorum"); err != nil {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeInvalidField,
+			`invalid value for "quorum"`,
+		)
+	}
+	if stream, err = getBool(r.Form, "stream"); err != nil {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeInvalidField,
+			`invalid value for "stream"`,
+		)
+	}
+
+	if wait && r.Method != "GET" {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodeInvalidField,
+			`"wait" can only be used with GET requests`,
+		)
+	}
+
+	pV := r.FormValue("prevValue")
+	if _, ok := r.Form["prevValue"]; ok && pV == "" {
+		return emptyReq, etcdErr.NewRequestError(
+			etcdErr.EcodePrevValueRequired,
+			`"prevValue" cannot be empty`,
+		)
+	}
+
+	// TTL is nullable, so leave it null if not specified
+	// or an empty string
+	var ttl *uint64
+	if len(r.FormValue("ttl")) > 0 {
+		i, err := getUint64(r.Form, "ttl")
+		if err != nil {
+			return emptyReq, etcdErr.NewRequestError(
+				etcdErr.EcodeTTLNaN,
+				`invalid value for "ttl"`,
+			)
+		}
+		ttl = &i
+	}
+
+	// prevExist is nullable, so leave it null if not specified
+	var pe *bool
+	if _, ok := r.Form["prevExist"]; ok {
+		bv, err := getBool(r.Form, "prevExist")
+		if err != nil {
+			return emptyReq, etcdErr.NewRequestError(
+				etcdErr.EcodeInvalidField,
+				"invalid value for prevExist",
+			)
+		}
+		pe = &bv
+	}
+
+	// refresh is nullable, so leave it null if not specified
+	var refresh *bool
+	if _, ok := r.Form["refresh"]; ok {
+		bv, err := getBool(r.Form, "refresh")
+		if err != nil {
+			return emptyReq, etcdErr.NewRequestError(
+				etcdErr.EcodeInvalidField,
+				"invalid value for refresh",
+			)
+		}
+		refresh = &bv
+		if refresh != nil && *refresh {
+			val := r.FormValue("value")
+			if _, ok := r.Form["value"]; ok && val != "" {
+				return emptyReq, etcdErr.NewRequestError(
+					etcdErr.EcodeRefreshValue,
+					`A value was provided on a refresh`,
+				)
+			}
+			if ttl == nil {
+				return emptyReq, etcdErr.NewRequestError(
+					etcdErr.EcodeRefreshTTLRequired,
+					`No TTL value set`,
+				)
+			}
+		}
+	}
+
+	rr := etcdserverpb.Request{
+		Method:    r.Method,
+		Path:      p,
+		Val:       r.FormValue("value"),
+		Dir:       dir,
+		PrevValue: pV,
+		PrevIndex: pIdx,
+		PrevExist: pe,
+		Wait:      wait,
+		Since:     wIdx,
+		Recursive: rec,
+		Sorted:    sort,
+		Quorum:    quorum,
+		Stream:    stream,
+	}
+
+	if pe != nil {
+		rr.PrevExist = pe
+	}
+
+	if refresh != nil {
+		rr.Refresh = refresh
+	}
+
+	// Null TTL is equivalent to unset Expiration
+	if ttl != nil {
+		expr := time.Duration(*ttl) * time.Second
+		rr.Expiration = clock.Now().Add(expr).UnixNano()
+	}
+
+	return rr, nil
+}
+
+// writeKeyEvent trims the prefix of key path in a single Event under
+// StoreKeysPrefix, serializes it and writes the resulting JSON to the given
+// ResponseWriter, along with the appropriate headers.
+func writeKeyEvent(w http.ResponseWriter, ev *store.Event, rt etcdserver.RaftTimer) error {
+	if ev == nil {
+		return errors.New("cannot write empty Event!")
+	}
+	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("X-Etcd-Index", fmt.Sprint(ev.EtcdIndex))
+	w.Header().Set("X-Raft-Index", fmt.Sprint(rt.Index()))
+	w.Header().Set("X-Raft-Term", fmt.Sprint(rt.Term()))
+
+	if ev.IsCreated() {
+		w.WriteHeader(http.StatusCreated)
+	}
+
+	ev = trimEventPrefix(ev, etcdserver.StoreKeysPrefix)
+	return json.NewEncoder(w).Encode(ev)
+}
+
+func writeKeyNoAuth(w http.ResponseWriter) {
+	e := etcdErr.NewError(etcdErr.EcodeUnauthorized, "Insufficient credentials", 0)
+	e.WriteTo(w)
+}
+
+// writeKeyError logs and writes the given Error to the ResponseWriter.
+// If Error is not an etcdErr, the error will be converted to an etcd error.
+func writeKeyError(w http.ResponseWriter, err error) {
+	if err == nil {
+		return
+	}
+	switch e := err.(type) {
+	case *etcdErr.Error:
+		e.WriteTo(w)
+	default:
+		switch err {
+		case etcdserver.ErrTimeoutDueToLeaderFail, etcdserver.ErrTimeoutDueToConnectionLost:
+			mlog.MergeError(err)
+		default:
+			mlog.MergeErrorf("got unexpected response error (%v)", err)
+		}
+		ee := etcdErr.NewError(etcdErr.EcodeRaftInternal, err.Error(), 0)
+		ee.WriteTo(w)
+	}
+}
+
+func handleKeyWatch(ctx context.Context, w http.ResponseWriter, wa store.Watcher, stream bool, rt etcdserver.RaftTimer) {
+	defer wa.Remove()
+	ech := wa.EventChan()
+	var nch <-chan bool
+	if x, ok := w.(http.CloseNotifier); ok {
+		nch = x.CloseNotify()
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("X-Etcd-Index", fmt.Sprint(wa.StartIndex()))
+	w.Header().Set("X-Raft-Index", fmt.Sprint(rt.Index()))
+	w.Header().Set("X-Raft-Term", fmt.Sprint(rt.Term()))
+	w.WriteHeader(http.StatusOK)
+
+	// Ensure headers are flushed early, in case of long polling
+	w.(http.Flusher).Flush()
+
+	for {
+		select {
+		case <-nch:
+			// Client closed connection. Nothing to do.
+			return
+		case <-ctx.Done():
+			// Timed out. net/http will close the connection for us, so nothing to do.
+			return
+		case ev, ok := <-ech:
+			if !ok {
+				// If the channel is closed this may be an indication of
+				// that notifications are much more than we are able to
+				// send to the client in time. Then we simply end streaming.
+				return
+			}
+			ev = trimEventPrefix(ev, etcdserver.StoreKeysPrefix)
+			if err := json.NewEncoder(w).Encode(ev); err != nil {
+				// Should never be reached
+				plog.Warningf("error writing event (%v)", err)
+				return
+			}
+			if !stream {
+				return
+			}
+			w.(http.Flusher).Flush()
+		}
+	}
+}
+
+func trimEventPrefix(ev *store.Event, prefix string) *store.Event {
+	if ev == nil {
+		return nil
+	}
+	// Since the *Event may reference one in the store history
+	// history, we must copy it before modifying
+	e := ev.Clone()
+	e.Node = trimNodeExternPrefix(e.Node, prefix)
+	e.PrevNode = trimNodeExternPrefix(e.PrevNode, prefix)
+	return e
+}
+
+func trimNodeExternPrefix(n *store.NodeExtern, prefix string) *store.NodeExtern {
+	if n == nil {
+		return nil
+	}
+	n.Key = strings.TrimPrefix(n.Key, prefix)
+	for _, nn := range n.Nodes {
+		nn = trimNodeExternPrefix(nn, prefix)
+	}
+	return n
+}
+
+func trimErrorPrefix(err error, prefix string) error {
+	if e, ok := err.(*etcdErr.Error); ok {
+		e.Cause = strings.TrimPrefix(e.Cause, prefix)
+	}
+	return err
+}
+
+func unmarshalRequest(r *http.Request, req json.Unmarshaler, w http.ResponseWriter) bool {
+	ctype := r.Header.Get("Content-Type")
+	if ctype != "application/json" {
+		writeError(w, r, httptypes.NewHTTPError(http.StatusUnsupportedMediaType, fmt.Sprintf("Bad Content-Type %s, accept application/json", ctype)))
+		return false
+	}
+	b, err := ioutil.ReadAll(r.Body)
+	if err != nil {
+		writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, err.Error()))
+		return false
+	}
+	if err := req.UnmarshalJSON(b); err != nil {
+		writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, err.Error()))
+		return false
+	}
+	return true
+}
+
+func getID(p string, w http.ResponseWriter) (types.ID, bool) {
+	idStr := trimPrefix(p, membersPrefix)
+	if idStr == "" {
+		http.Error(w, "Method Not Allowed", http.StatusMethodNotAllowed)
+		return 0, false
+	}
+	id, err := types.IDFromString(idStr)
+	if err != nil {
+		writeError(w, nil, httptypes.NewHTTPError(http.StatusNotFound, fmt.Sprintf("No such member: %s", idStr)))
+		return 0, false
+	}
+	return id, true
+}
+
+// getUint64 extracts a uint64 by the given key from a Form. If the key does
+// not exist in the form, 0 is returned. If the key exists but the value is
+// badly formed, an error is returned. If multiple values are present only the
+// first is considered.
+func getUint64(form url.Values, key string) (i uint64, err error) {
+	if vals, ok := form[key]; ok {
+		i, err = strconv.ParseUint(vals[0], 10, 64)
+	}
+	return
+}
+
+// getBool extracts a bool by the given key from a Form. If the key does not
+// exist in the form, false is returned. If the key exists but the value is
+// badly formed, an error is returned. If multiple values are present only the
+// first is considered.
+func getBool(form url.Values, key string) (b bool, err error) {
+	if vals, ok := form[key]; ok {
+		b, err = strconv.ParseBool(vals[0])
+	}
+	return
+}
+
+// trimPrefix removes a given prefix and any slash following the prefix
+// e.g.: trimPrefix("foo", "foo") == trimPrefix("foo/", "foo") == ""
+func trimPrefix(p, prefix string) (s string) {
+	s = strings.TrimPrefix(p, prefix)
+	s = strings.TrimPrefix(s, "/")
+	return
+}
+
+func newMemberCollection(ms []*membership.Member) *httptypes.MemberCollection {
+	c := httptypes.MemberCollection(make([]httptypes.Member, len(ms)))
+
+	for i, m := range ms {
+		c[i] = newMember(m)
+	}
+
+	return &c
+}
+
+func newMember(m *membership.Member) httptypes.Member {
+	tm := httptypes.Member{
+		ID:         m.ID.String(),
+		Name:       m.Name,
+		PeerURLs:   make([]string, len(m.PeerURLs)),
+		ClientURLs: make([]string, len(m.ClientURLs)),
+	}
+
+	copy(tm.PeerURLs, m.PeerURLs)
+	copy(tm.ClientURLs, m.ClientURLs)
+
+	return tm
+}

vendor/github.com/coreos/etcd/etcdserver/api/v2http/client_auth.go

@@ -0,0 +1,507 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v2http
+
+import (
+	"encoding/json"
+	"net/http"
+	"path"
+	"strings"
+
+	"github.com/coreos/etcd/etcdserver/api"
+	"github.com/coreos/etcd/etcdserver/api/v2http/httptypes"
+	"github.com/coreos/etcd/etcdserver/auth"
+)
+
+type authHandler struct {
+	sec     auth.Store
+	cluster api.Cluster
+}
+
+func hasWriteRootAccess(sec auth.Store, r *http.Request) bool {
+	if r.Method == "GET" || r.Method == "HEAD" {
+		return true
+	}
+	return hasRootAccess(sec, r)
+}
+
+func hasRootAccess(sec auth.Store, r *http.Request) bool {
+	if sec == nil {
+		// No store means no auth available, eg, tests.
+		return true
+	}
+	if !sec.AuthEnabled() {
+		return true
+	}
+	username, password, ok := r.BasicAuth()
+	if !ok {
+		return false
+	}
+	rootUser, err := sec.GetUser(username)
+	if err != nil {
+		return false
+	}
+
+	ok = sec.CheckPassword(rootUser, password)
+	if !ok {
+		plog.Warningf("auth: wrong password for user %s", username)
+		return false
+	}
+	for _, role := range rootUser.Roles {
+		if role == auth.RootRoleName {
+			return true
+		}
+	}
+	plog.Warningf("auth: user %s does not have the %s role for resource %s.", username, auth.RootRoleName, r.URL.Path)
+	return false
+}
+
+func hasKeyPrefixAccess(sec auth.Store, r *http.Request, key string, recursive bool) bool {
+	if sec == nil {
+		// No store means no auth available, eg, tests.
+		return true
+	}
+	if !sec.AuthEnabled() {
+		return true
+	}
+	if r.Header.Get("Authorization") == "" {
+		plog.Warningf("auth: no authorization provided, checking guest access")
+		return hasGuestAccess(sec, r, key)
+	}
+	username, password, ok := r.BasicAuth()
+	if !ok {
+		plog.Warningf("auth: malformed basic auth encoding")
+		return false
+	}
+	user, err := sec.GetUser(username)
+	if err != nil {
+		plog.Warningf("auth: no such user: %s.", username)
+		return false
+	}
+	authAsUser := sec.CheckPassword(user, password)
+	if !authAsUser {
+		plog.Warningf("auth: incorrect password for user: %s.", username)
+		return false
+	}
+	writeAccess := r.Method != "GET" && r.Method != "HEAD"
+	for _, roleName := range user.Roles {
+		role, err := sec.GetRole(roleName)
+		if err != nil {
+			continue
+		}
+		if recursive {
+			if role.HasRecursiveAccess(key, writeAccess) {
+				return true
+			}
+		} else if role.HasKeyAccess(key, writeAccess) {
+			return true
+		}
+	}
+	plog.Warningf("auth: invalid access for user %s on key %s.", username, key)
+	return false
+}
+
+func hasGuestAccess(sec auth.Store, r *http.Request, key string) bool {
+	writeAccess := r.Method != "GET" && r.Method != "HEAD"
+	role, err := sec.GetRole(auth.GuestRoleName)
+	if err != nil {
+		return false
+	}
+	if role.HasKeyAccess(key, writeAccess) {
+		return true
+	}
+	plog.Warningf("auth: invalid access for unauthenticated user on resource %s.", key)
+	return false
+}
+
+func writeNoAuth(w http.ResponseWriter, r *http.Request) {
+	herr := httptypes.NewHTTPError(http.StatusUnauthorized, "Insufficient credentials")
+	if err := herr.WriteTo(w); err != nil {
+		plog.Debugf("error writing HTTPError (%v) to %s", err, r.RemoteAddr)
+	}
+}
+
+func handleAuth(mux *http.ServeMux, sh *authHandler) {
+	mux.HandleFunc(authPrefix+"/roles", capabilityHandler(authCapability, sh.baseRoles))
+	mux.HandleFunc(authPrefix+"/roles/", capabilityHandler(authCapability, sh.handleRoles))
+	mux.HandleFunc(authPrefix+"/users", capabilityHandler(authCapability, sh.baseUsers))
+	mux.HandleFunc(authPrefix+"/users/", capabilityHandler(authCapability, sh.handleUsers))
+	mux.HandleFunc(authPrefix+"/enable", capabilityHandler(authCapability, sh.enableDisable))
+}
+
+func (sh *authHandler) baseRoles(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "GET") {
+		return
+	}
+	if !hasRootAccess(sh.sec, r) {
+		writeNoAuth(w, r)
+		return
+	}
+
+	w.Header().Set("X-Etcd-Cluster-ID", sh.cluster.ID().String())
+	w.Header().Set("Content-Type", "application/json")
+
+	roles, err := sh.sec.AllRoles()
+	if err != nil {
+		writeError(w, r, err)
+		return
+	}
+	if roles == nil {
+		roles = make([]string, 0)
+	}
+
+	err = r.ParseForm()
+	if err != nil {
+		writeError(w, r, err)
+		return
+	}
+
+	var rolesCollections struct {
+		Roles []auth.Role `json:"roles"`
+	}
+	for _, roleName := range roles {
+		var role auth.Role
+		role, err = sh.sec.GetRole(roleName)
+		if err != nil {
+			writeError(w, r, err)
+			return
+		}
+		rolesCollections.Roles = append(rolesCollections.Roles, role)
+	}
+	err = json.NewEncoder(w).Encode(rolesCollections)
+
+	if err != nil {
+		plog.Warningf("baseRoles error encoding on %s", r.URL)
+		writeError(w, r, err)
+		return
+	}
+}
+
+func (sh *authHandler) handleRoles(w http.ResponseWriter, r *http.Request) {
+	subpath := path.Clean(r.URL.Path[len(authPrefix):])
+	// Split "/roles/rolename/command".
+	// First item is an empty string, second is "roles"
+	pieces := strings.Split(subpath, "/")
+	if len(pieces) == 2 {
+		sh.baseRoles(w, r)
+		return
+	}
+	if len(pieces) != 3 {
+		writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, "Invalid path"))
+		return
+	}
+	sh.forRole(w, r, pieces[2])
+}
+
+func (sh *authHandler) forRole(w http.ResponseWriter, r *http.Request, role string) {
+	if !allowMethod(w, r.Method, "GET", "PUT", "DELETE") {
+		return
+	}
+	if !hasRootAccess(sh.sec, r) {
+		writeNoAuth(w, r)
+		return
+	}
+	w.Header().Set("X-Etcd-Cluster-ID", sh.cluster.ID().String())
+	w.Header().Set("Content-Type", "application/json")
+
+	switch r.Method {
+	case "GET":
+		data, err := sh.sec.GetRole(role)
+		if err != nil {
+			writeError(w, r, err)
+			return
+		}
+		err = json.NewEncoder(w).Encode(data)
+		if err != nil {
+			plog.Warningf("forRole error encoding on %s", r.URL)
+			return
+		}
+		return
+	case "PUT":
+		var in auth.Role
+		err := json.NewDecoder(r.Body).Decode(&in)
+		if err != nil {
+			writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, "Invalid JSON in request body."))
+			return
+		}
+		if in.Role != role {
+			writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, "Role JSON name does not match the name in the URL"))
+			return
+		}
+
+		var out auth.Role
+
+		// create
+		if in.Grant.IsEmpty() && in.Revoke.IsEmpty() {
+			err = sh.sec.CreateRole(in)
+			if err != nil {
+				writeError(w, r, err)
+				return
+			}
+			w.WriteHeader(http.StatusCreated)
+			out = in
+		} else {
+			if !in.Permissions.IsEmpty() {
+				writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, "Role JSON contains both permissions and grant/revoke"))
+				return
+			}
+			out, err = sh.sec.UpdateRole(in)
+			if err != nil {
+				writeError(w, r, err)
+				return
+			}
+			w.WriteHeader(http.StatusOK)
+		}
+
+		err = json.NewEncoder(w).Encode(out)
+		if err != nil {
+			plog.Warningf("forRole error encoding on %s", r.URL)
+			return
+		}
+		return
+	case "DELETE":
+		err := sh.sec.DeleteRole(role)
+		if err != nil {
+			writeError(w, r, err)
+			return
+		}
+	}
+}
+
+type userWithRoles struct {
+	User  string      `json:"user"`
+	Roles []auth.Role `json:"roles,omitempty"`
+}
+
+func (sh *authHandler) baseUsers(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "GET") {
+		return
+	}
+	if !hasRootAccess(sh.sec, r) {
+		writeNoAuth(w, r)
+		return
+	}
+	w.Header().Set("X-Etcd-Cluster-ID", sh.cluster.ID().String())
+	w.Header().Set("Content-Type", "application/json")
+
+	users, err := sh.sec.AllUsers()
+	if err != nil {
+		writeError(w, r, err)
+		return
+	}
+	if users == nil {
+		users = make([]string, 0)
+	}
+
+	err = r.ParseForm()
+	if err != nil {
+		writeError(w, r, err)
+		return
+	}
+
+	var usersCollections struct {
+		Users []userWithRoles `json:"users"`
+	}
+	for _, userName := range users {
+		var user auth.User
+		user, err = sh.sec.GetUser(userName)
+		if err != nil {
+			writeError(w, r, err)
+			return
+		}
+
+		uwr := userWithRoles{User: user.User}
+		for _, roleName := range user.Roles {
+			var role auth.Role
+			role, err = sh.sec.GetRole(roleName)
+			if err != nil {
+				writeError(w, r, err)
+				return
+			}
+			uwr.Roles = append(uwr.Roles, role)
+		}
+
+		usersCollections.Users = append(usersCollections.Users, uwr)
+	}
+	err = json.NewEncoder(w).Encode(usersCollections)
+
+	if err != nil {
+		plog.Warningf("baseUsers error encoding on %s", r.URL)
+		writeError(w, r, err)
+		return
+	}
+}
+
+func (sh *authHandler) handleUsers(w http.ResponseWriter, r *http.Request) {
+	subpath := path.Clean(r.URL.Path[len(authPrefix):])
+	// Split "/users/username".
+	// First item is an empty string, second is "users"
+	pieces := strings.Split(subpath, "/")
+	if len(pieces) == 2 {
+		sh.baseUsers(w, r)
+		return
+	}
+	if len(pieces) != 3 {
+		writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, "Invalid path"))
+		return
+	}
+	sh.forUser(w, r, pieces[2])
+}
+
+func (sh *authHandler) forUser(w http.ResponseWriter, r *http.Request, user string) {
+	if !allowMethod(w, r.Method, "GET", "PUT", "DELETE") {
+		return
+	}
+	if !hasRootAccess(sh.sec, r) {
+		writeNoAuth(w, r)
+		return
+	}
+	w.Header().Set("X-Etcd-Cluster-ID", sh.cluster.ID().String())
+	w.Header().Set("Content-Type", "application/json")
+
+	switch r.Method {
+	case "GET":
+		u, err := sh.sec.GetUser(user)
+		if err != nil {
+			writeError(w, r, err)
+			return
+		}
+
+		err = r.ParseForm()
+		if err != nil {
+			writeError(w, r, err)
+			return
+		}
+
+		uwr := userWithRoles{User: u.User}
+		for _, roleName := range u.Roles {
+			var role auth.Role
+			role, err = sh.sec.GetRole(roleName)
+			if err != nil {
+				writeError(w, r, err)
+				return
+			}
+			uwr.Roles = append(uwr.Roles, role)
+		}
+		err = json.NewEncoder(w).Encode(uwr)
+
+		if err != nil {
+			plog.Warningf("forUser error encoding on %s", r.URL)
+			return
+		}
+		return
+	case "PUT":
+		var u auth.User
+		err := json.NewDecoder(r.Body).Decode(&u)
+		if err != nil {
+			writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, "Invalid JSON in request body."))
+			return
+		}
+		if u.User != user {
+			writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, "User JSON name does not match the name in the URL"))
+			return
+		}
+
+		var (
+			out     auth.User
+			created bool
+		)
+
+		if len(u.Grant) == 0 && len(u.Revoke) == 0 {
+			// create or update
+			if len(u.Roles) != 0 {
+				out, err = sh.sec.CreateUser(u)
+			} else {
+				// if user passes in both password and roles, we are unsure about his/her
+				// intention.
+				out, created, err = sh.sec.CreateOrUpdateUser(u)
+			}
+
+			if err != nil {
+				writeError(w, r, err)
+				return
+			}
+		} else {
+			// update case
+			if len(u.Roles) != 0 {
+				writeError(w, r, httptypes.NewHTTPError(http.StatusBadRequest, "User JSON contains both roles and grant/revoke"))
+				return
+			}
+			out, err = sh.sec.UpdateUser(u)
+			if err != nil {
+				writeError(w, r, err)
+				return
+			}
+		}
+
+		if created {
+			w.WriteHeader(http.StatusCreated)
+		} else {
+			w.WriteHeader(http.StatusOK)
+		}
+
+		out.Password = ""
+
+		err = json.NewEncoder(w).Encode(out)
+		if err != nil {
+			plog.Warningf("forUser error encoding on %s", r.URL)
+			return
+		}
+		return
+	case "DELETE":
+		err := sh.sec.DeleteUser(user)
+		if err != nil {
+			writeError(w, r, err)
+			return
+		}
+	}
+}
+
+type enabled struct {
+	Enabled bool `json:"enabled"`
+}
+
+func (sh *authHandler) enableDisable(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "GET", "PUT", "DELETE") {
+		return
+	}
+	if !hasWriteRootAccess(sh.sec, r) {
+		writeNoAuth(w, r)
+		return
+	}
+	w.Header().Set("X-Etcd-Cluster-ID", sh.cluster.ID().String())
+	w.Header().Set("Content-Type", "application/json")
+	isEnabled := sh.sec.AuthEnabled()
+	switch r.Method {
+	case "GET":
+		jsonDict := enabled{isEnabled}
+		err := json.NewEncoder(w).Encode(jsonDict)
+		if err != nil {
+			plog.Warningf("error encoding auth state on %s", r.URL)
+		}
+	case "PUT":
+		err := sh.sec.EnableAuth()
+		if err != nil {
+			writeError(w, r, err)
+			return
+		}
+	case "DELETE":
+		err := sh.sec.DisableAuth()
+		if err != nil {
+			writeError(w, r, err)
+			return
+		}
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/api/v2http/doc.go

@@ -0,0 +1,16 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package v2http provides etcd client and server implementations.
+package v2http

vendor/github.com/coreos/etcd/etcdserver/api/v2http/http.go

@@ -0,0 +1,94 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v2http
+
+import (
+	"math"
+	"net/http"
+	"strings"
+	"time"
+
+	etcdErr "github.com/coreos/etcd/error"
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/api/v2http/httptypes"
+
+	"github.com/coreos/etcd/etcdserver/auth"
+	"github.com/coreos/etcd/pkg/logutil"
+	"github.com/coreos/pkg/capnslog"
+)
+
+const (
+	// time to wait for a Watch request
+	defaultWatchTimeout = time.Duration(math.MaxInt64)
+)
+
+var (
+	plog = capnslog.NewPackageLogger("github.com/coreos/etcd/etcdserver/api", "v2http")
+	mlog = logutil.NewMergeLogger(plog)
+)
+
+// writeError logs and writes the given Error to the ResponseWriter
+// If Error is an etcdErr, it is rendered to the ResponseWriter
+// Otherwise, it is assumed to be a StatusInternalServerError
+func writeError(w http.ResponseWriter, r *http.Request, err error) {
+	if err == nil {
+		return
+	}
+	switch e := err.(type) {
+	case *etcdErr.Error:
+		e.WriteTo(w)
+	case *httptypes.HTTPError:
+		if et := e.WriteTo(w); et != nil {
+			plog.Debugf("error writing HTTPError (%v) to %s", et, r.RemoteAddr)
+		}
+	case auth.Error:
+		herr := httptypes.NewHTTPError(e.HTTPStatus(), e.Error())
+		if et := herr.WriteTo(w); et != nil {
+			plog.Debugf("error writing HTTPError (%v) to %s", et, r.RemoteAddr)
+		}
+	default:
+		switch err {
+		case etcdserver.ErrTimeoutDueToLeaderFail, etcdserver.ErrTimeoutDueToConnectionLost, etcdserver.ErrNotEnoughStartedMembers:
+			mlog.MergeError(err)
+		default:
+			mlog.MergeErrorf("got unexpected response error (%v)", err)
+		}
+		herr := httptypes.NewHTTPError(http.StatusInternalServerError, "Internal Server Error")
+		if et := herr.WriteTo(w); et != nil {
+			plog.Debugf("error writing HTTPError (%v) to %s", et, r.RemoteAddr)
+		}
+	}
+}
+
+// allowMethod verifies that the given method is one of the allowed methods,
+// and if not, it writes an error to w.  A boolean is returned indicating
+// whether or not the method is allowed.
+func allowMethod(w http.ResponseWriter, m string, ms ...string) bool {
+	for _, meth := range ms {
+		if m == meth {
+			return true
+		}
+	}
+	w.Header().Set("Allow", strings.Join(ms, ","))
+	http.Error(w, "Method Not Allowed", http.StatusMethodNotAllowed)
+	return false
+}
+
+func requestLogger(handler http.Handler) http.Handler {
+	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		plog.Debugf("[%s] %s remote:%s", r.Method, r.RequestURI, r.RemoteAddr)
+		handler.ServeHTTP(w, r)
+	})
+}

vendor/github.com/coreos/etcd/etcdserver/api/v2http/httptypes/errors.go

@@ -0,0 +1,56 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package httptypes
+
+import (
+	"encoding/json"
+	"net/http"
+
+	"github.com/coreos/pkg/capnslog"
+)
+
+var (
+	plog = capnslog.NewPackageLogger("github.com/coreos/etcd/etcdserver/api/v2http", "httptypes")
+)
+
+type HTTPError struct {
+	Message string `json:"message"`
+	// Code is the HTTP status code
+	Code int `json:"-"`
+}
+
+func (e HTTPError) Error() string {
+	return e.Message
+}
+
+func (e HTTPError) WriteTo(w http.ResponseWriter) error {
+	w.Header().Set("Content-Type", "application/json")
+	w.WriteHeader(e.Code)
+	b, err := json.Marshal(e)
+	if err != nil {
+		plog.Panicf("marshal HTTPError should never fail (%v)", err)
+	}
+	if _, err := w.Write(b); err != nil {
+		return err
+	}
+	return nil
+}
+
+func NewHTTPError(code int, m string) *HTTPError {
+	return &HTTPError{
+		Message: m,
+		Code:    code,
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/api/v2http/httptypes/member.go

@@ -0,0 +1,69 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package httptypes defines how etcd's HTTP API entities are serialized to and
+// deserialized from JSON.
+package httptypes
+
+import (
+	"encoding/json"
+
+	"github.com/coreos/etcd/pkg/types"
+)
+
+type Member struct {
+	ID         string   `json:"id"`
+	Name       string   `json:"name"`
+	PeerURLs   []string `json:"peerURLs"`
+	ClientURLs []string `json:"clientURLs"`
+}
+
+type MemberCreateRequest struct {
+	PeerURLs types.URLs
+}
+
+type MemberUpdateRequest struct {
+	MemberCreateRequest
+}
+
+func (m *MemberCreateRequest) UnmarshalJSON(data []byte) error {
+	s := struct {
+		PeerURLs []string `json:"peerURLs"`
+	}{}
+
+	err := json.Unmarshal(data, &s)
+	if err != nil {
+		return err
+	}
+
+	urls, err := types.NewURLs(s.PeerURLs)
+	if err != nil {
+		return err
+	}
+
+	m.PeerURLs = urls
+	return nil
+}
+
+type MemberCollection []Member
+
+func (c *MemberCollection) MarshalJSON() ([]byte, error) {
+	d := struct {
+		Members []Member `json:"members"`
+	}{
+		Members: []Member(*c),
+	}
+
+	return json.Marshal(d)
+}

vendor/github.com/coreos/etcd/etcdserver/api/v2http/metrics.go

@@ -0,0 +1,96 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v2http
+
+import (
+	"strconv"
+	"time"
+
+	"net/http"
+
+	etcdErr "github.com/coreos/etcd/error"
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/api/v2http/httptypes"
+	"github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/prometheus/client_golang/prometheus"
+)
+
+var (
+	incomingEvents = prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Namespace: "etcd",
+			Subsystem: "http",
+			Name:      "received_total",
+			Help:      "Counter of requests received into the system (successfully parsed and authd).",
+		}, []string{"method"})
+
+	failedEvents = prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Namespace: "etcd",
+			Subsystem: "http",
+			Name:      "failed_total",
+			Help:      "Counter of handle failures of requests (non-watches), by method (GET/PUT etc.) and code (400, 500 etc.).",
+		}, []string{"method", "code"})
+
+	successfulEventsHandlingTime = prometheus.NewHistogramVec(
+		prometheus.HistogramOpts{
+			Namespace: "etcd",
+			Subsystem: "http",
+			Name:      "successful_duration_second",
+			Help:      "Bucketed histogram of processing time (s) of successfully handled requests (non-watches), by method (GET/PUT etc.).",
+			Buckets:   prometheus.ExponentialBuckets(0.0005, 2, 13),
+		}, []string{"method"})
+)
+
+func init() {
+	prometheus.MustRegister(incomingEvents)
+	prometheus.MustRegister(failedEvents)
+	prometheus.MustRegister(successfulEventsHandlingTime)
+}
+
+func reportRequestReceived(request etcdserverpb.Request) {
+	incomingEvents.WithLabelValues(methodFromRequest(request)).Inc()
+}
+
+func reportRequestCompleted(request etcdserverpb.Request, response etcdserver.Response, startTime time.Time) {
+	method := methodFromRequest(request)
+	successfulEventsHandlingTime.WithLabelValues(method).Observe(time.Since(startTime).Seconds())
+}
+
+func reportRequestFailed(request etcdserverpb.Request, err error) {
+	method := methodFromRequest(request)
+	failedEvents.WithLabelValues(method, strconv.Itoa(codeFromError(err))).Inc()
+}
+
+func methodFromRequest(request etcdserverpb.Request) string {
+	if request.Method == "GET" && request.Quorum {
+		return "QGET"
+	}
+	return request.Method
+}
+
+func codeFromError(err error) int {
+	if err == nil {
+		return http.StatusInternalServerError
+	}
+	switch e := err.(type) {
+	case *etcdErr.Error:
+		return (*etcdErr.Error)(e).StatusCode()
+	case *httptypes.HTTPError:
+		return (*httptypes.HTTPError)(e).Code
+	default:
+		return http.StatusInternalServerError
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/api/v2http/peer.go

@@ -0,0 +1,77 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v2http
+
+import (
+	"encoding/json"
+	"net/http"
+
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/api"
+	"github.com/coreos/etcd/lease/leasehttp"
+	"github.com/coreos/etcd/rafthttp"
+)
+
+const (
+	peerMembersPrefix = "/members"
+	leasesPrefix      = "/leases"
+)
+
+// NewPeerHandler generates an http.Handler to handle etcd peer requests.
+func NewPeerHandler(s *etcdserver.EtcdServer) http.Handler {
+	var lh http.Handler
+	if l := s.Lessor(); l != nil {
+		lh = leasehttp.NewHandler(l)
+	}
+	return newPeerHandler(s.Cluster(), s.RaftHandler(), lh)
+}
+
+func newPeerHandler(cluster api.Cluster, raftHandler http.Handler, leaseHandler http.Handler) http.Handler {
+	mh := &peerMembersHandler{
+		cluster: cluster,
+	}
+
+	mux := http.NewServeMux()
+	mux.HandleFunc("/", http.NotFound)
+	mux.Handle(rafthttp.RaftPrefix, raftHandler)
+	mux.Handle(rafthttp.RaftPrefix+"/", raftHandler)
+	mux.Handle(peerMembersPrefix, mh)
+	if leaseHandler != nil {
+		mux.Handle(leasesPrefix, leaseHandler)
+	}
+	mux.HandleFunc(versionPath, versionHandler(cluster, serveVersion))
+	return mux
+}
+
+type peerMembersHandler struct {
+	cluster api.Cluster
+}
+
+func (h *peerMembersHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
+	if !allowMethod(w, r.Method, "GET") {
+		return
+	}
+	w.Header().Set("X-Etcd-Cluster-ID", h.cluster.ID().String())
+
+	if r.URL.Path != peerMembersPrefix {
+		http.Error(w, "bad path", http.StatusBadRequest)
+		return
+	}
+	ms := h.cluster.Members()
+	w.Header().Set("Content-Type", "application/json")
+	if err := json.NewEncoder(w).Encode(ms); err != nil {
+		plog.Warningf("failed to encode members response (%v)", err)
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/auth.go

@@ -0,0 +1,114 @@
+// Copyright 2016 Nippon Telegraph and Telephone Corporation.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v3rpc
+
+import (
+	"github.com/coreos/etcd/etcdserver"
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"golang.org/x/net/context"
+)
+
+type AuthServer struct {
+	authenticator etcdserver.Authenticator
+}
+
+func NewAuthServer(s *etcdserver.EtcdServer) *AuthServer {
+	return &AuthServer{authenticator: s}
+}
+
+func (as *AuthServer) AuthEnable(ctx context.Context, r *pb.AuthEnableRequest) (*pb.AuthEnableResponse, error) {
+	resp, err := as.authenticator.AuthEnable(ctx, r)
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+	return resp, nil
+}
+
+func (as *AuthServer) AuthDisable(ctx context.Context, r *pb.AuthDisableRequest) (*pb.AuthDisableResponse, error) {
+	plog.Info("not implemented yet")
+	return nil, nil
+}
+
+func (as *AuthServer) Authenticate(ctx context.Context, r *pb.AuthenticateRequest) (*pb.AuthenticateResponse, error) {
+	plog.Info("not implemented yet")
+	return nil, nil
+}
+
+func (as *AuthServer) RoleAdd(ctx context.Context, r *pb.AuthRoleAddRequest) (*pb.AuthRoleAddResponse, error) {
+	resp, err := as.authenticator.RoleAdd(ctx, r)
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+	return resp, nil
+}
+
+func (as *AuthServer) RoleDelete(ctx context.Context, r *pb.AuthRoleDeleteRequest) (*pb.AuthRoleDeleteResponse, error) {
+	plog.Info("not implemented yet")
+	return nil, nil
+}
+
+func (as *AuthServer) RoleGet(ctx context.Context, r *pb.AuthRoleGetRequest) (*pb.AuthRoleGetResponse, error) {
+	plog.Info("not implemented yet")
+	return nil, nil
+}
+
+func (as *AuthServer) RoleRevoke(ctx context.Context, r *pb.AuthRoleRevokeRequest) (*pb.AuthRoleRevokeResponse, error) {
+	plog.Info("not implemented yet")
+	return nil, nil
+}
+
+func (as *AuthServer) RoleGrant(ctx context.Context, r *pb.AuthRoleGrantRequest) (*pb.AuthRoleGrantResponse, error) {
+	plog.Info("not implemented yet")
+	return nil, nil
+}
+
+func (as *AuthServer) UserAdd(ctx context.Context, r *pb.AuthUserAddRequest) (*pb.AuthUserAddResponse, error) {
+	resp, err := as.authenticator.UserAdd(ctx, r)
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+	return resp, nil
+}
+
+func (as *AuthServer) UserDelete(ctx context.Context, r *pb.AuthUserDeleteRequest) (*pb.AuthUserDeleteResponse, error) {
+	resp, err := as.authenticator.UserDelete(ctx, r)
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+	return resp, nil
+}
+
+func (as *AuthServer) UserGet(ctx context.Context, r *pb.AuthUserGetRequest) (*pb.AuthUserGetResponse, error) {
+	plog.Info("not implemented yet")
+	return nil, nil
+}
+
+func (as *AuthServer) UserGrant(ctx context.Context, r *pb.AuthUserGrantRequest) (*pb.AuthUserGrantResponse, error) {
+	plog.Info("not implemented yet")
+	return nil, nil
+}
+
+func (as *AuthServer) UserRevoke(ctx context.Context, r *pb.AuthUserRevokeRequest) (*pb.AuthUserRevokeResponse, error) {
+	plog.Info("not implemented yet")
+	return nil, nil
+}
+
+func (as *AuthServer) UserChangePassword(ctx context.Context, r *pb.AuthUserChangePasswordRequest) (*pb.AuthUserChangePasswordResponse, error) {
+	resp, err := as.authenticator.UserChangePassword(ctx, r)
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+	return resp, nil
+}

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/grpc.go

@@ -0,0 +1,40 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v3rpc
+
+import (
+	"crypto/tls"
+
+	"github.com/coreos/etcd/etcdserver"
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/credentials"
+)
+
+func Server(s *etcdserver.EtcdServer, tls *tls.Config) *grpc.Server {
+	var opts []grpc.ServerOption
+	if tls != nil {
+		opts = append(opts, grpc.Creds(credentials.NewTLS(tls)))
+	}
+
+	grpcServer := grpc.NewServer(opts...)
+	pb.RegisterKVServer(grpcServer, NewQuotaKVServer(s))
+	pb.RegisterWatchServer(grpcServer, NewWatchServer(s))
+	pb.RegisterLeaseServer(grpcServer, NewQuotaLeaseServer(s))
+	pb.RegisterClusterServer(grpcServer, NewClusterServer(s))
+	pb.RegisterAuthServer(grpcServer, NewAuthServer(s))
+	pb.RegisterMaintenanceServer(grpcServer, NewMaintenanceServer(s))
+	return grpcServer
+}

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/header.go

@@ -0,0 +1,43 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v3rpc
+
+import (
+	"github.com/coreos/etcd/etcdserver"
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+)
+
+type header struct {
+	clusterID int64
+	memberID  int64
+	raftTimer etcdserver.RaftTimer
+	rev       func() int64
+}
+
+func newHeader(s *etcdserver.EtcdServer) header {
+	return header{
+		clusterID: int64(s.Cluster().ID()),
+		memberID:  int64(s.ID()),
+		raftTimer: s,
+		rev:       func() int64 { return s.KV().Rev() },
+	}
+}
+
+// fill populates pb.ResponseHeader using etcdserver information
+func (h *header) fill(rh *pb.ResponseHeader) {
+	rh.ClusterId = uint64(h.clusterID)
+	rh.MemberId = uint64(h.memberID)
+	rh.RaftTerm = h.raftTimer.Term()
+}

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/key.go

@@ -0,0 +1,259 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package v3rpc implements etcd v3 RPC system based on gRPC.
+package v3rpc
+
+import (
+	"sort"
+
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/api/v3rpc/rpctypes"
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/pkg/capnslog"
+	"golang.org/x/net/context"
+)
+
+var (
+	plog = capnslog.NewPackageLogger("github.com/coreos/etcd/etcdserver/api", "v3rpc")
+
+	// Max operations per txn list. For example, Txn.Success can have at most 128 operations,
+	// and Txn.Failure can have at most 128 operations.
+	MaxOpsPerTxn = 128
+)
+
+type kvServer struct {
+	hdr header
+	kv  etcdserver.RaftKV
+}
+
+func NewKVServer(s *etcdserver.EtcdServer) pb.KVServer {
+	return &kvServer{hdr: newHeader(s), kv: s}
+}
+
+func (s *kvServer) Range(ctx context.Context, r *pb.RangeRequest) (*pb.RangeResponse, error) {
+	if err := checkRangeRequest(r); err != nil {
+		return nil, err
+	}
+
+	resp, err := s.kv.Range(ctx, r)
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+
+	if resp.Header == nil {
+		plog.Panic("unexpected nil resp.Header")
+	}
+	s.hdr.fill(resp.Header)
+	return resp, err
+}
+
+func (s *kvServer) Put(ctx context.Context, r *pb.PutRequest) (*pb.PutResponse, error) {
+	if err := checkPutRequest(r); err != nil {
+		return nil, err
+	}
+
+	resp, err := s.kv.Put(ctx, r)
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+
+	if resp.Header == nil {
+		plog.Panic("unexpected nil resp.Header")
+	}
+	s.hdr.fill(resp.Header)
+	return resp, err
+}
+
+func (s *kvServer) DeleteRange(ctx context.Context, r *pb.DeleteRangeRequest) (*pb.DeleteRangeResponse, error) {
+	if err := checkDeleteRequest(r); err != nil {
+		return nil, err
+	}
+
+	resp, err := s.kv.DeleteRange(ctx, r)
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+
+	if resp.Header == nil {
+		plog.Panic("unexpected nil resp.Header")
+	}
+	s.hdr.fill(resp.Header)
+	return resp, err
+}
+
+func (s *kvServer) Txn(ctx context.Context, r *pb.TxnRequest) (*pb.TxnResponse, error) {
+	if err := checkTxnRequest(r); err != nil {
+		return nil, err
+	}
+
+	resp, err := s.kv.Txn(ctx, r)
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+
+	if resp.Header == nil {
+		plog.Panic("unexpected nil resp.Header")
+	}
+	s.hdr.fill(resp.Header)
+	return resp, err
+}
+
+func (s *kvServer) Compact(ctx context.Context, r *pb.CompactionRequest) (*pb.CompactionResponse, error) {
+	resp, err := s.kv.Compact(ctx, r)
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+
+	if resp.Header == nil {
+		plog.Panic("unexpected nil resp.Header")
+	}
+	s.hdr.fill(resp.Header)
+	return resp, nil
+}
+
+func checkRangeRequest(r *pb.RangeRequest) error {
+	if len(r.Key) == 0 {
+		return rpctypes.ErrEmptyKey
+	}
+	return nil
+}
+
+func checkPutRequest(r *pb.PutRequest) error {
+	if len(r.Key) == 0 {
+		return rpctypes.ErrEmptyKey
+	}
+	return nil
+}
+
+func checkDeleteRequest(r *pb.DeleteRangeRequest) error {
+	if len(r.Key) == 0 {
+		return rpctypes.ErrEmptyKey
+	}
+	return nil
+}
+
+func checkTxnRequest(r *pb.TxnRequest) error {
+	if len(r.Compare) > MaxOpsPerTxn || len(r.Success) > MaxOpsPerTxn || len(r.Failure) > MaxOpsPerTxn {
+		return rpctypes.ErrTooManyOps
+	}
+
+	for _, c := range r.Compare {
+		if len(c.Key) == 0 {
+			return rpctypes.ErrEmptyKey
+		}
+	}
+
+	for _, u := range r.Success {
+		if err := checkRequestUnion(u); err != nil {
+			return err
+		}
+	}
+	if err := checkRequestDupKeys(r.Success); err != nil {
+		return err
+	}
+
+	for _, u := range r.Failure {
+		if err := checkRequestUnion(u); err != nil {
+			return err
+		}
+	}
+	if err := checkRequestDupKeys(r.Failure); err != nil {
+		return err
+	}
+
+	return nil
+}
+
+// checkRequestDupKeys gives rpctypes.ErrDuplicateKey if the same key is modified twice
+func checkRequestDupKeys(reqs []*pb.RequestUnion) error {
+	// check put overlap
+	keys := make(map[string]struct{})
+	for _, requ := range reqs {
+		tv, ok := requ.Request.(*pb.RequestUnion_RequestPut)
+		if !ok {
+			continue
+		}
+		preq := tv.RequestPut
+		if preq == nil {
+			continue
+		}
+		key := string(preq.Key)
+		if _, ok := keys[key]; ok {
+			return rpctypes.ErrDuplicateKey
+		}
+		keys[key] = struct{}{}
+	}
+
+	// no need to check deletes if no puts; delete overlaps are permitted
+	if len(keys) == 0 {
+		return nil
+	}
+
+	// sort keys for range checking
+	sortedKeys := []string{}
+	for k := range keys {
+		sortedKeys = append(sortedKeys, k)
+	}
+	sort.Strings(sortedKeys)
+
+	// check put overlap with deletes
+	for _, requ := range reqs {
+		tv, ok := requ.Request.(*pb.RequestUnion_RequestDeleteRange)
+		if !ok {
+			continue
+		}
+		dreq := tv.RequestDeleteRange
+		if dreq == nil {
+			continue
+		}
+		key := string(dreq.Key)
+		if dreq.RangeEnd == nil {
+			if _, found := keys[key]; found {
+				return rpctypes.ErrDuplicateKey
+			}
+		} else {
+			lo := sort.SearchStrings(sortedKeys, key)
+			hi := sort.SearchStrings(sortedKeys, string(dreq.RangeEnd))
+			if lo != hi {
+				// element between lo and hi => overlap
+				return rpctypes.ErrDuplicateKey
+			}
+		}
+	}
+
+	return nil
+}
+
+func checkRequestUnion(u *pb.RequestUnion) error {
+	// TODO: ensure only one of the field is set.
+	switch uv := u.Request.(type) {
+	case *pb.RequestUnion_RequestRange:
+		if uv.RequestRange != nil {
+			return checkRangeRequest(uv.RequestRange)
+		}
+	case *pb.RequestUnion_RequestPut:
+		if uv.RequestPut != nil {
+			return checkPutRequest(uv.RequestPut)
+		}
+	case *pb.RequestUnion_RequestDeleteRange:
+		if uv.RequestDeleteRange != nil {
+			return checkDeleteRequest(uv.RequestDeleteRange)
+		}
+	default:
+		// empty union
+		return nil
+	}
+	return nil
+}

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/lease.go

@@ -0,0 +1,76 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v3rpc
+
+import (
+	"io"
+
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/api/v3rpc/rpctypes"
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/lease"
+	"golang.org/x/net/context"
+)
+
+type LeaseServer struct {
+	le etcdserver.Lessor
+}
+
+func NewLeaseServer(le etcdserver.Lessor) pb.LeaseServer {
+	return &LeaseServer{le: le}
+}
+
+func (ls *LeaseServer) LeaseGrant(ctx context.Context, cr *pb.LeaseGrantRequest) (*pb.LeaseGrantResponse, error) {
+	resp, err := ls.le.LeaseGrant(ctx, cr)
+	if err == lease.ErrLeaseExists {
+		return nil, rpctypes.ErrLeaseExist
+	}
+	return resp, err
+}
+
+func (ls *LeaseServer) LeaseRevoke(ctx context.Context, rr *pb.LeaseRevokeRequest) (*pb.LeaseRevokeResponse, error) {
+	r, err := ls.le.LeaseRevoke(ctx, rr)
+	if err != nil {
+		return nil, rpctypes.ErrLeaseNotFound
+	}
+	return r, nil
+}
+
+func (ls *LeaseServer) LeaseKeepAlive(stream pb.Lease_LeaseKeepAliveServer) error {
+	for {
+		req, err := stream.Recv()
+		if err == io.EOF {
+			return nil
+		}
+		if err != nil {
+			return err
+		}
+
+		ttl, err := ls.le.LeaseRenew(lease.LeaseID(req.ID))
+		if err == lease.ErrLeaseNotFound {
+			return rpctypes.ErrLeaseNotFound
+		}
+
+		if err != nil && err != lease.ErrLeaseNotFound {
+			return err
+		}
+
+		resp := &pb.LeaseKeepAliveResponse{ID: req.ID, TTL: ttl}
+		err = stream.Send(resp)
+		if err != nil {
+			return err
+		}
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/maintenance.go

@@ -0,0 +1,72 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v3rpc
+
+import (
+	"github.com/coreos/etcd/etcdserver"
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/storage/backend"
+	"github.com/coreos/etcd/version"
+	"golang.org/x/net/context"
+)
+
+type BackendGetter interface {
+	Backend() backend.Backend
+}
+
+type Alarmer interface {
+	Alarm(ctx context.Context, ar *pb.AlarmRequest) (*pb.AlarmResponse, error)
+}
+
+type maintenanceServer struct {
+	bg  BackendGetter
+	a   Alarmer
+	hdr header
+}
+
+func NewMaintenanceServer(s *etcdserver.EtcdServer) pb.MaintenanceServer {
+	return &maintenanceServer{bg: s, a: s, hdr: newHeader(s)}
+}
+
+func (ms *maintenanceServer) Defragment(ctx context.Context, sr *pb.DefragmentRequest) (*pb.DefragmentResponse, error) {
+	plog.Noticef("starting to defragment the storage backend...")
+	err := ms.bg.Backend().Defrag()
+	if err != nil {
+		plog.Errorf("failed to deframent the storage backend (%v)", err)
+		return nil, err
+	}
+	plog.Noticef("finished defragmenting the storage backend")
+	return &pb.DefragmentResponse{}, nil
+}
+
+func (ms *maintenanceServer) Hash(ctx context.Context, r *pb.HashRequest) (*pb.HashResponse, error) {
+	h, err := ms.bg.Backend().Hash()
+	if err != nil {
+		return nil, togRPCError(err)
+	}
+	resp := &pb.HashResponse{Header: &pb.ResponseHeader{Revision: ms.hdr.rev()}, Hash: h}
+	ms.hdr.fill(resp.Header)
+	return resp, nil
+}
+
+func (ms *maintenanceServer) Alarm(ctx context.Context, ar *pb.AlarmRequest) (*pb.AlarmResponse, error) {
+	return ms.a.Alarm(ctx, ar)
+}
+
+func (ms *maintenanceServer) Status(ctx context.Context, ar *pb.StatusRequest) (*pb.StatusResponse, error) {
+	resp := &pb.StatusResponse{Header: &pb.ResponseHeader{Revision: ms.hdr.rev()}, Version: version.Version}
+	ms.hdr.fill(resp.Header)
+	return resp, nil
+}

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/member.go

@@ -0,0 +1,120 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v3rpc
+
+import (
+	"time"
+
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/api"
+	"github.com/coreos/etcd/etcdserver/api/v3rpc/rpctypes"
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/etcdserver/membership"
+	"github.com/coreos/etcd/pkg/types"
+	"golang.org/x/net/context"
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/codes"
+)
+
+type ClusterServer struct {
+	cluster   api.Cluster
+	server    etcdserver.Server
+	raftTimer etcdserver.RaftTimer
+}
+
+func NewClusterServer(s *etcdserver.EtcdServer) *ClusterServer {
+	return &ClusterServer{
+		cluster:   s.Cluster(),
+		server:    s,
+		raftTimer: s,
+	}
+}
+
+func (cs *ClusterServer) MemberAdd(ctx context.Context, r *pb.MemberAddRequest) (*pb.MemberAddResponse, error) {
+	urls, err := types.NewURLs(r.PeerURLs)
+	if err != nil {
+		return nil, rpctypes.ErrMemberBadURLs
+	}
+
+	now := time.Now()
+	m := membership.NewMember("", urls, "", &now)
+	err = cs.server.AddMember(ctx, *m)
+	switch {
+	case err == membership.ErrIDExists:
+		return nil, rpctypes.ErrMemberExist
+	case err == membership.ErrPeerURLexists:
+		return nil, rpctypes.ErrPeerURLExist
+	case err != nil:
+		return nil, grpc.Errorf(codes.Internal, err.Error())
+	}
+
+	return &pb.MemberAddResponse{
+		Header: cs.header(),
+		Member: &pb.Member{ID: uint64(m.ID), IsLeader: m.ID == cs.server.Leader(), PeerURLs: m.PeerURLs},
+	}, nil
+}
+
+func (cs *ClusterServer) MemberRemove(ctx context.Context, r *pb.MemberRemoveRequest) (*pb.MemberRemoveResponse, error) {
+	err := cs.server.RemoveMember(ctx, r.ID)
+	switch {
+	case err == membership.ErrIDRemoved:
+		fallthrough
+	case err == membership.ErrIDNotFound:
+		return nil, rpctypes.ErrMemberNotFound
+	case err != nil:
+		return nil, grpc.Errorf(codes.Internal, err.Error())
+	}
+
+	return &pb.MemberRemoveResponse{Header: cs.header()}, nil
+}
+
+func (cs *ClusterServer) MemberUpdate(ctx context.Context, r *pb.MemberUpdateRequest) (*pb.MemberUpdateResponse, error) {
+	m := membership.Member{
+		ID:             types.ID(r.ID),
+		RaftAttributes: membership.RaftAttributes{PeerURLs: r.PeerURLs},
+	}
+	err := cs.server.UpdateMember(ctx, m)
+	switch {
+	case err == membership.ErrPeerURLexists:
+		return nil, rpctypes.ErrPeerURLExist
+	case err == membership.ErrIDNotFound:
+		return nil, rpctypes.ErrMemberNotFound
+	case err != nil:
+		return nil, grpc.Errorf(codes.Internal, err.Error())
+	}
+
+	return &pb.MemberUpdateResponse{Header: cs.header()}, nil
+}
+
+func (cs *ClusterServer) MemberList(ctx context.Context, r *pb.MemberListRequest) (*pb.MemberListResponse, error) {
+	membs := cs.cluster.Members()
+
+	protoMembs := make([]*pb.Member, len(membs))
+	for i := range membs {
+		protoMembs[i] = &pb.Member{
+			Name:       membs[i].Name,
+			ID:         uint64(membs[i].ID),
+			IsLeader:   membs[i].ID == cs.server.Leader(),
+			PeerURLs:   membs[i].PeerURLs,
+			ClientURLs: membs[i].ClientURLs,
+		}
+	}
+
+	return &pb.MemberListResponse{Header: cs.header(), Members: protoMembs}, nil
+}
+
+func (cs *ClusterServer) header() *pb.ResponseHeader {
+	return &pb.ResponseHeader{ClusterId: uint64(cs.cluster.ID()), MemberId: uint64(cs.server.ID()), RaftTerm: cs.raftTimer.Term()}
+}

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/quota.go

@@ -0,0 +1,89 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v3rpc
+
+import (
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/api/v3rpc/rpctypes"
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/pkg/types"
+	"golang.org/x/net/context"
+)
+
+type quotaKVServer struct {
+	pb.KVServer
+	qa quotaAlarmer
+}
+
+type quotaAlarmer struct {
+	q  etcdserver.Quota
+	a  Alarmer
+	id types.ID
+}
+
+// check whether request satisfies the quota. If there is not enough space,
+// ignore request and raise the free space alarm.
+func (qa *quotaAlarmer) check(ctx context.Context, r interface{}) error {
+	if qa.q.Available(r) {
+		return nil
+	}
+	req := &pb.AlarmRequest{
+		MemberID: uint64(qa.id),
+		Action:   pb.AlarmRequest_ACTIVATE,
+		Alarm:    pb.AlarmType_NOSPACE,
+	}
+	qa.a.Alarm(ctx, req)
+	return rpctypes.ErrNoSpace
+}
+
+func NewQuotaKVServer(s *etcdserver.EtcdServer) pb.KVServer {
+	return &quotaKVServer{
+		NewKVServer(s),
+		quotaAlarmer{etcdserver.NewBackendQuota(s), s, s.ID()},
+	}
+}
+
+func (s *quotaKVServer) Put(ctx context.Context, r *pb.PutRequest) (*pb.PutResponse, error) {
+	if err := s.qa.check(ctx, r); err != nil {
+		return nil, err
+	}
+	return s.KVServer.Put(ctx, r)
+}
+
+func (s *quotaKVServer) Txn(ctx context.Context, r *pb.TxnRequest) (*pb.TxnResponse, error) {
+	if err := s.qa.check(ctx, r); err != nil {
+		return nil, err
+	}
+	return s.KVServer.Txn(ctx, r)
+}
+
+type quotaLeaseServer struct {
+	pb.LeaseServer
+	qa quotaAlarmer
+}
+
+func (s *quotaLeaseServer) LeaseGrant(ctx context.Context, cr *pb.LeaseGrantRequest) (*pb.LeaseGrantResponse, error) {
+	if err := s.qa.check(ctx, cr); err != nil {
+		return nil, err
+	}
+	return s.LeaseServer.LeaseGrant(ctx, cr)
+}
+
+func NewQuotaLeaseServer(s *etcdserver.EtcdServer) pb.LeaseServer {
+	return &quotaLeaseServer{
+		NewLeaseServer(s),
+		quotaAlarmer{etcdserver.NewBackendQuota(s), s, s.ID()},
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/rpctypes/error.go

@@ -0,0 +1,43 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package rpctypes
+
+import (
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/codes"
+)
+
+var (
+	ErrEmptyKey     = grpc.Errorf(codes.InvalidArgument, "etcdserver: key is not provided")
+	ErrTooManyOps   = grpc.Errorf(codes.InvalidArgument, "etcdserver: too many operations in txn request")
+	ErrDuplicateKey = grpc.Errorf(codes.InvalidArgument, "etcdserver: duplicate key given in txn request")
+	ErrCompacted    = grpc.Errorf(codes.OutOfRange, "etcdserver: storage: required revision has been compacted")
+	ErrFutureRev    = grpc.Errorf(codes.OutOfRange, "etcdserver: storage: required revision is a future revision")
+	ErrNoSpace      = grpc.Errorf(codes.ResourceExhausted, "etcdserver: storage: database space exceeded")
+
+	ErrLeaseNotFound = grpc.Errorf(codes.NotFound, "etcdserver: requested lease not found")
+	ErrLeaseExist    = grpc.Errorf(codes.FailedPrecondition, "etcdserver: lease already exists")
+
+	ErrMemberExist    = grpc.Errorf(codes.FailedPrecondition, "etcdserver: member ID already exist")
+	ErrPeerURLExist   = grpc.Errorf(codes.FailedPrecondition, "etcdserver: Peer URLs already exists")
+	ErrMemberBadURLs  = grpc.Errorf(codes.InvalidArgument, "etcdserver: given member URLs are invalid")
+	ErrMemberNotFound = grpc.Errorf(codes.NotFound, "etcdserver: member not found")
+
+	ErrRequestTooLarge = grpc.Errorf(codes.InvalidArgument, "etcdserver: request is too large")
+
+	ErrUserAlreadyExist = grpc.Errorf(codes.FailedPrecondition, "etcdserver: user name already exists")
+	ErrUserNotFound     = grpc.Errorf(codes.FailedPrecondition, "etcdserver: user name not found")
+	ErrRoleAlreadyExist = grpc.Errorf(codes.FailedPrecondition, "etcdserver: role name already exists")
+)

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/util.go

@@ -0,0 +1,49 @@
+// Copyright 2016 Nippon Telegraph and Telephone Corporation.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v3rpc
+
+import (
+	"github.com/coreos/etcd/auth"
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/api/v3rpc/rpctypes"
+	"github.com/coreos/etcd/lease"
+	"github.com/coreos/etcd/storage"
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/codes"
+)
+
+func togRPCError(err error) error {
+	switch err {
+	case storage.ErrCompacted:
+		return rpctypes.ErrCompacted
+	case storage.ErrFutureRev:
+		return rpctypes.ErrFutureRev
+	case lease.ErrLeaseNotFound:
+		return rpctypes.ErrLeaseNotFound
+	// TODO: handle error from raft and timeout
+	case etcdserver.ErrRequestTooLarge:
+		return rpctypes.ErrRequestTooLarge
+	case etcdserver.ErrNoSpace:
+		return rpctypes.ErrNoSpace
+	case auth.ErrUserAlreadyExist:
+		return rpctypes.ErrUserAlreadyExist
+	case auth.ErrUserNotFound:
+		return rpctypes.ErrUserNotFound
+	case auth.ErrRoleAlreadyExist:
+		return rpctypes.ErrRoleAlreadyExist
+	default:
+		return grpc.Errorf(codes.Internal, err.Error())
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/watch.go

@@ -0,0 +1,287 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package v3rpc
+
+import (
+	"io"
+	"sync"
+	"time"
+
+	"github.com/coreos/etcd/etcdserver"
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/storage"
+	"github.com/coreos/etcd/storage/storagepb"
+)
+
+type watchServer struct {
+	clusterID int64
+	memberID  int64
+	raftTimer etcdserver.RaftTimer
+	watchable storage.Watchable
+}
+
+func NewWatchServer(s *etcdserver.EtcdServer) pb.WatchServer {
+	return &watchServer{
+		clusterID: int64(s.Cluster().ID()),
+		memberID:  int64(s.ID()),
+		raftTimer: s,
+		watchable: s.Watchable(),
+	}
+}
+
+var (
+	// External test can read this with GetProgressReportInterval()
+	// and change this to a small value to finish fast with
+	// SetProgressReportInterval().
+	progressReportInterval   = 10 * time.Minute
+	progressReportIntervalMu sync.RWMutex
+)
+
+func GetProgressReportInterval() time.Duration {
+	progressReportIntervalMu.RLock()
+	defer progressReportIntervalMu.RUnlock()
+	return progressReportInterval
+}
+
+func SetProgressReportInterval(newTimeout time.Duration) {
+	progressReportIntervalMu.Lock()
+	defer progressReportIntervalMu.Unlock()
+	progressReportInterval = newTimeout
+}
+
+const (
+	// We send ctrl response inside the read loop. We do not want
+	// send to block read, but we still want ctrl response we sent to
+	// be serialized. Thus we use a buffered chan to solve the problem.
+	// A small buffer should be OK for most cases, since we expect the
+	// ctrl requests are infrequent.
+	ctrlStreamBufLen = 16
+)
+
+// serverWatchStream is an etcd server side stream. It receives requests
+// from client side gRPC stream. It receives watch events from storage.WatchStream,
+// and creates responses that forwarded to gRPC stream.
+// It also forwards control message like watch created and canceled.
+type serverWatchStream struct {
+	clusterID int64
+	memberID  int64
+	raftTimer etcdserver.RaftTimer
+
+	gRPCStream  pb.Watch_WatchServer
+	watchStream storage.WatchStream
+	ctrlStream  chan *pb.WatchResponse
+
+	// progress tracks the watchID that stream might need to send
+	// progress to.
+	progress map[storage.WatchID]bool
+	// mu protects progress
+	mu sync.Mutex
+
+	// closec indicates the stream is closed.
+	closec chan struct{}
+}
+
+func (ws *watchServer) Watch(stream pb.Watch_WatchServer) error {
+	sws := serverWatchStream{
+		clusterID:   ws.clusterID,
+		memberID:    ws.memberID,
+		raftTimer:   ws.raftTimer,
+		gRPCStream:  stream,
+		watchStream: ws.watchable.NewWatchStream(),
+		// chan for sending control response like watcher created and canceled.
+		ctrlStream: make(chan *pb.WatchResponse, ctrlStreamBufLen),
+		progress:   make(map[storage.WatchID]bool),
+		closec:     make(chan struct{}),
+	}
+	defer sws.close()
+
+	go sws.sendLoop()
+	return sws.recvLoop()
+}
+
+func (sws *serverWatchStream) recvLoop() error {
+	for {
+		req, err := sws.gRPCStream.Recv()
+		if err == io.EOF {
+			return nil
+		}
+		if err != nil {
+			return err
+		}
+
+		switch uv := req.RequestUnion.(type) {
+		case *pb.WatchRequest_CreateRequest:
+			if uv.CreateRequest == nil {
+				break
+			}
+
+			creq := uv.CreateRequest
+			if len(creq.Key) == 0 {
+				// \x00 is the smallest key
+				creq.Key = []byte{0}
+			}
+			if len(creq.RangeEnd) == 1 && creq.RangeEnd[0] == 0 {
+				// support  >= key queries
+				creq.RangeEnd = []byte{}
+			}
+			wsrev := sws.watchStream.Rev()
+			rev := creq.StartRevision
+			if rev == 0 {
+				rev = wsrev + 1
+			}
+			id := sws.watchStream.Watch(creq.Key, creq.RangeEnd, rev)
+			if id != -1 && creq.ProgressNotify {
+				sws.progress[id] = true
+			}
+			sws.ctrlStream <- &pb.WatchResponse{
+				Header:   sws.newResponseHeader(wsrev),
+				WatchId:  int64(id),
+				Created:  true,
+				Canceled: id == -1,
+			}
+		case *pb.WatchRequest_CancelRequest:
+			if uv.CancelRequest != nil {
+				id := uv.CancelRequest.WatchId
+				err := sws.watchStream.Cancel(storage.WatchID(id))
+				if err == nil {
+					sws.ctrlStream <- &pb.WatchResponse{
+						Header:   sws.newResponseHeader(sws.watchStream.Rev()),
+						WatchId:  id,
+						Canceled: true,
+					}
+					sws.mu.Lock()
+					delete(sws.progress, storage.WatchID(id))
+					sws.mu.Unlock()
+				}
+			}
+			// TODO: do we need to return error back to client?
+		default:
+			panic("not implemented")
+		}
+	}
+}
+
+func (sws *serverWatchStream) sendLoop() {
+	// watch ids that are currently active
+	ids := make(map[storage.WatchID]struct{})
+	// watch responses pending on a watch id creation message
+	pending := make(map[storage.WatchID][]*pb.WatchResponse)
+
+	interval := GetProgressReportInterval()
+	progressTicker := time.NewTicker(interval)
+	defer progressTicker.Stop()
+
+	for {
+		select {
+		case wresp, ok := <-sws.watchStream.Chan():
+			if !ok {
+				return
+			}
+
+			// TODO: evs is []storagepb.Event type
+			// either return []*storagepb.Event from storage package
+			// or define protocol buffer with []storagepb.Event.
+			evs := wresp.Events
+			events := make([]*storagepb.Event, len(evs))
+			for i := range evs {
+				events[i] = &evs[i]
+			}
+
+			wr := &pb.WatchResponse{
+				Header:          sws.newResponseHeader(wresp.Revision),
+				WatchId:         int64(wresp.WatchID),
+				Events:          events,
+				CompactRevision: wresp.CompactRevision,
+			}
+
+			if _, hasId := ids[wresp.WatchID]; !hasId {
+				// buffer if id not yet announced
+				wrs := append(pending[wresp.WatchID], wr)
+				pending[wresp.WatchID] = wrs
+				continue
+			}
+
+			storage.ReportEventReceived()
+			if err := sws.gRPCStream.Send(wr); err != nil {
+				return
+			}
+
+			sws.mu.Lock()
+			if _, ok := sws.progress[wresp.WatchID]; ok {
+				sws.progress[wresp.WatchID] = false
+			}
+			sws.mu.Unlock()
+
+		case c, ok := <-sws.ctrlStream:
+			if !ok {
+				return
+			}
+
+			if err := sws.gRPCStream.Send(c); err != nil {
+				return
+			}
+
+			// track id creation
+			wid := storage.WatchID(c.WatchId)
+			if c.Canceled {
+				delete(ids, wid)
+				continue
+			}
+			if c.Created {
+				// flush buffered events
+				ids[wid] = struct{}{}
+				for _, v := range pending[wid] {
+					storage.ReportEventReceived()
+					if err := sws.gRPCStream.Send(v); err != nil {
+						return
+					}
+				}
+				delete(pending, wid)
+			}
+		case <-progressTicker.C:
+			for id, ok := range sws.progress {
+				if ok {
+					sws.watchStream.RequestProgress(id)
+				}
+				sws.progress[id] = true
+			}
+		case <-sws.closec:
+			// drain the chan to clean up pending events
+			for range sws.watchStream.Chan() {
+				storage.ReportEventReceived()
+			}
+			for _, wrs := range pending {
+				for range wrs {
+					storage.ReportEventReceived()
+				}
+			}
+		}
+	}
+}
+
+func (sws *serverWatchStream) close() {
+	sws.watchStream.Close()
+	close(sws.closec)
+	close(sws.ctrlStream)
+}
+
+func (sws *serverWatchStream) newResponseHeader(rev int64) *pb.ResponseHeader {
+	return &pb.ResponseHeader{
+		ClusterId: uint64(sws.clusterID),
+		MemberId:  uint64(sws.memberID),
+		Revision:  rev,
+		RaftTerm:  sws.raftTimer.Term(),
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/apply.go

@@ -0,0 +1,627 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	"bytes"
+	"fmt"
+	"sort"
+
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/lease"
+	"github.com/coreos/etcd/pkg/types"
+	dstorage "github.com/coreos/etcd/storage"
+	"github.com/coreos/etcd/storage/storagepb"
+	"github.com/gogo/protobuf/proto"
+)
+
+const (
+	// noTxn is an invalid txn ID.
+	// To apply with independent Range, Put, Delete, you can pass noTxn
+	// to apply functions instead of a valid txn ID.
+	noTxn = -1
+)
+
+type applyResult struct {
+	resp proto.Message
+	err  error
+	// physc signals the physical effect of the request has completed in addition
+	// to being logically reflected by the node. Currently only used for
+	// Compaction requests.
+	physc <-chan struct{}
+}
+
+// applierV3 is the interface for processing V3 raft messages
+type applierV3 interface {
+	Put(txnID int64, p *pb.PutRequest) (*pb.PutResponse, error)
+	Range(txnID int64, r *pb.RangeRequest) (*pb.RangeResponse, error)
+	DeleteRange(txnID int64, dr *pb.DeleteRangeRequest) (*pb.DeleteRangeResponse, error)
+	Txn(rt *pb.TxnRequest) (*pb.TxnResponse, error)
+	Compaction(compaction *pb.CompactionRequest) (*pb.CompactionResponse, <-chan struct{}, error)
+	LeaseGrant(lc *pb.LeaseGrantRequest) (*pb.LeaseGrantResponse, error)
+	LeaseRevoke(lc *pb.LeaseRevokeRequest) (*pb.LeaseRevokeResponse, error)
+	Alarm(*pb.AlarmRequest) (*pb.AlarmResponse, error)
+	AuthEnable() (*pb.AuthEnableResponse, error)
+	UserAdd(ua *pb.AuthUserAddRequest) (*pb.AuthUserAddResponse, error)
+	UserDelete(ua *pb.AuthUserDeleteRequest) (*pb.AuthUserDeleteResponse, error)
+	UserChangePassword(ua *pb.AuthUserChangePasswordRequest) (*pb.AuthUserChangePasswordResponse, error)
+	RoleAdd(ua *pb.AuthRoleAddRequest) (*pb.AuthRoleAddResponse, error)
+}
+
+type applierV3backend struct {
+	s *EtcdServer
+}
+
+func (s *EtcdServer) applyV3Request(r *pb.InternalRaftRequest) *applyResult {
+	ar := &applyResult{}
+	switch {
+	case r.Range != nil:
+		ar.resp, ar.err = s.applyV3.Range(noTxn, r.Range)
+	case r.Put != nil:
+		ar.resp, ar.err = s.applyV3.Put(noTxn, r.Put)
+	case r.DeleteRange != nil:
+		ar.resp, ar.err = s.applyV3.DeleteRange(noTxn, r.DeleteRange)
+	case r.Txn != nil:
+		ar.resp, ar.err = s.applyV3.Txn(r.Txn)
+	case r.Compaction != nil:
+		ar.resp, ar.physc, ar.err = s.applyV3.Compaction(r.Compaction)
+	case r.LeaseGrant != nil:
+		ar.resp, ar.err = s.applyV3.LeaseGrant(r.LeaseGrant)
+	case r.LeaseRevoke != nil:
+		ar.resp, ar.err = s.applyV3.LeaseRevoke(r.LeaseRevoke)
+	case r.Alarm != nil:
+		ar.resp, ar.err = s.applyV3.Alarm(r.Alarm)
+	case r.AuthEnable != nil:
+		ar.resp, ar.err = s.applyV3.AuthEnable()
+	case r.AuthUserAdd != nil:
+		ar.resp, ar.err = s.applyV3.UserAdd(r.AuthUserAdd)
+	case r.AuthUserDelete != nil:
+		ar.resp, ar.err = s.applyV3.UserDelete(r.AuthUserDelete)
+	case r.AuthUserChangePassword != nil:
+		ar.resp, ar.err = s.applyV3.UserChangePassword(r.AuthUserChangePassword)
+	case r.AuthRoleAdd != nil:
+		ar.resp, ar.err = s.applyV3.RoleAdd(r.AuthRoleAdd)
+	default:
+		panic("not implemented")
+	}
+	return ar
+}
+
+func (a *applierV3backend) Put(txnID int64, p *pb.PutRequest) (*pb.PutResponse, error) {
+	resp := &pb.PutResponse{}
+	resp.Header = &pb.ResponseHeader{}
+	var (
+		rev int64
+		err error
+	)
+	if txnID != noTxn {
+		rev, err = a.s.KV().TxnPut(txnID, p.Key, p.Value, lease.LeaseID(p.Lease))
+		if err != nil {
+			return nil, err
+		}
+	} else {
+		leaseID := lease.LeaseID(p.Lease)
+		if leaseID != lease.NoLease {
+			if l := a.s.lessor.Lookup(leaseID); l == nil {
+				return nil, lease.ErrLeaseNotFound
+			}
+		}
+		rev = a.s.KV().Put(p.Key, p.Value, leaseID)
+	}
+	resp.Header.Revision = rev
+	return resp, nil
+}
+
+func (a *applierV3backend) DeleteRange(txnID int64, dr *pb.DeleteRangeRequest) (*pb.DeleteRangeResponse, error) {
+	resp := &pb.DeleteRangeResponse{}
+	resp.Header = &pb.ResponseHeader{}
+
+	var (
+		n   int64
+		rev int64
+		err error
+	)
+
+	if isGteRange(dr.RangeEnd) {
+		dr.RangeEnd = []byte{}
+	}
+
+	if txnID != noTxn {
+		n, rev, err = a.s.KV().TxnDeleteRange(txnID, dr.Key, dr.RangeEnd)
+		if err != nil {
+			return nil, err
+		}
+	} else {
+		n, rev = a.s.KV().DeleteRange(dr.Key, dr.RangeEnd)
+	}
+
+	resp.Deleted = n
+	resp.Header.Revision = rev
+	return resp, nil
+}
+
+func (a *applierV3backend) Range(txnID int64, r *pb.RangeRequest) (*pb.RangeResponse, error) {
+	resp := &pb.RangeResponse{}
+	resp.Header = &pb.ResponseHeader{}
+
+	var (
+		kvs []storagepb.KeyValue
+		rev int64
+		err error
+	)
+
+	if isGteRange(r.RangeEnd) {
+		r.RangeEnd = []byte{}
+	}
+
+	limit := r.Limit
+	if r.SortOrder != pb.RangeRequest_NONE {
+		// fetch everything; sort and truncate afterwards
+		limit = 0
+	}
+	if limit > 0 {
+		// fetch one extra for 'more' flag
+		limit = limit + 1
+	}
+
+	if txnID != noTxn {
+		kvs, rev, err = a.s.KV().TxnRange(txnID, r.Key, r.RangeEnd, limit, r.Revision)
+		if err != nil {
+			return nil, err
+		}
+	} else {
+		kvs, rev, err = a.s.KV().Range(r.Key, r.RangeEnd, limit, r.Revision)
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	if r.SortOrder != pb.RangeRequest_NONE {
+		var sorter sort.Interface
+		switch {
+		case r.SortTarget == pb.RangeRequest_KEY:
+			sorter = &kvSortByKey{&kvSort{kvs}}
+		case r.SortTarget == pb.RangeRequest_VERSION:
+			sorter = &kvSortByVersion{&kvSort{kvs}}
+		case r.SortTarget == pb.RangeRequest_CREATE:
+			sorter = &kvSortByCreate{&kvSort{kvs}}
+		case r.SortTarget == pb.RangeRequest_MOD:
+			sorter = &kvSortByMod{&kvSort{kvs}}
+		case r.SortTarget == pb.RangeRequest_VALUE:
+			sorter = &kvSortByValue{&kvSort{kvs}}
+		}
+		switch {
+		case r.SortOrder == pb.RangeRequest_ASCEND:
+			sort.Sort(sorter)
+		case r.SortOrder == pb.RangeRequest_DESCEND:
+			sort.Sort(sort.Reverse(sorter))
+		}
+	}
+
+	if r.Limit > 0 && len(kvs) > int(r.Limit) {
+		kvs = kvs[:r.Limit]
+		resp.More = true
+	}
+
+	resp.Header.Revision = rev
+	for i := range kvs {
+		resp.Kvs = append(resp.Kvs, &kvs[i])
+	}
+	return resp, nil
+}
+
+func (a *applierV3backend) Txn(rt *pb.TxnRequest) (*pb.TxnResponse, error) {
+	var revision int64
+
+	ok := true
+	for _, c := range rt.Compare {
+		if revision, ok = a.applyCompare(c); !ok {
+			break
+		}
+	}
+
+	var reqs []*pb.RequestUnion
+	if ok {
+		reqs = rt.Success
+	} else {
+		reqs = rt.Failure
+	}
+
+	if err := a.checkRequestLeases(reqs); err != nil {
+		return nil, err
+	}
+	if err := a.checkRequestRange(reqs); err != nil {
+		return nil, err
+	}
+
+	// When executing the operations of txn, we need to hold the txn lock.
+	// So the reader will not see any intermediate results.
+	txnID := a.s.KV().TxnBegin()
+	defer func() {
+		err := a.s.KV().TxnEnd(txnID)
+		if err != nil {
+			panic(fmt.Sprint("unexpected error when closing txn", txnID))
+		}
+	}()
+
+	resps := make([]*pb.ResponseUnion, len(reqs))
+	for i := range reqs {
+		resps[i] = a.applyUnion(txnID, reqs[i])
+	}
+
+	if len(resps) != 0 {
+		revision += 1
+	}
+
+	txnResp := &pb.TxnResponse{}
+	txnResp.Header = &pb.ResponseHeader{}
+	txnResp.Header.Revision = revision
+	txnResp.Responses = resps
+	txnResp.Succeeded = ok
+	return txnResp, nil
+}
+
+// applyCompare applies the compare request.
+// It returns the revision at which the comparison happens. If the comparison
+// succeeds, the it returns true. Otherwise it returns false.
+func (a *applierV3backend) applyCompare(c *pb.Compare) (int64, bool) {
+	ckvs, rev, err := a.s.KV().Range(c.Key, nil, 1, 0)
+	if err != nil {
+		if err == dstorage.ErrTxnIDMismatch {
+			panic("unexpected txn ID mismatch error")
+		}
+		return rev, false
+	}
+	var ckv storagepb.KeyValue
+	if len(ckvs) != 0 {
+		ckv = ckvs[0]
+	} else {
+		// Use the zero value of ckv normally. However...
+		if c.Target == pb.Compare_VALUE {
+			// Always fail if we're comparing a value on a key that doesn't exist.
+			// We can treat non-existence as the empty set explicitly, such that
+			// even a key with a value of length 0 bytes is still a real key
+			// that was written that way
+			return rev, false
+		}
+	}
+
+	// -1 is less, 0 is equal, 1 is greater
+	var result int
+	switch c.Target {
+	case pb.Compare_VALUE:
+		tv, _ := c.TargetUnion.(*pb.Compare_Value)
+		if tv != nil {
+			result = bytes.Compare(ckv.Value, tv.Value)
+		}
+	case pb.Compare_CREATE:
+		tv, _ := c.TargetUnion.(*pb.Compare_CreateRevision)
+		if tv != nil {
+			result = compareInt64(ckv.CreateRevision, tv.CreateRevision)
+		}
+
+	case pb.Compare_MOD:
+		tv, _ := c.TargetUnion.(*pb.Compare_ModRevision)
+		if tv != nil {
+			result = compareInt64(ckv.ModRevision, tv.ModRevision)
+		}
+	case pb.Compare_VERSION:
+		tv, _ := c.TargetUnion.(*pb.Compare_Version)
+		if tv != nil {
+			result = compareInt64(ckv.Version, tv.Version)
+		}
+	}
+
+	switch c.Result {
+	case pb.Compare_EQUAL:
+		if result != 0 {
+			return rev, false
+		}
+	case pb.Compare_GREATER:
+		if result != 1 {
+			return rev, false
+		}
+	case pb.Compare_LESS:
+		if result != -1 {
+			return rev, false
+		}
+	}
+	return rev, true
+}
+
+func (a *applierV3backend) applyUnion(txnID int64, union *pb.RequestUnion) *pb.ResponseUnion {
+	switch tv := union.Request.(type) {
+	case *pb.RequestUnion_RequestRange:
+		if tv.RequestRange != nil {
+			resp, err := a.Range(txnID, tv.RequestRange)
+			if err != nil {
+				panic("unexpected error during txn")
+			}
+			return &pb.ResponseUnion{Response: &pb.ResponseUnion_ResponseRange{ResponseRange: resp}}
+		}
+	case *pb.RequestUnion_RequestPut:
+		if tv.RequestPut != nil {
+			resp, err := a.Put(txnID, tv.RequestPut)
+			if err != nil {
+				panic("unexpected error during txn")
+			}
+			return &pb.ResponseUnion{Response: &pb.ResponseUnion_ResponsePut{ResponsePut: resp}}
+		}
+	case *pb.RequestUnion_RequestDeleteRange:
+		if tv.RequestDeleteRange != nil {
+			resp, err := a.DeleteRange(txnID, tv.RequestDeleteRange)
+			if err != nil {
+				panic("unexpected error during txn")
+			}
+			return &pb.ResponseUnion{Response: &pb.ResponseUnion_ResponseDeleteRange{ResponseDeleteRange: resp}}
+		}
+	default:
+		// empty union
+		return nil
+	}
+	return nil
+
+}
+
+func (a *applierV3backend) Compaction(compaction *pb.CompactionRequest) (*pb.CompactionResponse, <-chan struct{}, error) {
+	resp := &pb.CompactionResponse{}
+	resp.Header = &pb.ResponseHeader{}
+	ch, err := a.s.KV().Compact(compaction.Revision)
+	if err != nil {
+		return nil, ch, err
+	}
+	// get the current revision. which key to get is not important.
+	_, resp.Header.Revision, _ = a.s.KV().Range([]byte("compaction"), nil, 1, 0)
+	return resp, ch, err
+}
+
+func (a *applierV3backend) LeaseGrant(lc *pb.LeaseGrantRequest) (*pb.LeaseGrantResponse, error) {
+	l, err := a.s.lessor.Grant(lease.LeaseID(lc.ID), lc.TTL)
+	resp := &pb.LeaseGrantResponse{}
+	if err == nil {
+		resp.ID = int64(l.ID)
+		resp.TTL = l.TTL
+	}
+	return resp, err
+}
+
+func (a *applierV3backend) LeaseRevoke(lc *pb.LeaseRevokeRequest) (*pb.LeaseRevokeResponse, error) {
+	err := a.s.lessor.Revoke(lease.LeaseID(lc.ID))
+	return &pb.LeaseRevokeResponse{}, err
+}
+
+func (a *applierV3backend) Alarm(ar *pb.AlarmRequest) (*pb.AlarmResponse, error) {
+	resp := &pb.AlarmResponse{}
+	oldCount := len(a.s.alarmStore.Get(ar.Alarm))
+
+	switch ar.Action {
+	case pb.AlarmRequest_GET:
+		resp.Alarms = a.s.alarmStore.Get(ar.Alarm)
+	case pb.AlarmRequest_ACTIVATE:
+		m := a.s.alarmStore.Activate(types.ID(ar.MemberID), ar.Alarm)
+		if m == nil {
+			break
+		}
+		resp.Alarms = append(resp.Alarms, m)
+		activated := oldCount == 0 && len(a.s.alarmStore.Get(m.Alarm)) == 1
+		if !activated {
+			break
+		}
+
+		switch m.Alarm {
+		case pb.AlarmType_NOSPACE:
+			plog.Warningf("alarm raised %+v", m)
+			a.s.applyV3 = newApplierV3Capped(a)
+		default:
+			plog.Errorf("unimplemented alarm activation (%+v)", m)
+		}
+	case pb.AlarmRequest_DEACTIVATE:
+		m := a.s.alarmStore.Deactivate(types.ID(ar.MemberID), ar.Alarm)
+		if m == nil {
+			break
+		}
+		resp.Alarms = append(resp.Alarms, m)
+		deactivated := oldCount > 0 && len(a.s.alarmStore.Get(ar.Alarm)) == 0
+		if !deactivated {
+			break
+		}
+
+		switch m.Alarm {
+		case pb.AlarmType_NOSPACE:
+			plog.Infof("alarm disarmed %+v", ar)
+			a.s.applyV3 = newQuotaApplierV3(a.s, &applierV3backend{a.s})
+		default:
+			plog.Errorf("unimplemented alarm deactivation (%+v)", m)
+		}
+	default:
+		return nil, nil
+	}
+	return resp, nil
+}
+
+type applierV3Capped struct {
+	applierV3
+	q backendQuota
+}
+
+// newApplierV3Capped creates an applyV3 that will reject Puts and transactions
+// with Puts so that the number of keys in the store is capped.
+func newApplierV3Capped(base applierV3) applierV3 { return &applierV3Capped{applierV3: base} }
+
+func (a *applierV3Capped) Put(txnID int64, p *pb.PutRequest) (*pb.PutResponse, error) {
+	return nil, ErrNoSpace
+}
+
+func (a *applierV3Capped) Txn(r *pb.TxnRequest) (*pb.TxnResponse, error) {
+	if a.q.Cost(r) > 0 {
+		return nil, ErrNoSpace
+	}
+	return a.applierV3.Txn(r)
+}
+
+func (a *applierV3Capped) LeaseGrant(lc *pb.LeaseGrantRequest) (*pb.LeaseGrantResponse, error) {
+	return nil, ErrNoSpace
+}
+
+func (a *applierV3backend) AuthEnable() (*pb.AuthEnableResponse, error) {
+	a.s.AuthStore().AuthEnable()
+	return &pb.AuthEnableResponse{}, nil
+}
+
+func (a *applierV3backend) UserAdd(r *pb.AuthUserAddRequest) (*pb.AuthUserAddResponse, error) {
+	return a.s.AuthStore().UserAdd(r)
+}
+
+func (a *applierV3backend) UserDelete(r *pb.AuthUserDeleteRequest) (*pb.AuthUserDeleteResponse, error) {
+	return a.s.AuthStore().UserDelete(r)
+}
+
+func (a *applierV3backend) UserChangePassword(r *pb.AuthUserChangePasswordRequest) (*pb.AuthUserChangePasswordResponse, error) {
+	return a.s.AuthStore().UserChangePassword(r)
+}
+
+func (a *applierV3backend) RoleAdd(r *pb.AuthRoleAddRequest) (*pb.AuthRoleAddResponse, error) {
+	return a.s.AuthStore().RoleAdd(r)
+}
+
+type quotaApplierV3 struct {
+	applierV3
+	q Quota
+}
+
+func newQuotaApplierV3(s *EtcdServer, app applierV3) applierV3 {
+	return &quotaApplierV3{app, NewBackendQuota(s)}
+}
+
+func (a *quotaApplierV3) Put(txnID int64, p *pb.PutRequest) (*pb.PutResponse, error) {
+	ok := a.q.Available(p)
+	resp, err := a.applierV3.Put(txnID, p)
+	if err == nil && !ok {
+		err = ErrNoSpace
+	}
+	return resp, err
+}
+
+func (a *quotaApplierV3) Txn(rt *pb.TxnRequest) (*pb.TxnResponse, error) {
+	ok := a.q.Available(rt)
+	resp, err := a.applierV3.Txn(rt)
+	if err == nil && !ok {
+		err = ErrNoSpace
+	}
+	return resp, err
+}
+
+func (a *quotaApplierV3) LeaseGrant(lc *pb.LeaseGrantRequest) (*pb.LeaseGrantResponse, error) {
+	ok := a.q.Available(lc)
+	resp, err := a.applierV3.LeaseGrant(lc)
+	if err == nil && !ok {
+		err = ErrNoSpace
+	}
+	return resp, err
+}
+
+type kvSort struct{ kvs []storagepb.KeyValue }
+
+func (s *kvSort) Swap(i, j int) {
+	t := s.kvs[i]
+	s.kvs[i] = s.kvs[j]
+	s.kvs[j] = t
+}
+func (s *kvSort) Len() int { return len(s.kvs) }
+
+type kvSortByKey struct{ *kvSort }
+
+func (s *kvSortByKey) Less(i, j int) bool {
+	return bytes.Compare(s.kvs[i].Key, s.kvs[j].Key) < 0
+}
+
+type kvSortByVersion struct{ *kvSort }
+
+func (s *kvSortByVersion) Less(i, j int) bool {
+	return (s.kvs[i].Version - s.kvs[j].Version) < 0
+}
+
+type kvSortByCreate struct{ *kvSort }
+
+func (s *kvSortByCreate) Less(i, j int) bool {
+	return (s.kvs[i].CreateRevision - s.kvs[j].CreateRevision) < 0
+}
+
+type kvSortByMod struct{ *kvSort }
+
+func (s *kvSortByMod) Less(i, j int) bool {
+	return (s.kvs[i].ModRevision - s.kvs[j].ModRevision) < 0
+}
+
+type kvSortByValue struct{ *kvSort }
+
+func (s *kvSortByValue) Less(i, j int) bool {
+	return bytes.Compare(s.kvs[i].Value, s.kvs[j].Value) < 0
+}
+
+func (a *applierV3backend) checkRequestLeases(reqs []*pb.RequestUnion) error {
+	for _, requ := range reqs {
+		tv, ok := requ.Request.(*pb.RequestUnion_RequestPut)
+		if !ok {
+			continue
+		}
+		preq := tv.RequestPut
+		if preq == nil || lease.LeaseID(preq.Lease) == lease.NoLease {
+			continue
+		}
+		if l := a.s.lessor.Lookup(lease.LeaseID(preq.Lease)); l == nil {
+			return lease.ErrLeaseNotFound
+		}
+	}
+	return nil
+}
+
+func (a *applierV3backend) checkRequestRange(reqs []*pb.RequestUnion) error {
+	for _, requ := range reqs {
+		tv, ok := requ.Request.(*pb.RequestUnion_RequestRange)
+		if !ok {
+			continue
+		}
+		greq := tv.RequestRange
+		if greq == nil || greq.Revision == 0 {
+			continue
+		}
+
+		if greq.Revision > a.s.KV().Rev() {
+			return dstorage.ErrFutureRev
+		}
+		if greq.Revision < a.s.KV().FirstRev() {
+			return dstorage.ErrCompacted
+		}
+	}
+	return nil
+}
+
+func compareInt64(a, b int64) int {
+	switch {
+	case a < b:
+		return -1
+	case a > b:
+		return 1
+	default:
+		return 0
+	}
+}
+
+// isGteRange determines if the range end is a >= range. This works around grpc
+// sending empty byte strings as nil; >= is encoded in the range end as '\0'.
+func isGteRange(rangeEnd []byte) bool {
+	return len(rangeEnd) == 1 && rangeEnd[0] == 0
+}

vendor/github.com/coreos/etcd/etcdserver/auth/auth.go

@@ -0,0 +1,652 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package auth implements etcd authentication.
+package auth
+
+import (
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"path"
+	"reflect"
+	"sort"
+	"strings"
+	"time"
+
+	etcderr "github.com/coreos/etcd/error"
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/pkg/types"
+	"github.com/coreos/pkg/capnslog"
+	"golang.org/x/crypto/bcrypt"
+	"golang.org/x/net/context"
+)
+
+const (
+	// StorePermsPrefix is the internal prefix of the storage layer dedicated to storing user data.
+	StorePermsPrefix = "/2"
+
+	// RootRoleName is the name of the ROOT role, with privileges to manage the cluster.
+	RootRoleName = "root"
+
+	// GuestRoleName is the name of the role that defines the privileges of an unauthenticated user.
+	GuestRoleName = "guest"
+)
+
+var (
+	plog = capnslog.NewPackageLogger("github.com/coreos/etcd/etcdserver", "auth")
+)
+
+var rootRole = Role{
+	Role: RootRoleName,
+	Permissions: Permissions{
+		KV: RWPermission{
+			Read:  []string{"/*"},
+			Write: []string{"/*"},
+		},
+	},
+}
+
+var guestRole = Role{
+	Role: GuestRoleName,
+	Permissions: Permissions{
+		KV: RWPermission{
+			Read:  []string{"/*"},
+			Write: []string{"/*"},
+		},
+	},
+}
+
+type doer interface {
+	Do(context.Context, etcdserverpb.Request) (etcdserver.Response, error)
+}
+
+type Store interface {
+	AllUsers() ([]string, error)
+	GetUser(name string) (User, error)
+	CreateOrUpdateUser(user User) (out User, created bool, err error)
+	CreateUser(user User) (User, error)
+	DeleteUser(name string) error
+	UpdateUser(user User) (User, error)
+	AllRoles() ([]string, error)
+	GetRole(name string) (Role, error)
+	CreateRole(role Role) error
+	DeleteRole(name string) error
+	UpdateRole(role Role) (Role, error)
+	AuthEnabled() bool
+	EnableAuth() error
+	DisableAuth() error
+	PasswordStore
+}
+
+type PasswordStore interface {
+	CheckPassword(user User, password string) bool
+	HashPassword(password string) (string, error)
+}
+
+type store struct {
+	server      doer
+	timeout     time.Duration
+	ensuredOnce bool
+
+	PasswordStore
+}
+
+type User struct {
+	User     string   `json:"user"`
+	Password string   `json:"password,omitempty"`
+	Roles    []string `json:"roles"`
+	Grant    []string `json:"grant,omitempty"`
+	Revoke   []string `json:"revoke,omitempty"`
+}
+
+type Role struct {
+	Role        string       `json:"role"`
+	Permissions Permissions  `json:"permissions"`
+	Grant       *Permissions `json:"grant,omitempty"`
+	Revoke      *Permissions `json:"revoke,omitempty"`
+}
+
+type Permissions struct {
+	KV RWPermission `json:"kv"`
+}
+
+func (p *Permissions) IsEmpty() bool {
+	return p == nil || (len(p.KV.Read) == 0 && len(p.KV.Write) == 0)
+}
+
+type RWPermission struct {
+	Read  []string `json:"read"`
+	Write []string `json:"write"`
+}
+
+type Error struct {
+	Status int
+	Errmsg string
+}
+
+func (ae Error) Error() string   { return ae.Errmsg }
+func (ae Error) HTTPStatus() int { return ae.Status }
+
+func authErr(hs int, s string, v ...interface{}) Error {
+	return Error{Status: hs, Errmsg: fmt.Sprintf("auth: "+s, v...)}
+}
+
+func NewStore(server doer, timeout time.Duration) Store {
+	s := &store{
+		server:        server,
+		timeout:       timeout,
+		PasswordStore: passwordStore{},
+	}
+	return s
+}
+
+// passwordStore implements PasswordStore using bcrypt to hash user passwords
+type passwordStore struct{}
+
+func (_ passwordStore) CheckPassword(user User, password string) bool {
+	err := bcrypt.CompareHashAndPassword([]byte(user.Password), []byte(password))
+	return err == nil
+}
+
+func (_ passwordStore) HashPassword(password string) (string, error) {
+	hash, err := bcrypt.GenerateFromPassword([]byte(password), bcrypt.DefaultCost)
+	return string(hash), err
+}
+
+func (s *store) AllUsers() ([]string, error) {
+	resp, err := s.requestResource("/users/", false)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeKeyNotFound {
+				return []string{}, nil
+			}
+		}
+		return nil, err
+	}
+	var nodes []string
+	for _, n := range resp.Event.Node.Nodes {
+		_, user := path.Split(n.Key)
+		nodes = append(nodes, user)
+	}
+	sort.Strings(nodes)
+	return nodes, nil
+}
+
+func (s *store) GetUser(name string) (User, error) {
+	resp, err := s.requestResource("/users/"+name, false)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeKeyNotFound {
+				return User{}, authErr(http.StatusNotFound, "User %s does not exist.", name)
+			}
+		}
+		return User{}, err
+	}
+	var u User
+	err = json.Unmarshal([]byte(*resp.Event.Node.Value), &u)
+	if err != nil {
+		return u, err
+	}
+	// Attach root role to root user.
+	if u.User == "root" {
+		u = attachRootRole(u)
+	}
+	return u, nil
+}
+
+// CreateOrUpdateUser should be only used for creating the new user or when you are not
+// sure if it is a create or update. (When only password is passed in, we are not sure
+// if it is a update or create)
+func (s *store) CreateOrUpdateUser(user User) (out User, created bool, err error) {
+	_, err = s.GetUser(user.User)
+	if err == nil {
+		out, err = s.UpdateUser(user)
+		return out, false, err
+	}
+	u, err := s.CreateUser(user)
+	return u, true, err
+}
+
+func (s *store) CreateUser(user User) (User, error) {
+	// Attach root role to root user.
+	if user.User == "root" {
+		user = attachRootRole(user)
+	}
+	u, err := s.createUserInternal(user)
+	if err == nil {
+		plog.Noticef("created user %s", user.User)
+	}
+	return u, err
+}
+
+func (s *store) createUserInternal(user User) (User, error) {
+	if user.Password == "" {
+		return user, authErr(http.StatusBadRequest, "Cannot create user %s with an empty password", user.User)
+	}
+	hash, err := s.HashPassword(user.Password)
+	if err != nil {
+		return user, err
+	}
+	user.Password = hash
+
+	_, err = s.createResource("/users/"+user.User, user)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeNodeExist {
+				return user, authErr(http.StatusConflict, "User %s already exists.", user.User)
+			}
+		}
+	}
+	return user, err
+}
+
+func (s *store) DeleteUser(name string) error {
+	if s.AuthEnabled() && name == "root" {
+		return authErr(http.StatusForbidden, "Cannot delete root user while auth is enabled.")
+	}
+	_, err := s.deleteResource("/users/" + name)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeKeyNotFound {
+				return authErr(http.StatusNotFound, "User %s does not exist", name)
+			}
+		}
+		return err
+	}
+	plog.Noticef("deleted user %s", name)
+	return nil
+}
+
+func (s *store) UpdateUser(user User) (User, error) {
+	old, err := s.GetUser(user.User)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeKeyNotFound {
+				return user, authErr(http.StatusNotFound, "User %s doesn't exist.", user.User)
+			}
+		}
+		return old, err
+	}
+
+	hash, err := s.HashPassword(user.Password)
+	if err != nil {
+		return old, err
+	}
+	user.Password = hash
+
+	newUser, err := old.merge(user)
+	if err != nil {
+		return old, err
+	}
+	if reflect.DeepEqual(old, newUser) {
+		return old, authErr(http.StatusBadRequest, "User not updated. Use grant/revoke/password to update the user.")
+	}
+	_, err = s.updateResource("/users/"+user.User, newUser)
+	if err == nil {
+		plog.Noticef("updated user %s", user.User)
+	}
+	return newUser, err
+}
+
+func (s *store) AllRoles() ([]string, error) {
+	nodes := []string{RootRoleName}
+	resp, err := s.requestResource("/roles/", false)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeKeyNotFound {
+				return nodes, nil
+			}
+		}
+		return nil, err
+	}
+	for _, n := range resp.Event.Node.Nodes {
+		_, role := path.Split(n.Key)
+		nodes = append(nodes, role)
+	}
+	sort.Strings(nodes)
+	return nodes, nil
+}
+
+func (s *store) GetRole(name string) (Role, error) {
+	if name == RootRoleName {
+		return rootRole, nil
+	}
+	resp, err := s.requestResource("/roles/"+name, false)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeKeyNotFound {
+				return Role{}, authErr(http.StatusNotFound, "Role %s does not exist.", name)
+			}
+		}
+		return Role{}, err
+	}
+	var r Role
+	err = json.Unmarshal([]byte(*resp.Event.Node.Value), &r)
+	if err != nil {
+		return r, err
+	}
+
+	return r, nil
+}
+
+func (s *store) CreateRole(role Role) error {
+	if role.Role == RootRoleName {
+		return authErr(http.StatusForbidden, "Cannot modify role %s: is root role.", role.Role)
+	}
+	_, err := s.createResource("/roles/"+role.Role, role)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeNodeExist {
+				return authErr(http.StatusConflict, "Role %s already exists.", role.Role)
+			}
+		}
+	}
+	if err == nil {
+		plog.Noticef("created new role %s", role.Role)
+	}
+	return err
+}
+
+func (s *store) DeleteRole(name string) error {
+	if name == RootRoleName {
+		return authErr(http.StatusForbidden, "Cannot modify role %s: is root role.", name)
+	}
+	_, err := s.deleteResource("/roles/" + name)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeKeyNotFound {
+				return authErr(http.StatusNotFound, "Role %s doesn't exist.", name)
+			}
+		}
+	}
+	if err == nil {
+		plog.Noticef("deleted role %s", name)
+	}
+	return err
+}
+
+func (s *store) UpdateRole(role Role) (Role, error) {
+	if role.Role == RootRoleName {
+		return Role{}, authErr(http.StatusForbidden, "Cannot modify role %s: is root role.", role.Role)
+	}
+	old, err := s.GetRole(role.Role)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeKeyNotFound {
+				return role, authErr(http.StatusNotFound, "Role %s doesn't exist.", role.Role)
+			}
+		}
+		return old, err
+	}
+	newRole, err := old.merge(role)
+	if err != nil {
+		return old, err
+	}
+	if reflect.DeepEqual(old, newRole) {
+		return old, authErr(http.StatusBadRequest, "Role not updated. Use grant/revoke to update the role.")
+	}
+	_, err = s.updateResource("/roles/"+role.Role, newRole)
+	if err == nil {
+		plog.Noticef("updated role %s", role.Role)
+	}
+	return newRole, err
+}
+
+func (s *store) AuthEnabled() bool {
+	return s.detectAuth()
+}
+
+func (s *store) EnableAuth() error {
+	if s.AuthEnabled() {
+		return authErr(http.StatusConflict, "already enabled")
+	}
+
+	if _, err := s.GetUser("root"); err != nil {
+		return authErr(http.StatusConflict, "No root user available, please create one")
+	}
+	if _, err := s.GetRole(GuestRoleName); err != nil {
+		plog.Printf("no guest role access found, creating default")
+		if err := s.CreateRole(guestRole); err != nil {
+			plog.Errorf("error creating guest role. aborting auth enable.")
+			return err
+		}
+	}
+
+	if err := s.enableAuth(); err != nil {
+		plog.Errorf("error enabling auth (%v)", err)
+		return err
+	}
+
+	plog.Noticef("auth: enabled auth")
+	return nil
+}
+
+func (s *store) DisableAuth() error {
+	if !s.AuthEnabled() {
+		return authErr(http.StatusConflict, "already disabled")
+	}
+
+	err := s.disableAuth()
+	if err == nil {
+		plog.Noticef("auth: disabled auth")
+	} else {
+		plog.Errorf("error disabling auth (%v)", err)
+	}
+	return err
+}
+
+// merge applies the properties of the passed-in User to the User on which it
+// is called and returns a new User with these modifications applied. Think of
+// all Users as immutable sets of data. Merge allows you to perform the set
+// operations (desired grants and revokes) atomically
+func (u User) merge(n User) (User, error) {
+	var out User
+	if u.User != n.User {
+		return out, authErr(http.StatusConflict, "Merging user data with conflicting usernames: %s %s", u.User, n.User)
+	}
+	out.User = u.User
+	if n.Password != "" {
+		out.Password = n.Password
+	} else {
+		out.Password = u.Password
+	}
+	currentRoles := types.NewUnsafeSet(u.Roles...)
+	for _, g := range n.Grant {
+		if currentRoles.Contains(g) {
+			plog.Noticef("granting duplicate role %s for user %s", g, n.User)
+			return User{}, authErr(http.StatusConflict, fmt.Sprintf("Granting duplicate role %s for user %s", g, n.User))
+		}
+		currentRoles.Add(g)
+	}
+	for _, r := range n.Revoke {
+		if !currentRoles.Contains(r) {
+			plog.Noticef("revoking ungranted role %s for user %s", r, n.User)
+			return User{}, authErr(http.StatusConflict, fmt.Sprintf("Revoking ungranted role %s for user %s", r, n.User))
+		}
+		currentRoles.Remove(r)
+	}
+	out.Roles = currentRoles.Values()
+	sort.Strings(out.Roles)
+	return out, nil
+}
+
+// merge for a role works the same as User above -- atomic Role application to
+// each of the substructures.
+func (r Role) merge(n Role) (Role, error) {
+	var out Role
+	var err error
+	if r.Role != n.Role {
+		return out, authErr(http.StatusConflict, "Merging role with conflicting names: %s %s", r.Role, n.Role)
+	}
+	out.Role = r.Role
+	out.Permissions, err = r.Permissions.Grant(n.Grant)
+	if err != nil {
+		return out, err
+	}
+	out.Permissions, err = out.Permissions.Revoke(n.Revoke)
+	if err != nil {
+		return out, err
+	}
+	return out, nil
+}
+
+func (r Role) HasKeyAccess(key string, write bool) bool {
+	if r.Role == RootRoleName {
+		return true
+	}
+	return r.Permissions.KV.HasAccess(key, write)
+}
+
+func (r Role) HasRecursiveAccess(key string, write bool) bool {
+	if r.Role == RootRoleName {
+		return true
+	}
+	return r.Permissions.KV.HasRecursiveAccess(key, write)
+}
+
+// Grant adds a set of permissions to the permission object on which it is called,
+// returning a new permission object.
+func (p Permissions) Grant(n *Permissions) (Permissions, error) {
+	var out Permissions
+	var err error
+	if n == nil {
+		return p, nil
+	}
+	out.KV, err = p.KV.Grant(n.KV)
+	return out, err
+}
+
+// Revoke removes a set of permissions to the permission object on which it is called,
+// returning a new permission object.
+func (p Permissions) Revoke(n *Permissions) (Permissions, error) {
+	var out Permissions
+	var err error
+	if n == nil {
+		return p, nil
+	}
+	out.KV, err = p.KV.Revoke(n.KV)
+	return out, err
+}
+
+// Grant adds a set of permissions to the permission object on which it is called,
+// returning a new permission object.
+func (rw RWPermission) Grant(n RWPermission) (RWPermission, error) {
+	var out RWPermission
+	currentRead := types.NewUnsafeSet(rw.Read...)
+	for _, r := range n.Read {
+		if currentRead.Contains(r) {
+			return out, authErr(http.StatusConflict, "Granting duplicate read permission %s", r)
+		}
+		currentRead.Add(r)
+	}
+	currentWrite := types.NewUnsafeSet(rw.Write...)
+	for _, w := range n.Write {
+		if currentWrite.Contains(w) {
+			return out, authErr(http.StatusConflict, "Granting duplicate write permission %s", w)
+		}
+		currentWrite.Add(w)
+	}
+	out.Read = currentRead.Values()
+	out.Write = currentWrite.Values()
+	sort.Strings(out.Read)
+	sort.Strings(out.Write)
+	return out, nil
+}
+
+// Revoke removes a set of permissions to the permission object on which it is called,
+// returning a new permission object.
+func (rw RWPermission) Revoke(n RWPermission) (RWPermission, error) {
+	var out RWPermission
+	currentRead := types.NewUnsafeSet(rw.Read...)
+	for _, r := range n.Read {
+		if !currentRead.Contains(r) {
+			plog.Noticef("revoking ungranted read permission %s", r)
+			continue
+		}
+		currentRead.Remove(r)
+	}
+	currentWrite := types.NewUnsafeSet(rw.Write...)
+	for _, w := range n.Write {
+		if !currentWrite.Contains(w) {
+			plog.Noticef("revoking ungranted write permission %s", w)
+			continue
+		}
+		currentWrite.Remove(w)
+	}
+	out.Read = currentRead.Values()
+	out.Write = currentWrite.Values()
+	sort.Strings(out.Read)
+	sort.Strings(out.Write)
+	return out, nil
+}
+
+func (rw RWPermission) HasAccess(key string, write bool) bool {
+	var list []string
+	if write {
+		list = rw.Write
+	} else {
+		list = rw.Read
+	}
+	for _, pat := range list {
+		match, err := simpleMatch(pat, key)
+		if err == nil && match {
+			return true
+		}
+	}
+	return false
+}
+
+func (rw RWPermission) HasRecursiveAccess(key string, write bool) bool {
+	list := rw.Read
+	if write {
+		list = rw.Write
+	}
+	for _, pat := range list {
+		match, err := prefixMatch(pat, key)
+		if err == nil && match {
+			return true
+		}
+	}
+	return false
+}
+
+func simpleMatch(pattern string, key string) (match bool, err error) {
+	if pattern[len(pattern)-1] == '*' {
+		return strings.HasPrefix(key, pattern[:len(pattern)-1]), nil
+	}
+	return key == pattern, nil
+}
+
+func prefixMatch(pattern string, key string) (match bool, err error) {
+	if pattern[len(pattern)-1] != '*' {
+		return false, nil
+	}
+	return strings.HasPrefix(key, pattern[:len(pattern)-1]), nil
+}
+
+func attachRootRole(u User) User {
+	inRoles := false
+	for _, r := range u.Roles {
+		if r == RootRoleName {
+			inRoles = true
+			break
+		}
+	}
+	if !inRoles {
+		u.Roles = append(u.Roles, RootRoleName)
+	}
+	return u
+}

vendor/github.com/coreos/etcd/etcdserver/auth/auth_requests.go

@@ -0,0 +1,162 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package auth
+
+import (
+	"encoding/json"
+	"path"
+
+	etcderr "github.com/coreos/etcd/error"
+	"github.com/coreos/etcd/etcdserver"
+	"github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"golang.org/x/net/context"
+)
+
+func (s *store) ensureAuthDirectories() error {
+	if s.ensuredOnce {
+		return nil
+	}
+	for _, res := range []string{StorePermsPrefix, StorePermsPrefix + "/users/", StorePermsPrefix + "/roles/"} {
+		ctx, cancel := context.WithTimeout(context.Background(), s.timeout)
+		defer cancel()
+		pe := false
+		rr := etcdserverpb.Request{
+			Method:    "PUT",
+			Path:      res,
+			Dir:       true,
+			PrevExist: &pe,
+		}
+		_, err := s.server.Do(ctx, rr)
+		if err != nil {
+			if e, ok := err.(*etcderr.Error); ok {
+				if e.ErrorCode == etcderr.EcodeNodeExist {
+					continue
+				}
+			}
+			plog.Errorf("failed to create auth directories in the store (%v)", err)
+			return err
+		}
+	}
+	ctx, cancel := context.WithTimeout(context.Background(), s.timeout)
+	defer cancel()
+	pe := false
+	rr := etcdserverpb.Request{
+		Method:    "PUT",
+		Path:      StorePermsPrefix + "/enabled",
+		Val:       "false",
+		PrevExist: &pe,
+	}
+	_, err := s.server.Do(ctx, rr)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeNodeExist {
+				s.ensuredOnce = true
+				return nil
+			}
+		}
+		return err
+	}
+	s.ensuredOnce = true
+	return nil
+}
+
+func (s *store) enableAuth() error {
+	_, err := s.updateResource("/enabled", true)
+	return err
+}
+func (s *store) disableAuth() error {
+	_, err := s.updateResource("/enabled", false)
+	return err
+}
+
+func (s *store) detectAuth() bool {
+	if s.server == nil {
+		return false
+	}
+	value, err := s.requestResource("/enabled", false)
+	if err != nil {
+		if e, ok := err.(*etcderr.Error); ok {
+			if e.ErrorCode == etcderr.EcodeKeyNotFound {
+				return false
+			}
+		}
+		plog.Errorf("failed to detect auth settings (%s)", err)
+		return false
+	}
+
+	var u bool
+	err = json.Unmarshal([]byte(*value.Event.Node.Value), &u)
+	if err != nil {
+		plog.Errorf("internal bookkeeping value for enabled isn't valid JSON (%v)", err)
+		return false
+	}
+	return u
+}
+
+func (s *store) requestResource(res string, dir bool) (etcdserver.Response, error) {
+	ctx, cancel := context.WithTimeout(context.Background(), s.timeout)
+	defer cancel()
+	p := path.Join(StorePermsPrefix, res)
+	rr := etcdserverpb.Request{
+		Method: "GET",
+		Path:   p,
+		Dir:    dir,
+	}
+	return s.server.Do(ctx, rr)
+}
+
+func (s *store) updateResource(res string, value interface{}) (etcdserver.Response, error) {
+	return s.setResource(res, value, true)
+}
+func (s *store) createResource(res string, value interface{}) (etcdserver.Response, error) {
+	return s.setResource(res, value, false)
+}
+func (s *store) setResource(res string, value interface{}, prevexist bool) (etcdserver.Response, error) {
+	err := s.ensureAuthDirectories()
+	if err != nil {
+		return etcdserver.Response{}, err
+	}
+	ctx, cancel := context.WithTimeout(context.Background(), s.timeout)
+	defer cancel()
+	data, err := json.Marshal(value)
+	if err != nil {
+		return etcdserver.Response{}, err
+	}
+	p := path.Join(StorePermsPrefix, res)
+	rr := etcdserverpb.Request{
+		Method:    "PUT",
+		Path:      p,
+		Val:       string(data),
+		PrevExist: &prevexist,
+	}
+	return s.server.Do(ctx, rr)
+}
+
+func (s *store) deleteResource(res string) (etcdserver.Response, error) {
+	err := s.ensureAuthDirectories()
+	if err != nil {
+		return etcdserver.Response{}, err
+	}
+	ctx, cancel := context.WithTimeout(context.Background(), s.timeout)
+	defer cancel()
+	pex := true
+	p := path.Join(StorePermsPrefix, res)
+	rr := etcdserverpb.Request{
+		Method:    "DELETE",
+		Path:      p,
+		PrevExist: &pex,
+	}
+	return s.server.Do(ctx, rr)
+}

vendor/github.com/coreos/etcd/etcdserver/cluster_util.go

@@ -0,0 +1,258 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	"encoding/json"
+	"fmt"
+	"io/ioutil"
+	"net/http"
+	"sort"
+	"time"
+
+	"github.com/coreos/etcd/etcdserver/membership"
+	"github.com/coreos/etcd/pkg/httputil"
+	"github.com/coreos/etcd/pkg/types"
+	"github.com/coreos/etcd/version"
+	"github.com/coreos/go-semver/semver"
+)
+
+// isMemberBootstrapped tries to check if the given member has been bootstrapped
+// in the given cluster.
+func isMemberBootstrapped(cl *membership.RaftCluster, member string, rt http.RoundTripper, timeout time.Duration) bool {
+	rcl, err := getClusterFromRemotePeers(getRemotePeerURLs(cl, member), timeout, false, rt)
+	if err != nil {
+		return false
+	}
+	id := cl.MemberByName(member).ID
+	m := rcl.Member(id)
+	if m == nil {
+		return false
+	}
+	if len(m.ClientURLs) > 0 {
+		return true
+	}
+	return false
+}
+
+// GetClusterFromRemotePeers takes a set of URLs representing etcd peers, and
+// attempts to construct a Cluster by accessing the members endpoint on one of
+// these URLs. The first URL to provide a response is used. If no URLs provide
+// a response, or a Cluster cannot be successfully created from a received
+// response, an error is returned.
+// Each request has a 10-second timeout. Because the upper limit of TTL is 5s,
+// 10 second is enough for building connection and finishing request.
+func GetClusterFromRemotePeers(urls []string, rt http.RoundTripper) (*membership.RaftCluster, error) {
+	return getClusterFromRemotePeers(urls, 10*time.Second, true, rt)
+}
+
+// If logerr is true, it prints out more error messages.
+func getClusterFromRemotePeers(urls []string, timeout time.Duration, logerr bool, rt http.RoundTripper) (*membership.RaftCluster, error) {
+	cc := &http.Client{
+		Transport: rt,
+		Timeout:   timeout,
+	}
+	for _, u := range urls {
+		resp, err := cc.Get(u + "/members")
+		if err != nil {
+			if logerr {
+				plog.Warningf("could not get cluster response from %s: %v", u, err)
+			}
+			continue
+		}
+		b, err := ioutil.ReadAll(resp.Body)
+		if err != nil {
+			if logerr {
+				plog.Warningf("could not read the body of cluster response: %v", err)
+			}
+			continue
+		}
+		var membs []*membership.Member
+		if err = json.Unmarshal(b, &membs); err != nil {
+			if logerr {
+				plog.Warningf("could not unmarshal cluster response: %v", err)
+			}
+			continue
+		}
+		id, err := types.IDFromString(resp.Header.Get("X-Etcd-Cluster-ID"))
+		if err != nil {
+			if logerr {
+				plog.Warningf("could not parse the cluster ID from cluster res: %v", err)
+			}
+			continue
+		}
+		return membership.NewClusterFromMembers("", id, membs), nil
+	}
+	return nil, fmt.Errorf("could not retrieve cluster information from the given urls")
+}
+
+// getRemotePeerURLs returns peer urls of remote members in the cluster. The
+// returned list is sorted in ascending lexicographical order.
+func getRemotePeerURLs(cl *membership.RaftCluster, local string) []string {
+	us := make([]string, 0)
+	for _, m := range cl.Members() {
+		if m.Name == local {
+			continue
+		}
+		us = append(us, m.PeerURLs...)
+	}
+	sort.Strings(us)
+	return us
+}
+
+// getVersions returns the versions of the members in the given cluster.
+// The key of the returned map is the member's ID. The value of the returned map
+// is the semver versions string, including server and cluster.
+// If it fails to get the version of a member, the key will be nil.
+func getVersions(cl *membership.RaftCluster, local types.ID, rt http.RoundTripper) map[string]*version.Versions {
+	members := cl.Members()
+	vers := make(map[string]*version.Versions)
+	for _, m := range members {
+		if m.ID == local {
+			cv := "not_decided"
+			if cl.Version() != nil {
+				cv = cl.Version().String()
+			}
+			vers[m.ID.String()] = &version.Versions{Server: version.Version, Cluster: cv}
+			continue
+		}
+		ver, err := getVersion(m, rt)
+		if err != nil {
+			plog.Warningf("cannot get the version of member %s (%v)", m.ID, err)
+			vers[m.ID.String()] = nil
+		} else {
+			vers[m.ID.String()] = ver
+		}
+	}
+	return vers
+}
+
+// decideClusterVersion decides the cluster version based on the versions map.
+// The returned version is the min server version in the map, or nil if the min
+// version in unknown.
+func decideClusterVersion(vers map[string]*version.Versions) *semver.Version {
+	var cv *semver.Version
+	lv := semver.Must(semver.NewVersion(version.Version))
+
+	for mid, ver := range vers {
+		if ver == nil {
+			return nil
+		}
+		v, err := semver.NewVersion(ver.Server)
+		if err != nil {
+			plog.Errorf("cannot understand the version of member %s (%v)", mid, err)
+			return nil
+		}
+		if lv.LessThan(*v) {
+			plog.Warningf("the local etcd version %s is not up-to-date", lv.String())
+			plog.Warningf("member %s has a higher version %s", mid, ver.Server)
+		}
+		if cv == nil {
+			cv = v
+		} else if v.LessThan(*cv) {
+			cv = v
+		}
+	}
+	return cv
+}
+
+// isCompatibleWithCluster return true if the local member has a compatible version with
+// the current running cluster.
+// The version is considered as compatible when at least one of the other members in the cluster has a
+// cluster version in the range of [MinClusterVersion, Version] and no known members has a cluster version
+// out of the range.
+// We set this rule since when the local member joins, another member might be offline.
+func isCompatibleWithCluster(cl *membership.RaftCluster, local types.ID, rt http.RoundTripper) bool {
+	vers := getVersions(cl, local, rt)
+	minV := semver.Must(semver.NewVersion(version.MinClusterVersion))
+	maxV := semver.Must(semver.NewVersion(version.Version))
+	maxV = &semver.Version{
+		Major: maxV.Major,
+		Minor: maxV.Minor,
+	}
+
+	return isCompatibleWithVers(vers, local, minV, maxV)
+}
+
+func isCompatibleWithVers(vers map[string]*version.Versions, local types.ID, minV, maxV *semver.Version) bool {
+	var ok bool
+	for id, v := range vers {
+		// ignore comparison with local version
+		if id == local.String() {
+			continue
+		}
+		if v == nil {
+			continue
+		}
+		clusterv, err := semver.NewVersion(v.Cluster)
+		if err != nil {
+			plog.Errorf("cannot understand the cluster version of member %s (%v)", id, err)
+			continue
+		}
+		if clusterv.LessThan(*minV) {
+			plog.Warningf("the running cluster version(%v) is lower than the minimal cluster version(%v) supported", clusterv.String(), minV.String())
+			return false
+		}
+		if maxV.LessThan(*clusterv) {
+			plog.Warningf("the running cluster version(%v) is higher than the maximum cluster version(%v) supported", clusterv.String(), maxV.String())
+			return false
+		}
+		ok = true
+	}
+	return ok
+}
+
+// getVersion returns the Versions of the given member via its
+// peerURLs. Returns the last error if it fails to get the version.
+func getVersion(m *membership.Member, rt http.RoundTripper) (*version.Versions, error) {
+	cc := &http.Client{
+		Transport: rt,
+	}
+	var (
+		err  error
+		resp *http.Response
+	)
+
+	for _, u := range m.PeerURLs {
+		resp, err = cc.Get(u + "/version")
+		if err != nil {
+			plog.Warningf("failed to reach the peerURL(%s) of member %s (%v)", u, m.ID, err)
+			continue
+		}
+		// etcd 2.0 does not have version endpoint on peer url.
+		if resp.StatusCode == http.StatusNotFound {
+			httputil.GracefulClose(resp)
+			return &version.Versions{
+				Server:  "2.0.0",
+				Cluster: "2.0.0",
+			}, nil
+		}
+
+		var b []byte
+		b, err = ioutil.ReadAll(resp.Body)
+		resp.Body.Close()
+		if err != nil {
+			plog.Warningf("failed to read out the response body from the peerURL(%s) of member %s (%v)", u, m.ID, err)
+			continue
+		}
+		var vers version.Versions
+		if err = json.Unmarshal(b, &vers); err != nil {
+			plog.Warningf("failed to unmarshal the response body got from the peerURL(%s) of member %s (%v)", u, m.ID, err)
+			continue
+		}
+		return &vers, nil
+	}
+	return nil, err
+}

vendor/github.com/coreos/etcd/etcdserver/config.go

@@ -0,0 +1,196 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	"fmt"
+	"path"
+	"sort"
+	"strings"
+	"time"
+
+	"github.com/coreos/etcd/pkg/netutil"
+	"github.com/coreos/etcd/pkg/transport"
+	"github.com/coreos/etcd/pkg/types"
+)
+
+// ServerConfig holds the configuration of etcd as taken from the command line or discovery.
+type ServerConfig struct {
+	Name           string
+	DiscoveryURL   string
+	DiscoveryProxy string
+	ClientURLs     types.URLs
+	PeerURLs       types.URLs
+	DataDir        string
+	// DedicatedWALDir config will make the etcd to write the WAL to the WALDir
+	// rather than the dataDir/member/wal.
+	DedicatedWALDir     string
+	SnapCount           uint64
+	MaxSnapFiles        uint
+	MaxWALFiles         uint
+	InitialPeerURLsMap  types.URLsMap
+	InitialClusterToken string
+	NewCluster          bool
+	ForceNewCluster     bool
+	PeerTLSInfo         transport.TLSInfo
+
+	TickMs           uint
+	ElectionTicks    int
+	BootstrapTimeout time.Duration
+
+	AutoCompactionRetention int
+	QuotaBackendBytes       int64
+
+	StrictReconfigCheck bool
+
+	EnablePprof bool
+}
+
+// VerifyBootstrap sanity-checks the initial config for bootstrap case
+// and returns an error for things that should never happen.
+func (c *ServerConfig) VerifyBootstrap() error {
+	if err := c.verifyLocalMember(true); err != nil {
+		return err
+	}
+	if checkDuplicateURL(c.InitialPeerURLsMap) {
+		return fmt.Errorf("initial cluster %s has duplicate url", c.InitialPeerURLsMap)
+	}
+	if c.InitialPeerURLsMap.String() == "" && c.DiscoveryURL == "" {
+		return fmt.Errorf("initial cluster unset and no discovery URL found")
+	}
+	return nil
+}
+
+// VerifyJoinExisting sanity-checks the initial config for join existing cluster
+// case and returns an error for things that should never happen.
+func (c *ServerConfig) VerifyJoinExisting() error {
+	// no need for strict checking since the member have announced its
+	// peer urls to the cluster before starting and do not have to set
+	// it in the configuration again.
+	if err := c.verifyLocalMember(false); err != nil {
+		return err
+	}
+	if checkDuplicateURL(c.InitialPeerURLsMap) {
+		return fmt.Errorf("initial cluster %s has duplicate url", c.InitialPeerURLsMap)
+	}
+	if c.DiscoveryURL != "" {
+		return fmt.Errorf("discovery URL should not be set when joining existing initial cluster")
+	}
+	return nil
+}
+
+// verifyLocalMember verifies the configured member is in configured
+// cluster. If strict is set, it also verifies the configured member
+// has the same peer urls as configured advertised peer urls.
+func (c *ServerConfig) verifyLocalMember(strict bool) error {
+	urls := c.InitialPeerURLsMap[c.Name]
+	// Make sure the cluster at least contains the local server.
+	if urls == nil {
+		return fmt.Errorf("couldn't find local name %q in the initial cluster configuration", c.Name)
+	}
+
+	// Advertised peer URLs must match those in the cluster peer list
+	apurls := c.PeerURLs.StringSlice()
+	sort.Strings(apurls)
+	urls.Sort()
+	if strict {
+		if !netutil.URLStringsEqual(apurls, urls.StringSlice()) {
+			umap := map[string]types.URLs{c.Name: c.PeerURLs}
+			return fmt.Errorf("--initial-cluster must include %s given --initial-advertise-peer-urls=%s", types.URLsMap(umap).String(), strings.Join(apurls, ","))
+		}
+	}
+	return nil
+}
+
+func (c *ServerConfig) MemberDir() string { return path.Join(c.DataDir, "member") }
+
+func (c *ServerConfig) WALDir() string {
+	if c.DedicatedWALDir != "" {
+		return c.DedicatedWALDir
+	}
+	return path.Join(c.MemberDir(), "wal")
+}
+
+func (c *ServerConfig) SnapDir() string { return path.Join(c.MemberDir(), "snap") }
+
+func (c *ServerConfig) ShouldDiscover() bool { return c.DiscoveryURL != "" }
+
+// ReqTimeout returns timeout for request to finish.
+func (c *ServerConfig) ReqTimeout() time.Duration {
+	// 5s for queue waiting, computation and disk IO delay
+	// + 2 * election timeout for possible leader election
+	return 5*time.Second + 2*time.Duration(c.ElectionTicks)*time.Duration(c.TickMs)*time.Millisecond
+}
+
+func (c *ServerConfig) electionTimeout() time.Duration {
+	return time.Duration(c.ElectionTicks) * time.Duration(c.TickMs) * time.Millisecond
+}
+
+func (c *ServerConfig) peerDialTimeout() time.Duration {
+	// 1s for queue wait and system delay
+	// + one RTT, which is smaller than 1/5 election timeout
+	return time.Second + time.Duration(c.ElectionTicks)*time.Duration(c.TickMs)*time.Millisecond/5
+}
+
+func (c *ServerConfig) PrintWithInitial() { c.print(true) }
+
+func (c *ServerConfig) Print() { c.print(false) }
+
+func (c *ServerConfig) print(initial bool) {
+	plog.Infof("name = %s", c.Name)
+	if c.ForceNewCluster {
+		plog.Infof("force new cluster")
+	}
+	plog.Infof("data dir = %s", c.DataDir)
+	plog.Infof("member dir = %s", c.MemberDir())
+	if c.DedicatedWALDir != "" {
+		plog.Infof("dedicated WAL dir = %s", c.DedicatedWALDir)
+	}
+	plog.Infof("heartbeat = %dms", c.TickMs)
+	plog.Infof("election = %dms", c.ElectionTicks*int(c.TickMs))
+	plog.Infof("snapshot count = %d", c.SnapCount)
+	if len(c.DiscoveryURL) != 0 {
+		plog.Infof("discovery URL= %s", c.DiscoveryURL)
+		if len(c.DiscoveryProxy) != 0 {
+			plog.Infof("discovery proxy = %s", c.DiscoveryProxy)
+		}
+	}
+	plog.Infof("advertise client URLs = %s", c.ClientURLs)
+	if initial {
+		plog.Infof("initial advertise peer URLs = %s", c.PeerURLs)
+		plog.Infof("initial cluster = %s", c.InitialPeerURLsMap)
+	}
+}
+
+func checkDuplicateURL(urlsmap types.URLsMap) bool {
+	um := make(map[string]bool)
+	for _, urls := range urlsmap {
+		for _, url := range urls {
+			u := url.String()
+			if um[u] {
+				return true
+			}
+			um[u] = true
+		}
+	}
+	return false
+}
+
+func (c *ServerConfig) bootstrapTimeout() time.Duration {
+	if c.BootstrapTimeout != 0 {
+		return c.BootstrapTimeout
+	}
+	return time.Second
+}

vendor/github.com/coreos/etcd/etcdserver/consistent_index.go

@@ -0,0 +1,25 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+// consistentIndex represents the offset of an entry in a consistent replica log.
+// It implements the storage.ConsistentIndexGetter interface.
+// It is always set to the offset of current entry before executing the entry,
+// so ConsistentWatchableKV could get the consistent index from it.
+type consistentIndex uint64
+
+func (i *consistentIndex) setConsistentIndex(v uint64) { *i = consistentIndex(v) }
+
+func (i *consistentIndex) ConsistentIndex() uint64 { return uint64(*i) }

vendor/github.com/coreos/etcd/etcdserver/doc.go

@@ -0,0 +1,16 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package etcdserver defines how etcd servers interact and store their states.
+package etcdserver

vendor/github.com/coreos/etcd/etcdserver/errors.go

@@ -0,0 +1,42 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	"errors"
+	"fmt"
+)
+
+var (
+	ErrUnknownMethod              = errors.New("etcdserver: unknown method")
+	ErrStopped                    = errors.New("etcdserver: server stopped")
+	ErrCanceled                   = errors.New("etcdserver: request cancelled")
+	ErrTimeout                    = errors.New("etcdserver: request timed out")
+	ErrTimeoutDueToLeaderFail     = errors.New("etcdserver: request timed out, possibly due to previous leader failure")
+	ErrTimeoutDueToConnectionLost = errors.New("etcdserver: request timed out, possibly due to connection lost")
+	ErrNotEnoughStartedMembers    = errors.New("etcdserver: re-configuration failed due to not enough started members")
+	ErrNoLeader                   = errors.New("etcdserver: no leader")
+	ErrRequestTooLarge            = errors.New("etcdserver: request is too large")
+	ErrNoSpace                    = errors.New("etcdserver: no space")
+)
+
+type DiscoveryError struct {
+	Op  string
+	Err error
+}
+
+func (e DiscoveryError) Error() string {
+	return fmt.Sprintf("failed to %s discovery cluster (%v)", e.Op, e.Err)
+}

vendor/github.com/coreos/etcd/etcdserver/etcdserverpb/etcdserver.pb.go

@@ -0,0 +1,997 @@
+// Code generated by protoc-gen-gogo.
+// source: etcdserver.proto
+// DO NOT EDIT!
+
+/*
+	Package etcdserverpb is a generated protocol buffer package.
+
+	It is generated from these files:
+		etcdserver.proto
+		raft_internal.proto
+		rpc.proto
+
+	It has these top-level messages:
+		Request
+		Metadata
+		InternalRaftRequest
+		EmptyResponse
+		ResponseHeader
+		RangeRequest
+		RangeResponse
+		PutRequest
+		PutResponse
+		DeleteRangeRequest
+		DeleteRangeResponse
+		RequestUnion
+		ResponseUnion
+		Compare
+		TxnRequest
+		TxnResponse
+		CompactionRequest
+		CompactionResponse
+		HashRequest
+		HashResponse
+		WatchRequest
+		WatchCreateRequest
+		WatchCancelRequest
+		WatchResponse
+		LeaseGrantRequest
+		LeaseGrantResponse
+		LeaseRevokeRequest
+		LeaseRevokeResponse
+		LeaseKeepAliveRequest
+		LeaseKeepAliveResponse
+		Member
+		MemberAddRequest
+		MemberAddResponse
+		MemberRemoveRequest
+		MemberRemoveResponse
+		MemberUpdateRequest
+		MemberUpdateResponse
+		MemberListRequest
+		MemberListResponse
+		DefragmentRequest
+		DefragmentResponse
+		AlarmRequest
+		AlarmMember
+		AlarmResponse
+		StatusRequest
+		StatusResponse
+		AuthEnableRequest
+		AuthDisableRequest
+		AuthenticateRequest
+		AuthUserAddRequest
+		AuthUserGetRequest
+		AuthUserDeleteRequest
+		AuthUserChangePasswordRequest
+		AuthUserGrantRequest
+		AuthUserRevokeRequest
+		AuthRoleAddRequest
+		AuthRoleGetRequest
+		AuthRoleDeleteRequest
+		AuthRoleGrantRequest
+		AuthRoleRevokeRequest
+		AuthEnableResponse
+		AuthDisableResponse
+		AuthenticateResponse
+		AuthUserAddResponse
+		AuthUserGetResponse
+		AuthUserDeleteResponse
+		AuthUserChangePasswordResponse
+		AuthUserGrantResponse
+		AuthUserRevokeResponse
+		AuthRoleAddResponse
+		AuthRoleGetResponse
+		AuthRoleDeleteResponse
+		AuthRoleGrantResponse
+		AuthRoleRevokeResponse
+*/
+package etcdserverpb
+
+import (
+	"fmt"
+
+	proto "github.com/gogo/protobuf/proto"
+
+	math "math"
+)
+
+import io "io"
+
+// Reference imports to suppress errors if they are not otherwise used.
+var _ = proto.Marshal
+var _ = fmt.Errorf
+var _ = math.Inf
+
+type Request struct {
+	ID               uint64 `protobuf:"varint,1,opt,name=ID" json:"ID"`
+	Method           string `protobuf:"bytes,2,opt,name=Method" json:"Method"`
+	Path             string `protobuf:"bytes,3,opt,name=Path" json:"Path"`
+	Val              string `protobuf:"bytes,4,opt,name=Val" json:"Val"`
+	Dir              bool   `protobuf:"varint,5,opt,name=Dir" json:"Dir"`
+	PrevValue        string `protobuf:"bytes,6,opt,name=PrevValue" json:"PrevValue"`
+	PrevIndex        uint64 `protobuf:"varint,7,opt,name=PrevIndex" json:"PrevIndex"`
+	PrevExist        *bool  `protobuf:"varint,8,opt,name=PrevExist" json:"PrevExist,omitempty"`
+	Expiration       int64  `protobuf:"varint,9,opt,name=Expiration" json:"Expiration"`
+	Wait             bool   `protobuf:"varint,10,opt,name=Wait" json:"Wait"`
+	Since            uint64 `protobuf:"varint,11,opt,name=Since" json:"Since"`
+	Recursive        bool   `protobuf:"varint,12,opt,name=Recursive" json:"Recursive"`
+	Sorted           bool   `protobuf:"varint,13,opt,name=Sorted" json:"Sorted"`
+	Quorum           bool   `protobuf:"varint,14,opt,name=Quorum" json:"Quorum"`
+	Time             int64  `protobuf:"varint,15,opt,name=Time" json:"Time"`
+	Stream           bool   `protobuf:"varint,16,opt,name=Stream" json:"Stream"`
+	Refresh          *bool  `protobuf:"varint,17,opt,name=Refresh" json:"Refresh,omitempty"`
+	XXX_unrecognized []byte `json:"-"`
+}
+
+func (m *Request) Reset()         { *m = Request{} }
+func (m *Request) String() string { return proto.CompactTextString(m) }
+func (*Request) ProtoMessage()    {}
+
+type Metadata struct {
+	NodeID           uint64 `protobuf:"varint,1,opt,name=NodeID" json:"NodeID"`
+	ClusterID        uint64 `protobuf:"varint,2,opt,name=ClusterID" json:"ClusterID"`
+	XXX_unrecognized []byte `json:"-"`
+}
+
+func (m *Metadata) Reset()         { *m = Metadata{} }
+func (m *Metadata) String() string { return proto.CompactTextString(m) }
+func (*Metadata) ProtoMessage()    {}
+
+func init() {
+	proto.RegisterType((*Request)(nil), "etcdserverpb.Request")
+	proto.RegisterType((*Metadata)(nil), "etcdserverpb.Metadata")
+}
+func (m *Request) Marshal() (data []byte, err error) {
+	size := m.Size()
+	data = make([]byte, size)
+	n, err := m.MarshalTo(data)
+	if err != nil {
+		return nil, err
+	}
+	return data[:n], nil
+}
+
+func (m *Request) MarshalTo(data []byte) (int, error) {
+	var i int
+	_ = i
+	var l int
+	_ = l
+	data[i] = 0x8
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(m.ID))
+	data[i] = 0x12
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(len(m.Method)))
+	i += copy(data[i:], m.Method)
+	data[i] = 0x1a
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(len(m.Path)))
+	i += copy(data[i:], m.Path)
+	data[i] = 0x22
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(len(m.Val)))
+	i += copy(data[i:], m.Val)
+	data[i] = 0x28
+	i++
+	if m.Dir {
+		data[i] = 1
+	} else {
+		data[i] = 0
+	}
+	i++
+	data[i] = 0x32
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(len(m.PrevValue)))
+	i += copy(data[i:], m.PrevValue)
+	data[i] = 0x38
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(m.PrevIndex))
+	if m.PrevExist != nil {
+		data[i] = 0x40
+		i++
+		if *m.PrevExist {
+			data[i] = 1
+		} else {
+			data[i] = 0
+		}
+		i++
+	}
+	data[i] = 0x48
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(m.Expiration))
+	data[i] = 0x50
+	i++
+	if m.Wait {
+		data[i] = 1
+	} else {
+		data[i] = 0
+	}
+	i++
+	data[i] = 0x58
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(m.Since))
+	data[i] = 0x60
+	i++
+	if m.Recursive {
+		data[i] = 1
+	} else {
+		data[i] = 0
+	}
+	i++
+	data[i] = 0x68
+	i++
+	if m.Sorted {
+		data[i] = 1
+	} else {
+		data[i] = 0
+	}
+	i++
+	data[i] = 0x70
+	i++
+	if m.Quorum {
+		data[i] = 1
+	} else {
+		data[i] = 0
+	}
+	i++
+	data[i] = 0x78
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(m.Time))
+	data[i] = 0x80
+	i++
+	data[i] = 0x1
+	i++
+	if m.Stream {
+		data[i] = 1
+	} else {
+		data[i] = 0
+	}
+	i++
+	if m.Refresh != nil {
+		data[i] = 0x88
+		i++
+		data[i] = 0x1
+		i++
+		if *m.Refresh {
+			data[i] = 1
+		} else {
+			data[i] = 0
+		}
+		i++
+	}
+	if m.XXX_unrecognized != nil {
+		i += copy(data[i:], m.XXX_unrecognized)
+	}
+	return i, nil
+}
+
+func (m *Metadata) Marshal() (data []byte, err error) {
+	size := m.Size()
+	data = make([]byte, size)
+	n, err := m.MarshalTo(data)
+	if err != nil {
+		return nil, err
+	}
+	return data[:n], nil
+}
+
+func (m *Metadata) MarshalTo(data []byte) (int, error) {
+	var i int
+	_ = i
+	var l int
+	_ = l
+	data[i] = 0x8
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(m.NodeID))
+	data[i] = 0x10
+	i++
+	i = encodeVarintEtcdserver(data, i, uint64(m.ClusterID))
+	if m.XXX_unrecognized != nil {
+		i += copy(data[i:], m.XXX_unrecognized)
+	}
+	return i, nil
+}
+
+func encodeFixed64Etcdserver(data []byte, offset int, v uint64) int {
+	data[offset] = uint8(v)
+	data[offset+1] = uint8(v >> 8)
+	data[offset+2] = uint8(v >> 16)
+	data[offset+3] = uint8(v >> 24)
+	data[offset+4] = uint8(v >> 32)
+	data[offset+5] = uint8(v >> 40)
+	data[offset+6] = uint8(v >> 48)
+	data[offset+7] = uint8(v >> 56)
+	return offset + 8
+}
+func encodeFixed32Etcdserver(data []byte, offset int, v uint32) int {
+	data[offset] = uint8(v)
+	data[offset+1] = uint8(v >> 8)
+	data[offset+2] = uint8(v >> 16)
+	data[offset+3] = uint8(v >> 24)
+	return offset + 4
+}
+func encodeVarintEtcdserver(data []byte, offset int, v uint64) int {
+	for v >= 1<<7 {
+		data[offset] = uint8(v&0x7f | 0x80)
+		v >>= 7
+		offset++
+	}
+	data[offset] = uint8(v)
+	return offset + 1
+}
+func (m *Request) Size() (n int) {
+	var l int
+	_ = l
+	n += 1 + sovEtcdserver(uint64(m.ID))
+	l = len(m.Method)
+	n += 1 + l + sovEtcdserver(uint64(l))
+	l = len(m.Path)
+	n += 1 + l + sovEtcdserver(uint64(l))
+	l = len(m.Val)
+	n += 1 + l + sovEtcdserver(uint64(l))
+	n += 2
+	l = len(m.PrevValue)
+	n += 1 + l + sovEtcdserver(uint64(l))
+	n += 1 + sovEtcdserver(uint64(m.PrevIndex))
+	if m.PrevExist != nil {
+		n += 2
+	}
+	n += 1 + sovEtcdserver(uint64(m.Expiration))
+	n += 2
+	n += 1 + sovEtcdserver(uint64(m.Since))
+	n += 2
+	n += 2
+	n += 2
+	n += 1 + sovEtcdserver(uint64(m.Time))
+	n += 3
+	if m.Refresh != nil {
+		n += 3
+	}
+	if m.XXX_unrecognized != nil {
+		n += len(m.XXX_unrecognized)
+	}
+	return n
+}
+
+func (m *Metadata) Size() (n int) {
+	var l int
+	_ = l
+	n += 1 + sovEtcdserver(uint64(m.NodeID))
+	n += 1 + sovEtcdserver(uint64(m.ClusterID))
+	if m.XXX_unrecognized != nil {
+		n += len(m.XXX_unrecognized)
+	}
+	return n
+}
+
+func sovEtcdserver(x uint64) (n int) {
+	for {
+		n++
+		x >>= 7
+		if x == 0 {
+			break
+		}
+	}
+	return n
+}
+func sozEtcdserver(x uint64) (n int) {
+	return sovEtcdserver(uint64((x << 1) ^ uint64((int64(x) >> 63))))
+}
+func (m *Request) Unmarshal(data []byte) error {
+	l := len(data)
+	iNdEx := 0
+	for iNdEx < l {
+		preIndex := iNdEx
+		var wire uint64
+		for shift := uint(0); ; shift += 7 {
+			if shift >= 64 {
+				return ErrIntOverflowEtcdserver
+			}
+			if iNdEx >= l {
+				return io.ErrUnexpectedEOF
+			}
+			b := data[iNdEx]
+			iNdEx++
+			wire |= (uint64(b) & 0x7F) << shift
+			if b < 0x80 {
+				break
+			}
+		}
+		fieldNum := int32(wire >> 3)
+		wireType := int(wire & 0x7)
+		if wireType == 4 {
+			return fmt.Errorf("proto: Request: wiretype end group for non-group")
+		}
+		if fieldNum <= 0 {
+			return fmt.Errorf("proto: Request: illegal tag %d (wire type %d)", fieldNum, wire)
+		}
+		switch fieldNum {
+		case 1:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field ID", wireType)
+			}
+			m.ID = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				m.ID |= (uint64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		case 2:
+			if wireType != 2 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Method", wireType)
+			}
+			var stringLen uint64
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				stringLen |= (uint64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			intStringLen := int(stringLen)
+			if intStringLen < 0 {
+				return ErrInvalidLengthEtcdserver
+			}
+			postIndex := iNdEx + intStringLen
+			if postIndex > l {
+				return io.ErrUnexpectedEOF
+			}
+			m.Method = string(data[iNdEx:postIndex])
+			iNdEx = postIndex
+		case 3:
+			if wireType != 2 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Path", wireType)
+			}
+			var stringLen uint64
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				stringLen |= (uint64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			intStringLen := int(stringLen)
+			if intStringLen < 0 {
+				return ErrInvalidLengthEtcdserver
+			}
+			postIndex := iNdEx + intStringLen
+			if postIndex > l {
+				return io.ErrUnexpectedEOF
+			}
+			m.Path = string(data[iNdEx:postIndex])
+			iNdEx = postIndex
+		case 4:
+			if wireType != 2 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Val", wireType)
+			}
+			var stringLen uint64
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				stringLen |= (uint64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			intStringLen := int(stringLen)
+			if intStringLen < 0 {
+				return ErrInvalidLengthEtcdserver
+			}
+			postIndex := iNdEx + intStringLen
+			if postIndex > l {
+				return io.ErrUnexpectedEOF
+			}
+			m.Val = string(data[iNdEx:postIndex])
+			iNdEx = postIndex
+		case 5:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Dir", wireType)
+			}
+			var v int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				v |= (int(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			m.Dir = bool(v != 0)
+		case 6:
+			if wireType != 2 {
+				return fmt.Errorf("proto: wrong wireType = %d for field PrevValue", wireType)
+			}
+			var stringLen uint64
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				stringLen |= (uint64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			intStringLen := int(stringLen)
+			if intStringLen < 0 {
+				return ErrInvalidLengthEtcdserver
+			}
+			postIndex := iNdEx + intStringLen
+			if postIndex > l {
+				return io.ErrUnexpectedEOF
+			}
+			m.PrevValue = string(data[iNdEx:postIndex])
+			iNdEx = postIndex
+		case 7:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field PrevIndex", wireType)
+			}
+			m.PrevIndex = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				m.PrevIndex |= (uint64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		case 8:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field PrevExist", wireType)
+			}
+			var v int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				v |= (int(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			b := bool(v != 0)
+			m.PrevExist = &b
+		case 9:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Expiration", wireType)
+			}
+			m.Expiration = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				m.Expiration |= (int64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		case 10:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Wait", wireType)
+			}
+			var v int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				v |= (int(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			m.Wait = bool(v != 0)
+		case 11:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Since", wireType)
+			}
+			m.Since = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				m.Since |= (uint64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		case 12:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Recursive", wireType)
+			}
+			var v int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				v |= (int(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			m.Recursive = bool(v != 0)
+		case 13:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Sorted", wireType)
+			}
+			var v int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				v |= (int(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			m.Sorted = bool(v != 0)
+		case 14:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Quorum", wireType)
+			}
+			var v int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				v |= (int(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			m.Quorum = bool(v != 0)
+		case 15:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Time", wireType)
+			}
+			m.Time = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				m.Time |= (int64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		case 16:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Stream", wireType)
+			}
+			var v int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				v |= (int(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			m.Stream = bool(v != 0)
+		case 17:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field Refresh", wireType)
+			}
+			var v int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				v |= (int(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			b := bool(v != 0)
+			m.Refresh = &b
+		default:
+			iNdEx = preIndex
+			skippy, err := skipEtcdserver(data[iNdEx:])
+			if err != nil {
+				return err
+			}
+			if skippy < 0 {
+				return ErrInvalidLengthEtcdserver
+			}
+			if (iNdEx + skippy) > l {
+				return io.ErrUnexpectedEOF
+			}
+			m.XXX_unrecognized = append(m.XXX_unrecognized, data[iNdEx:iNdEx+skippy]...)
+			iNdEx += skippy
+		}
+	}
+
+	if iNdEx > l {
+		return io.ErrUnexpectedEOF
+	}
+	return nil
+}
+func (m *Metadata) Unmarshal(data []byte) error {
+	l := len(data)
+	iNdEx := 0
+	for iNdEx < l {
+		preIndex := iNdEx
+		var wire uint64
+		for shift := uint(0); ; shift += 7 {
+			if shift >= 64 {
+				return ErrIntOverflowEtcdserver
+			}
+			if iNdEx >= l {
+				return io.ErrUnexpectedEOF
+			}
+			b := data[iNdEx]
+			iNdEx++
+			wire |= (uint64(b) & 0x7F) << shift
+			if b < 0x80 {
+				break
+			}
+		}
+		fieldNum := int32(wire >> 3)
+		wireType := int(wire & 0x7)
+		if wireType == 4 {
+			return fmt.Errorf("proto: Metadata: wiretype end group for non-group")
+		}
+		if fieldNum <= 0 {
+			return fmt.Errorf("proto: Metadata: illegal tag %d (wire type %d)", fieldNum, wire)
+		}
+		switch fieldNum {
+		case 1:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field NodeID", wireType)
+			}
+			m.NodeID = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				m.NodeID |= (uint64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		case 2:
+			if wireType != 0 {
+				return fmt.Errorf("proto: wrong wireType = %d for field ClusterID", wireType)
+			}
+			m.ClusterID = 0
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				m.ClusterID |= (uint64(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+		default:
+			iNdEx = preIndex
+			skippy, err := skipEtcdserver(data[iNdEx:])
+			if err != nil {
+				return err
+			}
+			if skippy < 0 {
+				return ErrInvalidLengthEtcdserver
+			}
+			if (iNdEx + skippy) > l {
+				return io.ErrUnexpectedEOF
+			}
+			m.XXX_unrecognized = append(m.XXX_unrecognized, data[iNdEx:iNdEx+skippy]...)
+			iNdEx += skippy
+		}
+	}
+
+	if iNdEx > l {
+		return io.ErrUnexpectedEOF
+	}
+	return nil
+}
+func skipEtcdserver(data []byte) (n int, err error) {
+	l := len(data)
+	iNdEx := 0
+	for iNdEx < l {
+		var wire uint64
+		for shift := uint(0); ; shift += 7 {
+			if shift >= 64 {
+				return 0, ErrIntOverflowEtcdserver
+			}
+			if iNdEx >= l {
+				return 0, io.ErrUnexpectedEOF
+			}
+			b := data[iNdEx]
+			iNdEx++
+			wire |= (uint64(b) & 0x7F) << shift
+			if b < 0x80 {
+				break
+			}
+		}
+		wireType := int(wire & 0x7)
+		switch wireType {
+		case 0:
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return 0, ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return 0, io.ErrUnexpectedEOF
+				}
+				iNdEx++
+				if data[iNdEx-1] < 0x80 {
+					break
+				}
+			}
+			return iNdEx, nil
+		case 1:
+			iNdEx += 8
+			return iNdEx, nil
+		case 2:
+			var length int
+			for shift := uint(0); ; shift += 7 {
+				if shift >= 64 {
+					return 0, ErrIntOverflowEtcdserver
+				}
+				if iNdEx >= l {
+					return 0, io.ErrUnexpectedEOF
+				}
+				b := data[iNdEx]
+				iNdEx++
+				length |= (int(b) & 0x7F) << shift
+				if b < 0x80 {
+					break
+				}
+			}
+			iNdEx += length
+			if length < 0 {
+				return 0, ErrInvalidLengthEtcdserver
+			}
+			return iNdEx, nil
+		case 3:
+			for {
+				var innerWire uint64
+				var start int = iNdEx
+				for shift := uint(0); ; shift += 7 {
+					if shift >= 64 {
+						return 0, ErrIntOverflowEtcdserver
+					}
+					if iNdEx >= l {
+						return 0, io.ErrUnexpectedEOF
+					}
+					b := data[iNdEx]
+					iNdEx++
+					innerWire |= (uint64(b) & 0x7F) << shift
+					if b < 0x80 {
+						break
+					}
+				}
+				innerWireType := int(innerWire & 0x7)
+				if innerWireType == 4 {
+					break
+				}
+				next, err := skipEtcdserver(data[start:])
+				if err != nil {
+					return 0, err
+				}
+				iNdEx = start + next
+			}
+			return iNdEx, nil
+		case 4:
+			return iNdEx, nil
+		case 5:
+			iNdEx += 4
+			return iNdEx, nil
+		default:
+			return 0, fmt.Errorf("proto: illegal wireType %d", wireType)
+		}
+	}
+	panic("unreachable")
+}
+
+var (
+	ErrInvalidLengthEtcdserver = fmt.Errorf("proto: negative length found during unmarshaling")
+	ErrIntOverflowEtcdserver   = fmt.Errorf("proto: integer overflow")
+)

vendor/github.com/coreos/etcd/etcdserver/etcdserverpb/etcdserver.proto

@@ -0,0 +1,34 @@
+syntax = "proto2";
+package etcdserverpb;
+
+import "gogoproto/gogo.proto";
+
+option (gogoproto.marshaler_all) = true;
+option (gogoproto.sizer_all) = true;
+option (gogoproto.unmarshaler_all) = true;
+option (gogoproto.goproto_getters_all) = false;
+
+message Request {
+	optional uint64 ID         =  1 [(gogoproto.nullable) = false];
+	optional string Method     =  2 [(gogoproto.nullable) = false];
+	optional string Path       =  3 [(gogoproto.nullable) = false];
+	optional string Val        =  4 [(gogoproto.nullable) = false];
+	optional bool   Dir        =  5 [(gogoproto.nullable) = false];
+	optional string PrevValue  =  6 [(gogoproto.nullable) = false];
+	optional uint64 PrevIndex  =  7 [(gogoproto.nullable) = false];
+	optional bool   PrevExist  =  8 [(gogoproto.nullable) = true];
+	optional int64  Expiration =  9 [(gogoproto.nullable) = false];
+	optional bool   Wait       = 10 [(gogoproto.nullable) = false];
+	optional uint64 Since      = 11 [(gogoproto.nullable) = false];
+	optional bool   Recursive  = 12 [(gogoproto.nullable) = false];
+	optional bool   Sorted     = 13 [(gogoproto.nullable) = false];
+	optional bool   Quorum     = 14 [(gogoproto.nullable) = false];
+	optional int64  Time       = 15 [(gogoproto.nullable) = false];
+	optional bool   Stream     = 16 [(gogoproto.nullable) = false];
+	optional bool   Refresh    = 17 [(gogoproto.nullable) = true];
+}
+
+message Metadata {
+	optional uint64 NodeID    = 1 [(gogoproto.nullable) = false];
+	optional uint64 ClusterID = 2 [(gogoproto.nullable) = false];
+}

vendor/github.com/coreos/etcd/etcdserver/etcdserverpb/raft_internal.proto

@@ -0,0 +1,38 @@
+syntax = "proto3";
+package etcdserverpb;
+
+import "gogoproto/gogo.proto";
+import "etcdserver.proto";
+import "rpc.proto";
+
+option (gogoproto.marshaler_all) = true;
+option (gogoproto.sizer_all) = true;
+option (gogoproto.unmarshaler_all) = true;
+option (gogoproto.goproto_getters_all) = false;
+
+// An InternalRaftRequest is the union of all requests which can be
+// sent via raft.
+message InternalRaftRequest {
+  uint64 ID = 1;
+  Request v2 = 2;
+
+  RangeRequest range = 3;
+  PutRequest put = 4;
+  DeleteRangeRequest delete_range = 5;
+  TxnRequest txn = 6;
+  CompactionRequest compaction = 7;
+
+  LeaseGrantRequest lease_grant = 8;
+  LeaseRevokeRequest lease_revoke = 9;
+
+  AuthEnableRequest auth_enable = 10;
+  AuthUserAddRequest auth_user_add = 11;
+  AuthUserDeleteRequest auth_user_delete = 12;
+  AuthUserChangePasswordRequest auth_user_change_password = 13;
+  AuthRoleAddRequest auth_role_add = 14;
+
+  AlarmRequest alarm = 15;
+}
+
+message EmptyResponse {
+}

vendor/github.com/coreos/etcd/etcdserver/etcdserverpb/rpc.proto

@@ -0,0 +1,587 @@
+syntax = "proto3";
+package etcdserverpb;
+
+import "gogoproto/gogo.proto";
+import "etcd/storage/storagepb/kv.proto";
+
+option (gogoproto.marshaler_all) = true;
+option (gogoproto.unmarshaler_all) = true;
+
+service KV {
+  // Range gets the keys in the range from the store.
+  rpc Range(RangeRequest) returns (RangeResponse) {}
+
+  // Put puts the given key into the store.
+  // A put request increases the revision of the store,
+  // and generates one event in the event history.
+  rpc Put(PutRequest) returns (PutResponse) {}
+
+  // Delete deletes the given range from the store.
+  // A delete request increase the revision of the store,
+  // and generates one event in the event history.
+  rpc DeleteRange(DeleteRangeRequest) returns (DeleteRangeResponse) {}
+
+  // Txn processes all the requests in one transaction.
+  // A txn request increases the revision of the store,
+  // and generates events with the same revision in the event history.
+  // It is not allowed to modify the same key several times within one txn.
+  rpc Txn(TxnRequest) returns (TxnResponse) {}
+
+  // Compact compacts the event history in etcd. User should compact the
+  // event history periodically, or it will grow infinitely.
+  rpc Compact(CompactionRequest) returns (CompactionResponse) {}
+}
+
+service Watch {
+  // Watch watches the events happening or happened. Both input and output
+  // are stream. One watch rpc can watch for multiple keys or prefixs and
+  // get a stream of events. The whole events history can be watched unless
+  // compacted.
+  rpc Watch(stream WatchRequest) returns (stream WatchResponse) {}
+}
+
+service Lease {
+  // LeaseGrant creates a lease. A lease has a TTL. The lease will expire if the
+  // server does not receive a keepAlive within TTL from the lease holder.
+  // All keys attached to the lease will be expired and deleted if the lease expires.
+  // The key expiration generates an event in event history.
+  rpc LeaseGrant(LeaseGrantRequest) returns (LeaseGrantResponse) {}
+
+  // LeaseRevoke revokes a lease. All the key attached to the lease will be expired and deleted.
+  rpc LeaseRevoke(LeaseRevokeRequest) returns (LeaseRevokeResponse) {}
+
+  // KeepAlive keeps the lease alive.
+  rpc LeaseKeepAlive(stream LeaseKeepAliveRequest) returns (stream LeaseKeepAliveResponse) {}
+
+  // TODO(xiangli) List all existing Leases?
+  // TODO(xiangli) Get details information (expirations, leased keys, etc.) of a lease?
+}
+
+service Cluster {
+  // MemberAdd adds a member into the cluster.
+  rpc MemberAdd(MemberAddRequest) returns (MemberAddResponse) {}
+
+  // MemberRemove removes an existing member from the cluster.
+  rpc MemberRemove(MemberRemoveRequest) returns (MemberRemoveResponse) {}
+
+  // MemberUpdate updates the member configuration.
+  rpc MemberUpdate(MemberUpdateRequest) returns (MemberUpdateResponse) {}
+
+  // MemberList lists all the members in the cluster.
+  rpc MemberList(MemberListRequest) returns (MemberListResponse) {}
+}
+
+service Maintenance {
+  // Alarm activates, deactivates, and queries alarms regarding cluster health.
+  rpc Alarm(AlarmRequest) returns (AlarmResponse) {}
+
+  // Status gets the status of the member.
+  rpc Status(StatusRequest) returns (StatusResponse) {}
+
+  rpc Defragment(DefragmentRequest) returns (DefragmentResponse) {}
+
+  // Hash returns the hash of the local KV state for consistency checking purpose.
+  // This is designed for testing; do not use this in production when there
+  // are ongoing transactions.
+  rpc Hash(HashRequest) returns (HashResponse) {}
+}
+
+service Auth {
+  // AuthEnable enables authentication.
+  rpc AuthEnable(AuthEnableRequest) returns (AuthEnableResponse) {}
+
+  // AuthDisable disables authentication.
+  rpc AuthDisable(AuthDisableRequest) returns (AuthDisableResponse) {}
+
+  // Authenticate processes authenticate request.
+  rpc Authenticate(AuthenticateRequest) returns (AuthenticateResponse) {}
+
+  // UserAdd adds a new user.
+  rpc UserAdd(AuthUserAddRequest) returns (AuthUserAddResponse) {}
+
+  // UserGet gets a detailed information of a user or lists entire users.
+  rpc UserGet(AuthUserGetRequest) returns (AuthUserGetResponse) {}
+
+  // UserDelete deletes a specified user.
+  rpc UserDelete(AuthUserDeleteRequest) returns (AuthUserDeleteResponse) {}
+
+  // UserChangePassword changes password of a specified user.
+  rpc UserChangePassword(AuthUserChangePasswordRequest) returns (AuthUserChangePasswordResponse) {}
+
+  // UserGrant grants a role to a specified user.
+  rpc UserGrant(AuthUserGrantRequest) returns (AuthUserGrantResponse) {}
+
+  // UserRevoke revokes a role of specified user.
+  rpc UserRevoke(AuthUserRevokeRequest) returns (AuthUserRevokeResponse) {}
+
+  // RoleAdd adds a new role.
+  rpc RoleAdd(AuthRoleAddRequest) returns (AuthRoleAddResponse) {}
+
+  // RoleGet gets a detailed information of a role or lists entire roles.
+  rpc RoleGet(AuthRoleGetRequest) returns (AuthRoleGetResponse) {}
+
+  // RoleDelete deletes a specified role.
+  rpc RoleDelete(AuthRoleDeleteRequest) returns (AuthRoleDeleteResponse) {}
+
+  // RoleGrant grants a permission of a specified key or range to a specified role.
+  rpc RoleGrant(AuthRoleGrantRequest) returns (AuthRoleGrantResponse) {}
+
+  // RoleRevoke revokes a key or range permission of a specified role.
+  rpc RoleRevoke(AuthRoleRevokeRequest) returns (AuthRoleRevokeResponse) {}
+}
+
+message ResponseHeader {
+  uint64 cluster_id = 1;
+  uint64 member_id = 2;
+  // revision of the store when the request was applied.
+  int64 revision = 3;
+  // term of raft when the request was applied.
+  uint64 raft_term = 4;
+}
+
+message RangeRequest {
+  enum SortOrder {
+	NONE = 0; // default, no sorting
+	ASCEND = 1; // lowest target value first
+	DESCEND = 2; // highest target value first
+  }
+  enum SortTarget {
+	KEY = 0;
+	VERSION = 1;
+	CREATE = 2;
+	MOD = 3;
+	VALUE = 4;
+  }
+
+  // if the range_end is not given, the request returns the key.
+  bytes key = 1;
+  // if the range_end is given, it gets the keys in range [key, range_end)
+  // if range_end is nonempty, otherwise it returns all keys >= key.
+  bytes range_end = 2;
+  // limit the number of keys returned.
+  int64 limit = 3;
+  // range over the store at the given revision.
+  // if revision is less or equal to zero, range over the newest store.
+  // if the revision has been compacted, ErrCompaction will be returned in
+  // response.
+  int64 revision = 4;
+
+  // sort_order is the requested order for returned the results
+  SortOrder sort_order = 5;
+
+  // sort_target is the kv field to use for sorting
+  SortTarget sort_target = 6;
+
+  // range request is linearizable by default. Linearizable requests has a higher
+  // latency and lower throughput than serializable request.
+  // To reduce latency, serializable can be set. If serializable is set, range request 
+  // will be serializable, but not linearizable with other requests.
+  // Serializable range can be served locally without waiting for other nodes in the cluster.
+  bool serializable = 7;
+}
+
+message RangeResponse {
+  ResponseHeader header = 1;
+  repeated storagepb.KeyValue kvs = 2;
+  // more indicates if there are more keys to return in the requested range.
+  bool more = 3;
+}
+
+message PutRequest {
+  bytes key = 1;
+  bytes value = 2;
+  int64 lease = 3;
+}
+
+message PutResponse {
+  ResponseHeader header = 1;
+}
+
+message DeleteRangeRequest {
+  // if the range_end is not given, the request deletes the key.
+  bytes key = 1;
+  // if the range_end is given, it deletes the keys in range [key, range_end).
+  bytes range_end = 2;
+}
+
+message DeleteRangeResponse {
+  ResponseHeader header = 1;
+  // Deleted is the number of keys that got deleted.
+  int64 deleted = 2;
+}
+
+message RequestUnion {
+  oneof request {
+    RangeRequest request_range = 1;
+    PutRequest request_put = 2;
+    DeleteRangeRequest request_delete_range = 3;
+  }
+}
+
+message ResponseUnion {
+  oneof response {
+    RangeResponse response_range = 1;
+    PutResponse response_put = 2;
+    DeleteRangeResponse response_delete_range = 3;
+  }
+}
+
+message Compare {
+  enum CompareResult {
+    EQUAL = 0;
+    GREATER = 1;
+    LESS = 2;
+  }
+  enum CompareTarget {
+    VERSION = 0;
+    CREATE = 1;
+    MOD = 2;
+    VALUE= 3;
+  }
+  CompareResult result = 1;
+  CompareTarget target = 2;
+  // key path
+  bytes key = 3;
+  oneof target_union {
+    // version of the given key
+    int64 version = 4;
+    // create revision of the given key
+    int64 create_revision = 5;
+    // last modified revision of the given key
+    int64 mod_revision = 6;
+    // value of the given key
+    bytes value = 7;
+  }
+}
+
+// If the comparisons succeed, then the success requests will be processed in order,
+// and the response will contain their respective responses in order.
+// If the comparisons fail, then the failure requests will be processed in order,
+// and the response will contain their respective responses in order.
+
+// From google paxosdb paper:
+// Our implementation hinges around a powerful primitive which we call MultiOp. All other database
+// operations except for iteration are implemented as a single call to MultiOp. A MultiOp is applied atomically
+// and consists of three components:
+// 1. A list of tests called guard. Each test in guard checks a single entry in the database. It may check
+// for the absence or presence of a value, or compare with a given value. Two different tests in the guard
+// may apply to the same or different entries in the database. All tests in the guard are applied and
+// MultiOp returns the results. If all tests are true, MultiOp executes t op (see item 2 below), otherwise
+// it executes f op (see item 3 below).
+// 2. A list of database operations called t op. Each operation in the list is either an insert, delete, or
+// lookup operation, and applies to a single database entry. Two different operations in the list may apply
+// to the same or different entries in the database. These operations are executed
+// if guard evaluates to
+// true.
+// 3. A list of database operations called f op. Like t op, but executed if guard evaluates to false.
+message TxnRequest {
+  repeated Compare compare = 1;
+  repeated RequestUnion success = 2;
+  repeated RequestUnion failure = 3;
+}
+
+message TxnResponse {
+  ResponseHeader header = 1;
+  bool succeeded = 2;
+  repeated ResponseUnion responses = 3;
+}
+
+// Compaction compacts the kv store upto the given revision (including).
+// It removes the old versions of a key. It keeps the newest version of
+// the key even if its latest modification revision is smaller than the given
+// revision.
+message CompactionRequest {
+  int64 revision = 1;
+  // physical is set so the RPC will wait until the compaction is physically
+  // applied to the local database such that compacted entries are totally
+  // removed from the backing store.
+  bool physical = 2;
+}
+
+message CompactionResponse {
+  ResponseHeader header = 1;
+}
+
+message HashRequest {
+}
+
+message HashResponse {
+  ResponseHeader header = 1;
+  uint32 hash = 2;
+}
+
+message WatchRequest {
+  oneof request_union {
+    WatchCreateRequest create_request = 1;
+    WatchCancelRequest cancel_request = 2;
+  }
+}
+
+message WatchCreateRequest {
+  // the key to be watched
+  bytes key = 1;
+  // if the range_end is given, keys in [key, range_end) are watched
+  // NOTE: only range_end == prefixEnd(key) is accepted now
+  bytes range_end = 2;
+  // start_revision is an optional revision (including) to watch from. No start_revision is "now".
+  int64 start_revision = 3;
+  // if progress_notify is set, etcd server sends WatchResponse with empty events to the 
+  // created watcher when there are no recent events. It is useful when clients want always to be
+  // able to recover a disconnected watcher from a recent known revision.
+  // etcdsever can decide how long it should send a notification based on current load.
+  bool progress_notify = 4;
+}
+
+message WatchCancelRequest {
+  int64 watch_id = 1;
+}
+
+message WatchResponse {
+  ResponseHeader header = 1;
+  // watch_id is the ID of the watching the response sent to.
+  int64 watch_id = 2;
+  // If the response is for a create watch request, created is set to true.
+  // Client should record the watch_id and prepare for receiving events for
+  // that watching from the same stream.
+  // All events sent to the created watching will attach with the same watch_id.
+  bool created = 3;
+  // If the response is for a cancel watch request, cancel is set to true.
+  // No further events will be sent to the canceled watching.
+  bool canceled = 4;
+  // CompactRevision is set to the minimum index if a watching tries to watch
+  // at a compacted index.
+  //
+  // This happens when creating a watching at a compacted revision or the watching cannot
+  // catch up with the progress of the KV.
+  //
+  // Client should treat the watching as canceled and should not try to create any
+  // watching with same start_revision again.
+  int64 compact_revision  = 5;
+
+  repeated storagepb.Event events = 11;
+}
+
+message LeaseGrantRequest {
+  // advisory ttl in seconds
+  int64 TTL = 1;
+  // requested ID to create; 0 lets lessor choose
+  int64 ID = 2;
+}
+
+message LeaseGrantResponse {
+  ResponseHeader header = 1;
+  int64 ID = 2;
+  // server decided ttl in second
+  int64 TTL = 3;
+  string error = 4;
+}
+
+message LeaseRevokeRequest {
+  int64 ID = 1;
+}
+
+message LeaseRevokeResponse {
+  ResponseHeader header = 1;
+}
+
+message LeaseKeepAliveRequest {
+  int64 ID = 1;
+}
+
+message LeaseKeepAliveResponse {
+  ResponseHeader header = 1;
+  int64 ID = 2;
+  int64 TTL = 3;
+}
+
+message Member {
+  uint64 ID = 1;
+  // If the member is not started, name will be an empty string.
+  string name = 2;
+  bool IsLeader = 3;
+  repeated string peerURLs = 4;
+  // If the member is not started, client_URLs will be an zero length
+  // string array.
+  repeated string clientURLs = 5;
+}
+
+message MemberAddRequest {
+  repeated string peerURLs = 1;
+}
+
+message MemberAddResponse {
+  ResponseHeader header = 1;
+  Member member = 2;
+}
+
+message MemberRemoveRequest {
+  uint64 ID = 1;
+}
+
+message MemberRemoveResponse {
+  ResponseHeader header = 1;
+}
+
+message MemberUpdateRequest {
+  uint64 ID = 1;
+  repeated string peerURLs = 2;
+}
+
+message MemberUpdateResponse{
+  ResponseHeader header = 1;
+}
+
+message MemberListRequest {
+}
+
+message MemberListResponse {
+  ResponseHeader header = 1;
+  repeated Member members = 2;
+}
+
+message DefragmentRequest {
+
+}
+
+message DefragmentResponse {
+  ResponseHeader header = 1;
+}
+
+enum AlarmType {
+	NONE = 0; // default, used to query if any alarm is active
+	NOSPACE = 1;
+}
+
+message AlarmRequest {
+  enum AlarmAction {
+	GET = 0;
+	ACTIVATE = 1;
+	DEACTIVATE = 2;
+  }
+  AlarmAction action = 1;
+  // MemberID is the member raising the alarm request
+  uint64 memberID = 2;
+  AlarmType alarm = 3;
+}
+
+message AlarmMember {
+	uint64	memberID = 1;
+	AlarmType alarm = 2;
+}
+
+message AlarmResponse {
+  ResponseHeader header = 1;
+  repeated AlarmMember alarms = 2;
+}
+
+message StatusRequest {
+}
+
+message StatusResponse {
+  ResponseHeader header = 1;
+  string version = 2;
+}
+
+message AuthEnableRequest {
+}
+
+message AuthDisableRequest {
+}
+
+message AuthenticateRequest {
+}
+
+message AuthUserAddRequest {
+  string name = 1;
+  string password = 2;
+}
+
+message AuthUserGetRequest {
+}
+
+message AuthUserDeleteRequest {
+  string name = 1;
+}
+
+message AuthUserChangePasswordRequest {
+  string name = 1;
+  string password = 2;
+}
+
+message AuthUserGrantRequest {
+}
+
+message AuthUserRevokeRequest {
+}
+
+message AuthRoleAddRequest {
+  string name = 1;
+}
+
+message AuthRoleGetRequest {
+}
+
+message AuthRoleDeleteRequest {
+}
+
+message AuthRoleGrantRequest {
+}
+
+message AuthRoleRevokeRequest {
+}
+
+message AuthEnableResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthDisableResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthenticateResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthUserAddResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthUserGetResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthUserDeleteResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthUserChangePasswordResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthUserGrantResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthUserRevokeResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthRoleAddResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthRoleGetResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthRoleDeleteResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthRoleGrantResponse {
+  ResponseHeader header = 1;
+}
+
+message AuthRoleRevokeResponse {
+  ResponseHeader header = 1;
+}

vendor/github.com/coreos/etcd/etcdserver/membership/cluster.go

@@ -0,0 +1,489 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package membership
+
+import (
+	"bytes"
+	"crypto/sha1"
+	"encoding/binary"
+	"encoding/json"
+	"fmt"
+	"path"
+	"sort"
+	"strings"
+	"sync"
+
+	"github.com/coreos/etcd/pkg/netutil"
+	"github.com/coreos/etcd/pkg/types"
+	"github.com/coreos/etcd/raft"
+	"github.com/coreos/etcd/raft/raftpb"
+	"github.com/coreos/etcd/storage/backend"
+	"github.com/coreos/etcd/store"
+	"github.com/coreos/etcd/version"
+	"github.com/coreos/go-semver/semver"
+)
+
+// RaftCluster is a list of Members that belong to the same raft cluster
+type RaftCluster struct {
+	id    types.ID
+	token string
+
+	store store.Store
+	be    backend.Backend
+
+	sync.Mutex // guards the fields below
+	version    *semver.Version
+	members    map[types.ID]*Member
+	// removed contains the ids of removed members in the cluster.
+	// removed id cannot be reused.
+	removed map[types.ID]bool
+}
+
+func NewClusterFromURLsMap(token string, urlsmap types.URLsMap) (*RaftCluster, error) {
+	c := NewCluster(token)
+	for name, urls := range urlsmap {
+		m := NewMember(name, urls, token, nil)
+		if _, ok := c.members[m.ID]; ok {
+			return nil, fmt.Errorf("member exists with identical ID %v", m)
+		}
+		if uint64(m.ID) == raft.None {
+			return nil, fmt.Errorf("cannot use %x as member id", raft.None)
+		}
+		c.members[m.ID] = m
+	}
+	c.genID()
+	return c, nil
+}
+
+func NewClusterFromMembers(token string, id types.ID, membs []*Member) *RaftCluster {
+	c := NewCluster(token)
+	c.id = id
+	for _, m := range membs {
+		c.members[m.ID] = m
+	}
+	return c
+}
+
+func NewCluster(token string) *RaftCluster {
+	return &RaftCluster{
+		token:   token,
+		members: make(map[types.ID]*Member),
+		removed: make(map[types.ID]bool),
+	}
+}
+
+func (c *RaftCluster) ID() types.ID { return c.id }
+
+func (c *RaftCluster) Members() []*Member {
+	c.Lock()
+	defer c.Unlock()
+	var ms MembersByID
+	for _, m := range c.members {
+		ms = append(ms, m.Clone())
+	}
+	sort.Sort(ms)
+	return []*Member(ms)
+}
+
+func (c *RaftCluster) Member(id types.ID) *Member {
+	c.Lock()
+	defer c.Unlock()
+	return c.members[id].Clone()
+}
+
+// MemberByName returns a Member with the given name if exists.
+// If more than one member has the given name, it will panic.
+func (c *RaftCluster) MemberByName(name string) *Member {
+	c.Lock()
+	defer c.Unlock()
+	var memb *Member
+	for _, m := range c.members {
+		if m.Name == name {
+			if memb != nil {
+				plog.Panicf("two members with the given name %q exist", name)
+			}
+			memb = m
+		}
+	}
+	return memb.Clone()
+}
+
+func (c *RaftCluster) MemberIDs() []types.ID {
+	c.Lock()
+	defer c.Unlock()
+	var ids []types.ID
+	for _, m := range c.members {
+		ids = append(ids, m.ID)
+	}
+	sort.Sort(types.IDSlice(ids))
+	return ids
+}
+
+func (c *RaftCluster) IsIDRemoved(id types.ID) bool {
+	c.Lock()
+	defer c.Unlock()
+	return c.removed[id]
+}
+
+// PeerURLs returns a list of all peer addresses.
+// The returned list is sorted in ascending lexicographical order.
+func (c *RaftCluster) PeerURLs() []string {
+	c.Lock()
+	defer c.Unlock()
+	urls := make([]string, 0)
+	for _, p := range c.members {
+		for _, addr := range p.PeerURLs {
+			urls = append(urls, addr)
+		}
+	}
+	sort.Strings(urls)
+	return urls
+}
+
+// ClientURLs returns a list of all client addresses.
+// The returned list is sorted in ascending lexicographical order.
+func (c *RaftCluster) ClientURLs() []string {
+	c.Lock()
+	defer c.Unlock()
+	urls := make([]string, 0)
+	for _, p := range c.members {
+		for _, url := range p.ClientURLs {
+			urls = append(urls, url)
+		}
+	}
+	sort.Strings(urls)
+	return urls
+}
+
+func (c *RaftCluster) String() string {
+	c.Lock()
+	defer c.Unlock()
+	b := &bytes.Buffer{}
+	fmt.Fprintf(b, "{ClusterID:%s ", c.id)
+	var ms []string
+	for _, m := range c.members {
+		ms = append(ms, fmt.Sprintf("%+v", m))
+	}
+	fmt.Fprintf(b, "Members:[%s] ", strings.Join(ms, " "))
+	var ids []string
+	for id := range c.removed {
+		ids = append(ids, fmt.Sprintf("%s", id))
+	}
+	fmt.Fprintf(b, "RemovedMemberIDs:[%s]}", strings.Join(ids, " "))
+	return b.String()
+}
+
+func (c *RaftCluster) genID() {
+	mIDs := c.MemberIDs()
+	b := make([]byte, 8*len(mIDs))
+	for i, id := range mIDs {
+		binary.BigEndian.PutUint64(b[8*i:], uint64(id))
+	}
+	hash := sha1.Sum(b)
+	c.id = types.ID(binary.BigEndian.Uint64(hash[:8]))
+}
+
+func (c *RaftCluster) SetID(id types.ID) { c.id = id }
+
+func (c *RaftCluster) SetStore(st store.Store) { c.store = st }
+
+func (c *RaftCluster) Recover() {
+	c.Lock()
+	defer c.Unlock()
+
+	c.members, c.removed = membersFromStore(c.store)
+	c.version = clusterVersionFromStore(c.store)
+	mustDetectDowngrade(c.version)
+
+	for _, m := range c.members {
+		plog.Infof("added member %s %v to cluster %s from store", m.ID, m.PeerURLs, c.id)
+	}
+	if c.version != nil {
+		plog.Infof("set the cluster version to %v from store", version.Cluster(c.version.String()))
+	}
+}
+
+// ValidateConfigurationChange takes a proposed ConfChange and
+// ensures that it is still valid.
+func (c *RaftCluster) ValidateConfigurationChange(cc raftpb.ConfChange) error {
+	members, removed := membersFromStore(c.store)
+	id := types.ID(cc.NodeID)
+	if removed[id] {
+		return ErrIDRemoved
+	}
+	switch cc.Type {
+	case raftpb.ConfChangeAddNode:
+		if members[id] != nil {
+			return ErrIDExists
+		}
+		urls := make(map[string]bool)
+		for _, m := range members {
+			for _, u := range m.PeerURLs {
+				urls[u] = true
+			}
+		}
+		m := new(Member)
+		if err := json.Unmarshal(cc.Context, m); err != nil {
+			plog.Panicf("unmarshal member should never fail: %v", err)
+		}
+		for _, u := range m.PeerURLs {
+			if urls[u] {
+				return ErrPeerURLexists
+			}
+		}
+	case raftpb.ConfChangeRemoveNode:
+		if members[id] == nil {
+			return ErrIDNotFound
+		}
+	case raftpb.ConfChangeUpdateNode:
+		if members[id] == nil {
+			return ErrIDNotFound
+		}
+		urls := make(map[string]bool)
+		for _, m := range members {
+			if m.ID == id {
+				continue
+			}
+			for _, u := range m.PeerURLs {
+				urls[u] = true
+			}
+		}
+		m := new(Member)
+		if err := json.Unmarshal(cc.Context, m); err != nil {
+			plog.Panicf("unmarshal member should never fail: %v", err)
+		}
+		for _, u := range m.PeerURLs {
+			if urls[u] {
+				return ErrPeerURLexists
+			}
+		}
+	default:
+		plog.Panicf("ConfChange type should be either AddNode, RemoveNode or UpdateNode")
+	}
+	return nil
+}
+
+// AddMember adds a new Member into the cluster, and saves the given member's
+// raftAttributes into the store. The given member should have empty attributes.
+// A Member with a matching id must not exist.
+func (c *RaftCluster) AddMember(m *Member) {
+	c.Lock()
+	defer c.Unlock()
+	if c.store != nil {
+		mustSaveMemberToStore(c.store, m)
+	}
+	if c.be != nil {
+		mustSaveMemberToBackend(c.be, m)
+	}
+
+	c.members[m.ID] = m
+}
+
+// RemoveMember removes a member from the store.
+// The given id MUST exist, or the function panics.
+func (c *RaftCluster) RemoveMember(id types.ID) {
+	c.Lock()
+	defer c.Unlock()
+	if c.store != nil {
+		mustDeleteMemberFromStore(c.store, id)
+	}
+	if c.be != nil {
+		mustDeleteMemberFromBackend(c.be, id)
+	}
+
+	delete(c.members, id)
+	c.removed[id] = true
+}
+
+func (c *RaftCluster) UpdateAttributes(id types.ID, attr Attributes) bool {
+	c.Lock()
+	defer c.Unlock()
+	if m, ok := c.members[id]; ok {
+		m.Attributes = attr
+		return true
+	}
+	_, ok := c.removed[id]
+	if ok {
+		plog.Warningf("skipped updating attributes of removed member %s", id)
+	} else {
+		plog.Panicf("error updating attributes of unknown member %s", id)
+	}
+	// TODO: update store in this function
+	return false
+}
+
+func (c *RaftCluster) UpdateRaftAttributes(id types.ID, raftAttr RaftAttributes) {
+	c.Lock()
+	defer c.Unlock()
+
+	c.members[id].RaftAttributes = raftAttr
+	if c.store != nil {
+		mustUpdateMemberInStore(c.store, c.members[id])
+	}
+	if c.be != nil {
+		mustSaveMemberToBackend(c.be, c.members[id])
+	}
+}
+
+func (c *RaftCluster) Version() *semver.Version {
+	c.Lock()
+	defer c.Unlock()
+	if c.version == nil {
+		return nil
+	}
+	return semver.Must(semver.NewVersion(c.version.String()))
+}
+
+func (c *RaftCluster) SetVersion(ver *semver.Version) {
+	c.Lock()
+	defer c.Unlock()
+	if c.version != nil {
+		plog.Noticef("updated the cluster version from %v to %v", version.Cluster(c.version.String()), version.Cluster(ver.String()))
+	} else {
+		plog.Noticef("set the initial cluster version to %v", version.Cluster(ver.String()))
+	}
+	c.version = ver
+	mustDetectDowngrade(c.version)
+}
+
+func (c *RaftCluster) IsReadyToAddNewMember() bool {
+	nmembers := 1
+	nstarted := 0
+
+	for _, member := range c.members {
+		if member.IsStarted() {
+			nstarted++
+		}
+		nmembers++
+	}
+
+	if nstarted == 1 && nmembers == 2 {
+		// a case of adding a new node to 1-member cluster for restoring cluster data
+		// https://github.com/coreos/etcd/blob/master/Documentation/admin_guide.md#restoring-the-cluster
+
+		plog.Debugf("The number of started member is 1. This cluster can accept add member request.")
+		return true
+	}
+
+	nquorum := nmembers/2 + 1
+	if nstarted < nquorum {
+		plog.Warningf("Reject add member request: the number of started member (%d) will be less than the quorum number of the cluster (%d)", nstarted, nquorum)
+		return false
+	}
+
+	return true
+}
+
+func (c *RaftCluster) IsReadyToRemoveMember(id uint64) bool {
+	nmembers := 0
+	nstarted := 0
+
+	for _, member := range c.members {
+		if uint64(member.ID) == id {
+			continue
+		}
+
+		if member.IsStarted() {
+			nstarted++
+		}
+		nmembers++
+	}
+
+	nquorum := nmembers/2 + 1
+	if nstarted < nquorum {
+		plog.Warningf("Reject remove member request: the number of started member (%d) will be less than the quorum number of the cluster (%d)", nstarted, nquorum)
+		return false
+	}
+
+	return true
+}
+
+func membersFromStore(st store.Store) (map[types.ID]*Member, map[types.ID]bool) {
+	members := make(map[types.ID]*Member)
+	removed := make(map[types.ID]bool)
+	e, err := st.Get(StoreMembersPrefix, true, true)
+	if err != nil {
+		if isKeyNotFound(err) {
+			return members, removed
+		}
+		plog.Panicf("get storeMembers should never fail: %v", err)
+	}
+	for _, n := range e.Node.Nodes {
+		var m *Member
+		m, err = nodeToMember(n)
+		if err != nil {
+			plog.Panicf("nodeToMember should never fail: %v", err)
+		}
+		members[m.ID] = m
+	}
+
+	e, err = st.Get(storeRemovedMembersPrefix, true, true)
+	if err != nil {
+		if isKeyNotFound(err) {
+			return members, removed
+		}
+		plog.Panicf("get storeRemovedMembers should never fail: %v", err)
+	}
+	for _, n := range e.Node.Nodes {
+		removed[MustParseMemberIDFromKey(n.Key)] = true
+	}
+	return members, removed
+}
+
+func clusterVersionFromStore(st store.Store) *semver.Version {
+	e, err := st.Get(path.Join(storePrefix, "version"), false, false)
+	if err != nil {
+		if isKeyNotFound(err) {
+			return nil
+		}
+		plog.Panicf("unexpected error (%v) when getting cluster version from store", err)
+	}
+	return semver.Must(semver.NewVersion(*e.Node.Value))
+}
+
+// ValidateClusterAndAssignIDs validates the local cluster by matching the PeerURLs
+// with the existing cluster. If the validation succeeds, it assigns the IDs
+// from the existing cluster to the local cluster.
+// If the validation fails, an error will be returned.
+func ValidateClusterAndAssignIDs(local *RaftCluster, existing *RaftCluster) error {
+	ems := existing.Members()
+	lms := local.Members()
+	if len(ems) != len(lms) {
+		return fmt.Errorf("member count is unequal")
+	}
+	sort.Sort(MembersByPeerURLs(ems))
+	sort.Sort(MembersByPeerURLs(lms))
+
+	for i := range ems {
+		if !netutil.URLStringsEqual(ems[i].PeerURLs, lms[i].PeerURLs) {
+			return fmt.Errorf("unmatched member while checking PeerURLs")
+		}
+		lms[i].ID = ems[i].ID
+	}
+	local.members = make(map[types.ID]*Member)
+	for _, m := range lms {
+		local.members[m.ID] = m
+	}
+	return nil
+}
+
+func mustDetectDowngrade(cv *semver.Version) {
+	lv := semver.Must(semver.NewVersion(version.Version))
+	// only keep major.minor version for comparison against cluster version
+	lv = &semver.Version{Major: lv.Major, Minor: lv.Minor}
+	if cv != nil && lv.LessThan(*cv) {
+		plog.Fatalf("cluster cannot be downgraded (current version: %s is lower than determined cluster version: %s).", version.Version, version.Cluster(cv.String()))
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/membership/errors.go

@@ -0,0 +1,33 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package membership
+
+import (
+	"errors"
+
+	etcdErr "github.com/coreos/etcd/error"
+)
+
+var (
+	ErrIDRemoved     = errors.New("membership: ID removed")
+	ErrIDExists      = errors.New("membership: ID exists")
+	ErrIDNotFound    = errors.New("membership: ID not found")
+	ErrPeerURLexists = errors.New("membership: peerURL exists")
+)
+
+func isKeyNotFound(err error) bool {
+	e, ok := err.(*etcdErr.Error)
+	return ok && e.ErrorCode == etcdErr.EcodeKeyNotFound
+}

vendor/github.com/coreos/etcd/etcdserver/membership/member.go

@@ -0,0 +1,124 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package membership
+
+import (
+	"crypto/sha1"
+	"encoding/binary"
+	"fmt"
+	"math/rand"
+	"sort"
+	"time"
+
+	"github.com/coreos/etcd/pkg/types"
+	"github.com/coreos/pkg/capnslog"
+)
+
+var (
+	plog = capnslog.NewPackageLogger("github.com/coreos/etcd/etcdserver", "membership")
+)
+
+// RaftAttributes represents the raft related attributes of an etcd member.
+type RaftAttributes struct {
+	// PeerURLs is the list of peers in the raft cluster.
+	// TODO(philips): ensure these are URLs
+	PeerURLs []string `json:"peerURLs"`
+}
+
+// Attributes represents all the non-raft related attributes of an etcd member.
+type Attributes struct {
+	Name       string   `json:"name,omitempty"`
+	ClientURLs []string `json:"clientURLs,omitempty"`
+}
+
+type Member struct {
+	ID types.ID `json:"id"`
+	RaftAttributes
+	Attributes
+}
+
+// NewMember creates a Member without an ID and generates one based on the
+// cluster name, peer URLs, and time. This is used for bootstrapping/adding new member.
+func NewMember(name string, peerURLs types.URLs, clusterName string, now *time.Time) *Member {
+	m := &Member{
+		RaftAttributes: RaftAttributes{PeerURLs: peerURLs.StringSlice()},
+		Attributes:     Attributes{Name: name},
+	}
+
+	var b []byte
+	sort.Strings(m.PeerURLs)
+	for _, p := range m.PeerURLs {
+		b = append(b, []byte(p)...)
+	}
+
+	b = append(b, []byte(clusterName)...)
+	if now != nil {
+		b = append(b, []byte(fmt.Sprintf("%d", now.Unix()))...)
+	}
+
+	hash := sha1.Sum(b)
+	m.ID = types.ID(binary.BigEndian.Uint64(hash[:8]))
+	return m
+}
+
+// PickPeerURL chooses a random address from a given Member's PeerURLs.
+// It will panic if there is no PeerURLs available in Member.
+func (m *Member) PickPeerURL() string {
+	if len(m.PeerURLs) == 0 {
+		plog.Panicf("member should always have some peer url")
+	}
+	return m.PeerURLs[rand.Intn(len(m.PeerURLs))]
+}
+
+func (m *Member) Clone() *Member {
+	if m == nil {
+		return nil
+	}
+	mm := &Member{
+		ID: m.ID,
+		Attributes: Attributes{
+			Name: m.Name,
+		},
+	}
+	if m.PeerURLs != nil {
+		mm.PeerURLs = make([]string, len(m.PeerURLs))
+		copy(mm.PeerURLs, m.PeerURLs)
+	}
+	if m.ClientURLs != nil {
+		mm.ClientURLs = make([]string, len(m.ClientURLs))
+		copy(mm.ClientURLs, m.ClientURLs)
+	}
+	return mm
+}
+
+func (m *Member) IsStarted() bool {
+	return len(m.Name) != 0
+}
+
+// MembersByID implements sort by ID interface
+type MembersByID []*Member
+
+func (ms MembersByID) Len() int           { return len(ms) }
+func (ms MembersByID) Less(i, j int) bool { return ms[i].ID < ms[j].ID }
+func (ms MembersByID) Swap(i, j int)      { ms[i], ms[j] = ms[j], ms[i] }
+
+// MembersByPeerURLs implements sort by peer urls interface
+type MembersByPeerURLs []*Member
+
+func (ms MembersByPeerURLs) Len() int { return len(ms) }
+func (ms MembersByPeerURLs) Less(i, j int) bool {
+	return ms[i].PeerURLs[0] < ms[j].PeerURLs[0]
+}
+func (ms MembersByPeerURLs) Swap(i, j int) { ms[i], ms[j] = ms[j], ms[i] }

vendor/github.com/coreos/etcd/etcdserver/membership/store.go

@@ -0,0 +1,149 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package membership
+
+import (
+	"encoding/json"
+	"fmt"
+	"path"
+
+	"github.com/coreos/etcd/pkg/types"
+	"github.com/coreos/etcd/storage/backend"
+	"github.com/coreos/etcd/store"
+)
+
+const (
+	// TODO: make this private after moving all membership storage logic
+	// from etcdserver pkg
+	AttributesSuffix     = "attributes"
+	raftAttributesSuffix = "raftAttributes"
+
+	// the prefix for stroing membership related information in store provided by store pkg.
+	storePrefix = "/0"
+)
+
+var (
+	membersBucketName        = []byte("members")
+	membersRemovedBuckedName = []byte("members_removed")
+
+	StoreMembersPrefix        = path.Join(storePrefix, "members")
+	storeRemovedMembersPrefix = path.Join(storePrefix, "removed_members")
+)
+
+func mustSaveMemberToBackend(be backend.Backend, m *Member) {
+	mkey := backendMemberKey(m.ID)
+	mvalue, err := json.Marshal(m.RaftAttributes)
+	if err != nil {
+		plog.Panicf("marshal raftAttributes should never fail: %v", err)
+	}
+
+	tx := be.BatchTx()
+	tx.Lock()
+	tx.UnsafePut(membersBucketName, mkey, mvalue)
+	tx.Unlock()
+}
+
+func mustDeleteMemberFromBackend(be backend.Backend, id types.ID) {
+	mkey := backendMemberKey(id)
+
+	tx := be.BatchTx()
+	tx.Lock()
+	tx.UnsafeDelete(membersBucketName, mkey)
+	tx.UnsafePut(membersRemovedBuckedName, mkey, []byte("removed"))
+	tx.Unlock()
+}
+
+func mustSaveMemberToStore(s store.Store, m *Member) {
+	b, err := json.Marshal(m.RaftAttributes)
+	if err != nil {
+		plog.Panicf("marshal raftAttributes should never fail: %v", err)
+	}
+	p := path.Join(MemberStoreKey(m.ID), raftAttributesSuffix)
+	if _, err := s.Create(p, false, string(b), false, store.TTLOptionSet{ExpireTime: store.Permanent}); err != nil {
+		plog.Panicf("create raftAttributes should never fail: %v", err)
+	}
+}
+
+func mustDeleteMemberFromStore(s store.Store, id types.ID) {
+	if _, err := s.Delete(MemberStoreKey(id), true, true); err != nil {
+		plog.Panicf("delete member should never fail: %v", err)
+	}
+	if _, err := s.Create(RemovedMemberStoreKey(id), false, "", false, store.TTLOptionSet{ExpireTime: store.Permanent}); err != nil {
+		plog.Panicf("create removedMember should never fail: %v", err)
+	}
+}
+
+func mustUpdateMemberInStore(s store.Store, m *Member) {
+	b, err := json.Marshal(m.RaftAttributes)
+	if err != nil {
+		plog.Panicf("marshal raftAttributes should never fail: %v", err)
+	}
+	p := path.Join(MemberStoreKey(m.ID), raftAttributesSuffix)
+	if _, err := s.Update(p, string(b), store.TTLOptionSet{ExpireTime: store.Permanent}); err != nil {
+		plog.Panicf("update raftAttributes should never fail: %v", err)
+	}
+}
+
+// nodeToMember builds member from a key value node.
+// the child nodes of the given node MUST be sorted by key.
+func nodeToMember(n *store.NodeExtern) (*Member, error) {
+	m := &Member{ID: MustParseMemberIDFromKey(n.Key)}
+	attrs := make(map[string][]byte)
+	raftAttrKey := path.Join(n.Key, raftAttributesSuffix)
+	attrKey := path.Join(n.Key, AttributesSuffix)
+	for _, nn := range n.Nodes {
+		if nn.Key != raftAttrKey && nn.Key != attrKey {
+			return nil, fmt.Errorf("unknown key %q", nn.Key)
+		}
+		attrs[nn.Key] = []byte(*nn.Value)
+	}
+	if data := attrs[raftAttrKey]; data != nil {
+		if err := json.Unmarshal(data, &m.RaftAttributes); err != nil {
+			return nil, fmt.Errorf("unmarshal raftAttributes error: %v", err)
+		}
+	} else {
+		return nil, fmt.Errorf("raftAttributes key doesn't exist")
+	}
+	if data := attrs[attrKey]; data != nil {
+		if err := json.Unmarshal(data, &m.Attributes); err != nil {
+			return m, fmt.Errorf("unmarshal attributes error: %v", err)
+		}
+	}
+	return m, nil
+}
+
+func backendMemberKey(id types.ID) []byte {
+	return []byte(path.Join(id.String(), raftAttributesSuffix))
+}
+
+func MemberStoreKey(id types.ID) string {
+	return path.Join(StoreMembersPrefix, id.String())
+}
+
+func MemberAttributesStorePath(id types.ID) string {
+	return path.Join(MemberStoreKey(id), AttributesSuffix)
+}
+
+func MustParseMemberIDFromKey(key string) types.ID {
+	id, err := types.IDFromString(path.Base(key))
+	if err != nil {
+		plog.Panicf("unexpected parse member id error: %v", err)
+	}
+	return id
+}
+
+func RemovedMemberStoreKey(id types.ID) string {
+	return path.Join(storeRemovedMembersPrefix, id.String())
+}

vendor/github.com/coreos/etcd/etcdserver/metrics.go

@@ -0,0 +1,87 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	"time"
+
+	"github.com/coreos/etcd/pkg/runtime"
+	"github.com/prometheus/client_golang/prometheus"
+)
+
+var (
+	// TODO: with label in v3?
+	proposeDurations = prometheus.NewHistogram(prometheus.HistogramOpts{
+		Namespace: "etcd",
+		Subsystem: "server",
+		Name:      "proposal_durations_seconds",
+		Help:      "The latency distributions of committing proposal.",
+		Buckets:   prometheus.ExponentialBuckets(0.001, 2, 14),
+	})
+	proposePending = prometheus.NewGauge(prometheus.GaugeOpts{
+		Namespace: "etcd",
+		Subsystem: "server",
+		Name:      "pending_proposal_total",
+		Help:      "The total number of pending proposals.",
+	})
+	// This is number of proposal failed in client's view.
+	// The proposal might be later got committed in raft.
+	proposeFailed = prometheus.NewCounter(prometheus.CounterOpts{
+		Namespace: "etcd",
+		Subsystem: "server",
+		Name:      "proposal_failed_total",
+		Help:      "The total number of failed proposals.",
+	})
+
+	fileDescriptorUsed = prometheus.NewGauge(prometheus.GaugeOpts{
+		Namespace: "etcd",
+		Subsystem: "server",
+		Name:      "file_descriptors_used_total",
+		Help:      "The total number of file descriptors used.",
+	})
+)
+
+func init() {
+	prometheus.MustRegister(proposeDurations)
+	prometheus.MustRegister(proposePending)
+	prometheus.MustRegister(proposeFailed)
+	prometheus.MustRegister(fileDescriptorUsed)
+}
+
+func monitorFileDescriptor(done <-chan struct{}) {
+	ticker := time.NewTicker(5 * time.Second)
+	defer ticker.Stop()
+	for {
+		used, err := runtime.FDUsage()
+		if err != nil {
+			plog.Errorf("cannot monitor file descriptor usage (%v)", err)
+			return
+		}
+		fileDescriptorUsed.Set(float64(used))
+		limit, err := runtime.FDLimit()
+		if err != nil {
+			plog.Errorf("cannot monitor file descriptor usage (%v)", err)
+			return
+		}
+		if used >= limit/5*4 {
+			plog.Warningf("80%% of the file descriptor limit is used [used = %d, limit = %d]", used, limit)
+		}
+		select {
+		case <-ticker.C:
+		case <-done:
+			return
+		}
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/quota.go

@@ -0,0 +1,114 @@
+// Copyright 2016 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/storage/backend"
+)
+
+// Quota represents an arbitrary quota against arbitrary requests. Each request
+// costs some charge; if there is not enough remaining charge, then there are
+// too few resources available within the quota to apply the request.
+type Quota interface {
+	// Available judges whether the given request fits within the quota.
+	Available(req interface{}) bool
+	// Cost computes the charge against the quota for a given request.
+	Cost(req interface{}) int
+	// Remaining is the amount of charge left for the quota.
+	Remaining() int64
+}
+
+type passthroughQuota struct{}
+
+func (*passthroughQuota) Available(interface{}) bool { return true }
+func (*passthroughQuota) Cost(interface{}) int       { return 0 }
+func (*passthroughQuota) Remaining() int64           { return 1 }
+
+type backendQuota struct {
+	s               *EtcdServer
+	maxBackendBytes int64
+}
+
+const (
+	// leaseOverhead is an estimate for the cost of storing a lease
+	leaseOverhead = 64
+	// kvOverhead is an estimate for the cost of storing a key's metadata
+	kvOverhead = 256
+)
+
+func NewBackendQuota(s *EtcdServer) Quota {
+	if s.cfg.QuotaBackendBytes < 0 {
+		// disable quotas if negative
+		plog.Warningf("disabling backend quota")
+		return &passthroughQuota{}
+	}
+	if s.cfg.QuotaBackendBytes == 0 {
+		// use default size if no quota size given
+		return &backendQuota{s, backend.DefaultQuotaBytes}
+	}
+	if s.cfg.QuotaBackendBytes > backend.MaxQuotaBytes {
+		plog.Warningf("backend quota %v exceeds maximum quota %v; using maximum", s.cfg.QuotaBackendBytes, backend.MaxQuotaBytes)
+		return &backendQuota{s, backend.MaxQuotaBytes}
+	}
+	return &backendQuota{s, s.cfg.QuotaBackendBytes}
+}
+
+func (b *backendQuota) Available(v interface{}) bool {
+	// TODO: maybe optimize backend.Size()
+	return b.s.Backend().Size()+int64(b.Cost(v)) < b.maxBackendBytes
+}
+
+func (b *backendQuota) Cost(v interface{}) int {
+	switch r := v.(type) {
+	case *pb.PutRequest:
+		return costPut(r)
+	case *pb.TxnRequest:
+		return costTxn(r)
+	case *pb.LeaseGrantRequest:
+		return leaseOverhead
+	default:
+		panic("unexpected cost")
+	}
+}
+
+func costPut(r *pb.PutRequest) int { return kvOverhead + len(r.Key) + len(r.Value) }
+
+func costTxnReq(u *pb.RequestUnion) int {
+	r := u.GetRequestPut()
+	if r == nil {
+		return 0
+	}
+	return costPut(r)
+}
+
+func costTxn(r *pb.TxnRequest) int {
+	sizeSuccess := 0
+	for _, u := range r.Success {
+		sizeSuccess += costTxnReq(u)
+	}
+	sizeFailure := 0
+	for _, u := range r.Failure {
+		sizeFailure += costTxnReq(u)
+	}
+	if sizeFailure > sizeSuccess {
+		return sizeFailure
+	}
+	return sizeSuccess
+}
+
+func (b *backendQuota) Remaining() int64 {
+	return b.maxBackendBytes - b.s.Backend().Size()
+}

vendor/github.com/coreos/etcd/etcdserver/raft.go

@@ -0,0 +1,499 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	"encoding/json"
+	"expvar"
+	"os"
+	"sort"
+	"sync"
+	"sync/atomic"
+	"time"
+
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/etcdserver/membership"
+	"github.com/coreos/etcd/pkg/contention"
+	"github.com/coreos/etcd/pkg/pbutil"
+	"github.com/coreos/etcd/pkg/types"
+	"github.com/coreos/etcd/raft"
+	"github.com/coreos/etcd/raft/raftpb"
+	"github.com/coreos/etcd/rafthttp"
+	"github.com/coreos/etcd/wal"
+	"github.com/coreos/etcd/wal/walpb"
+	"github.com/coreos/pkg/capnslog"
+)
+
+const (
+	// Number of entries for slow follower to catch-up after compacting
+	// the raft storage entries.
+	// We expect the follower has a millisecond level latency with the leader.
+	// The max throughput is around 10K. Keep a 5K entries is enough for helping
+	// follower to catch up.
+	numberOfCatchUpEntries = 5000
+
+	// The max throughput of etcd will not exceed 100MB/s (100K * 1KB value).
+	// Assuming the RTT is around 10ms, 1MB max size is large enough.
+	maxSizePerMsg = 1 * 1024 * 1024
+	// Never overflow the rafthttp buffer, which is 4096.
+	// TODO: a better const?
+	maxInflightMsgs = 4096 / 8
+)
+
+var (
+	// protects raftStatus
+	raftStatusMu sync.Mutex
+	// indirection for expvar func interface
+	// expvar panics when publishing duplicate name
+	// expvar does not support remove a registered name
+	// so only register a func that calls raftStatus
+	// and change raftStatus as we need.
+	raftStatus func() raft.Status
+)
+
+func init() {
+	raft.SetLogger(capnslog.NewPackageLogger("github.com/coreos/etcd", "raft"))
+	expvar.Publish("raft.status", expvar.Func(func() interface{} {
+		raftStatusMu.Lock()
+		defer raftStatusMu.Unlock()
+		return raftStatus()
+	}))
+}
+
+type RaftTimer interface {
+	Index() uint64
+	Term() uint64
+}
+
+// apply contains entries, snapshot to be applied. Once
+// an apply is consumed, the entries will be persisted to
+// to raft storage concurrently; the application must read
+// raftDone before assuming the raft messages are stable.
+type apply struct {
+	entries  []raftpb.Entry
+	snapshot raftpb.Snapshot
+	raftDone <-chan struct{} // rx {} after raft has persisted messages
+}
+
+type raftNode struct {
+	// Cache of the latest raft index and raft term the server has seen.
+	// These three unit64 fields must be the first elements to keep 64-bit
+	// alignment for atomic access to the fields.
+	index uint64
+	term  uint64
+	lead  uint64
+
+	mu sync.Mutex
+	// last lead elected time
+	lt time.Time
+
+	raft.Node
+
+	// a chan to send out apply
+	applyc chan apply
+
+	// TODO: remove the etcdserver related logic from raftNode
+	// TODO: add a state machine interface to apply the commit entries
+	// and do snapshot/recover
+	s *EtcdServer
+
+	// utility
+	ticker      <-chan time.Time
+	raftStorage *raft.MemoryStorage
+	storage     Storage
+	// transport specifies the transport to send and receive msgs to members.
+	// Sending messages MUST NOT block. It is okay to drop messages, since
+	// clients should timeout and reissue their messages.
+	// If transport is nil, server will panic.
+	transport rafthttp.Transporter
+
+	td *contention.TimeoutDetector
+
+	stopped chan struct{}
+	done    chan struct{}
+}
+
+// start prepares and starts raftNode in a new goroutine. It is no longer safe
+// to modify the fields after it has been started.
+// TODO: Ideally raftNode should get rid of the passed in server structure.
+func (r *raftNode) start(s *EtcdServer) {
+	r.s = s
+	r.applyc = make(chan apply)
+	r.stopped = make(chan struct{})
+	r.done = make(chan struct{})
+
+	heartbeat := 200 * time.Millisecond
+	if s.cfg != nil {
+		heartbeat = time.Duration(s.cfg.TickMs) * time.Millisecond
+	}
+	// set up contention detectors for raft heartbeat message.
+	// expect to send a heartbeat within 2 heartbeat intervals.
+	r.td = contention.NewTimeoutDetector(2 * heartbeat)
+
+	go func() {
+		var syncC <-chan time.Time
+
+		defer r.onStop()
+		islead := false
+
+		for {
+			select {
+			case <-r.ticker:
+				r.Tick()
+			case rd := <-r.Ready():
+				if rd.SoftState != nil {
+					if lead := atomic.LoadUint64(&r.lead); rd.SoftState.Lead != raft.None && lead != rd.SoftState.Lead {
+						r.mu.Lock()
+						r.lt = time.Now()
+						r.mu.Unlock()
+					}
+					atomic.StoreUint64(&r.lead, rd.SoftState.Lead)
+					if rd.RaftState == raft.StateLeader {
+						islead = true
+						// TODO: raft should send server a notification through chan when
+						// it promotes or demotes instead of modifying server directly.
+						syncC = r.s.SyncTicker
+						if r.s.lessor != nil {
+							r.s.lessor.Promote(r.s.cfg.electionTimeout())
+						}
+						// TODO: remove the nil checking
+						// current test utility does not provide the stats
+						if r.s.stats != nil {
+							r.s.stats.BecomeLeader()
+						}
+						if r.s.compactor != nil {
+							r.s.compactor.Resume()
+						}
+						r.td.Reset()
+					} else {
+						islead = false
+						if r.s.lessor != nil {
+							r.s.lessor.Demote()
+						}
+						if r.s.compactor != nil {
+							r.s.compactor.Pause()
+						}
+						syncC = nil
+					}
+				}
+
+				raftDone := make(chan struct{}, 1)
+				ap := apply{
+					entries:  rd.CommittedEntries,
+					snapshot: rd.Snapshot,
+					raftDone: raftDone,
+				}
+
+				select {
+				case r.applyc <- ap:
+				case <-r.stopped:
+					return
+				}
+
+				// the leader can write to its disk in parallel with replicating to the followers and them
+				// writing to their disks.
+				// For more details, check raft thesis 10.2.1
+				if islead {
+					r.s.send(rd.Messages)
+				}
+
+				if !raft.IsEmptySnap(rd.Snapshot) {
+					if err := r.storage.SaveSnap(rd.Snapshot); err != nil {
+						plog.Fatalf("raft save snapshot error: %v", err)
+					}
+					r.raftStorage.ApplySnapshot(rd.Snapshot)
+					plog.Infof("raft applied incoming snapshot at index %d", rd.Snapshot.Metadata.Index)
+				}
+				if err := r.storage.Save(rd.HardState, rd.Entries); err != nil {
+					plog.Fatalf("raft save state and entries error: %v", err)
+				}
+				r.raftStorage.Append(rd.Entries)
+
+				if !islead {
+					r.s.send(rd.Messages)
+				}
+				raftDone <- struct{}{}
+				r.Advance()
+			case <-syncC:
+				r.s.sync(r.s.cfg.ReqTimeout())
+			case <-r.stopped:
+				return
+			}
+		}
+	}()
+}
+
+func (r *raftNode) apply() chan apply {
+	return r.applyc
+}
+
+func (r *raftNode) leadElectedTime() time.Time {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	return r.lt
+}
+
+func (r *raftNode) stop() {
+	r.stopped <- struct{}{}
+	<-r.done
+}
+
+func (r *raftNode) onStop() {
+	r.Stop()
+	r.transport.Stop()
+	if err := r.storage.Close(); err != nil {
+		plog.Panicf("raft close storage error: %v", err)
+	}
+	close(r.done)
+}
+
+// for testing
+func (r *raftNode) pauseSending() {
+	p := r.transport.(rafthttp.Pausable)
+	p.Pause()
+}
+
+func (r *raftNode) resumeSending() {
+	p := r.transport.(rafthttp.Pausable)
+	p.Resume()
+}
+
+// advanceTicksForElection advances ticks to the node for fast election.
+// This reduces the time to wait for first leader election if bootstrapping the whole
+// cluster, while leaving at least 1 heartbeat for possible existing leader
+// to contact it.
+func advanceTicksForElection(n raft.Node, electionTicks int) {
+	for i := 0; i < electionTicks-1; i++ {
+		n.Tick()
+	}
+}
+
+func startNode(cfg *ServerConfig, cl *membership.RaftCluster, ids []types.ID) (id types.ID, n raft.Node, s *raft.MemoryStorage, w *wal.WAL) {
+	var err error
+	member := cl.MemberByName(cfg.Name)
+	metadata := pbutil.MustMarshal(
+		&pb.Metadata{
+			NodeID:    uint64(member.ID),
+			ClusterID: uint64(cl.ID()),
+		},
+	)
+	if err = os.MkdirAll(cfg.SnapDir(), privateDirMode); err != nil {
+		plog.Fatalf("create snapshot directory error: %v", err)
+	}
+	if w, err = wal.Create(cfg.WALDir(), metadata); err != nil {
+		plog.Fatalf("create wal error: %v", err)
+	}
+	peers := make([]raft.Peer, len(ids))
+	for i, id := range ids {
+		ctx, err := json.Marshal((*cl).Member(id))
+		if err != nil {
+			plog.Panicf("marshal member should never fail: %v", err)
+		}
+		peers[i] = raft.Peer{ID: uint64(id), Context: ctx}
+	}
+	id = member.ID
+	plog.Infof("starting member %s in cluster %s", id, cl.ID())
+	s = raft.NewMemoryStorage()
+	c := &raft.Config{
+		ID:              uint64(id),
+		ElectionTick:    cfg.ElectionTicks,
+		HeartbeatTick:   1,
+		Storage:         s,
+		MaxSizePerMsg:   maxSizePerMsg,
+		MaxInflightMsgs: maxInflightMsgs,
+		CheckQuorum:     true,
+	}
+
+	n = raft.StartNode(c, peers)
+	raftStatusMu.Lock()
+	raftStatus = n.Status
+	raftStatusMu.Unlock()
+	advanceTicksForElection(n, c.ElectionTick)
+	return
+}
+
+func restartNode(cfg *ServerConfig, snapshot *raftpb.Snapshot) (types.ID, *membership.RaftCluster, raft.Node, *raft.MemoryStorage, *wal.WAL) {
+	var walsnap walpb.Snapshot
+	if snapshot != nil {
+		walsnap.Index, walsnap.Term = snapshot.Metadata.Index, snapshot.Metadata.Term
+	}
+	w, id, cid, st, ents := readWAL(cfg.WALDir(), walsnap)
+
+	plog.Infof("restarting member %s in cluster %s at commit index %d", id, cid, st.Commit)
+	cl := membership.NewCluster("")
+	cl.SetID(cid)
+	s := raft.NewMemoryStorage()
+	if snapshot != nil {
+		s.ApplySnapshot(*snapshot)
+	}
+	s.SetHardState(st)
+	s.Append(ents)
+	c := &raft.Config{
+		ID:              uint64(id),
+		ElectionTick:    cfg.ElectionTicks,
+		HeartbeatTick:   1,
+		Storage:         s,
+		MaxSizePerMsg:   maxSizePerMsg,
+		MaxInflightMsgs: maxInflightMsgs,
+		CheckQuorum:     true,
+	}
+
+	n := raft.RestartNode(c)
+	raftStatusMu.Lock()
+	raftStatus = n.Status
+	raftStatusMu.Unlock()
+	advanceTicksForElection(n, c.ElectionTick)
+	return id, cl, n, s, w
+}
+
+func restartAsStandaloneNode(cfg *ServerConfig, snapshot *raftpb.Snapshot) (types.ID, *membership.RaftCluster, raft.Node, *raft.MemoryStorage, *wal.WAL) {
+	var walsnap walpb.Snapshot
+	if snapshot != nil {
+		walsnap.Index, walsnap.Term = snapshot.Metadata.Index, snapshot.Metadata.Term
+	}
+	w, id, cid, st, ents := readWAL(cfg.WALDir(), walsnap)
+
+	// discard the previously uncommitted entries
+	for i, ent := range ents {
+		if ent.Index > st.Commit {
+			plog.Infof("discarding %d uncommitted WAL entries ", len(ents)-i)
+			ents = ents[:i]
+			break
+		}
+	}
+
+	// force append the configuration change entries
+	toAppEnts := createConfigChangeEnts(getIDs(snapshot, ents), uint64(id), st.Term, st.Commit)
+	ents = append(ents, toAppEnts...)
+
+	// force commit newly appended entries
+	err := w.Save(raftpb.HardState{}, toAppEnts)
+	if err != nil {
+		plog.Fatalf("%v", err)
+	}
+	if len(ents) != 0 {
+		st.Commit = ents[len(ents)-1].Index
+	}
+
+	plog.Printf("forcing restart of member %s in cluster %s at commit index %d", id, cid, st.Commit)
+	cl := membership.NewCluster("")
+	cl.SetID(cid)
+	s := raft.NewMemoryStorage()
+	if snapshot != nil {
+		s.ApplySnapshot(*snapshot)
+	}
+	s.SetHardState(st)
+	s.Append(ents)
+	c := &raft.Config{
+		ID:              uint64(id),
+		ElectionTick:    cfg.ElectionTicks,
+		HeartbeatTick:   1,
+		Storage:         s,
+		MaxSizePerMsg:   maxSizePerMsg,
+		MaxInflightMsgs: maxInflightMsgs,
+	}
+	n := raft.RestartNode(c)
+	raftStatus = n.Status
+	return id, cl, n, s, w
+}
+
+// getIDs returns an ordered set of IDs included in the given snapshot and
+// the entries. The given snapshot/entries can contain two kinds of
+// ID-related entry:
+// - ConfChangeAddNode, in which case the contained ID will be added into the set.
+// - ConfChangeRemoveNode, in which case the contained ID will be removed from the set.
+func getIDs(snap *raftpb.Snapshot, ents []raftpb.Entry) []uint64 {
+	ids := make(map[uint64]bool)
+	if snap != nil {
+		for _, id := range snap.Metadata.ConfState.Nodes {
+			ids[id] = true
+		}
+	}
+	for _, e := range ents {
+		if e.Type != raftpb.EntryConfChange {
+			continue
+		}
+		var cc raftpb.ConfChange
+		pbutil.MustUnmarshal(&cc, e.Data)
+		switch cc.Type {
+		case raftpb.ConfChangeAddNode:
+			ids[cc.NodeID] = true
+		case raftpb.ConfChangeRemoveNode:
+			delete(ids, cc.NodeID)
+		case raftpb.ConfChangeUpdateNode:
+			// do nothing
+		default:
+			plog.Panicf("ConfChange Type should be either ConfChangeAddNode or ConfChangeRemoveNode!")
+		}
+	}
+	sids := make(types.Uint64Slice, 0)
+	for id := range ids {
+		sids = append(sids, id)
+	}
+	sort.Sort(sids)
+	return []uint64(sids)
+}
+
+// createConfigChangeEnts creates a series of Raft entries (i.e.
+// EntryConfChange) to remove the set of given IDs from the cluster. The ID
+// `self` is _not_ removed, even if present in the set.
+// If `self` is not inside the given ids, it creates a Raft entry to add a
+// default member with the given `self`.
+func createConfigChangeEnts(ids []uint64, self uint64, term, index uint64) []raftpb.Entry {
+	ents := make([]raftpb.Entry, 0)
+	next := index + 1
+	found := false
+	for _, id := range ids {
+		if id == self {
+			found = true
+			continue
+		}
+		cc := &raftpb.ConfChange{
+			Type:   raftpb.ConfChangeRemoveNode,
+			NodeID: id,
+		}
+		e := raftpb.Entry{
+			Type:  raftpb.EntryConfChange,
+			Data:  pbutil.MustMarshal(cc),
+			Term:  term,
+			Index: next,
+		}
+		ents = append(ents, e)
+		next++
+	}
+	if !found {
+		m := membership.Member{
+			ID:             types.ID(self),
+			RaftAttributes: membership.RaftAttributes{PeerURLs: []string{"http://localhost:7001", "http://localhost:2380"}},
+		}
+		ctx, err := json.Marshal(m)
+		if err != nil {
+			plog.Panicf("marshal member should never fail: %v", err)
+		}
+		cc := &raftpb.ConfChange{
+			Type:    raftpb.ConfChangeAddNode,
+			NodeID:  self,
+			Context: ctx,
+		}
+		e := raftpb.Entry{
+			Type:  raftpb.EntryConfChange,
+			Data:  pbutil.MustMarshal(cc),
+			Term:  term,
+			Index: next,
+		}
+		ents = append(ents, e)
+	}
+	return ents
+}

vendor/github.com/coreos/etcd/etcdserver/snapshot_merge.go

@@ -0,0 +1,71 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	"io"
+	"log"
+
+	"github.com/coreos/etcd/raft/raftpb"
+	"github.com/coreos/etcd/snap"
+	"github.com/coreos/etcd/storage/backend"
+)
+
+// createMergedSnapshotMessage creates a snapshot message that contains: raft status (term, conf),
+// a snapshot of v2 store inside raft.Snapshot as []byte, a snapshot of v3 KV in the top level message
+// as ReadCloser.
+func (s *EtcdServer) createMergedSnapshotMessage(m raftpb.Message, snapi uint64, confState raftpb.ConfState) snap.Message {
+	snapt, err := s.r.raftStorage.Term(snapi)
+	if err != nil {
+		log.Panicf("get term should never fail: %v", err)
+	}
+
+	// get a snapshot of v2 store as []byte
+	clone := s.store.Clone()
+	d, err := clone.SaveNoCopy()
+	if err != nil {
+		plog.Panicf("store save should never fail: %v", err)
+	}
+
+	// get a snapshot of v3 KV as readCloser
+	rc := newSnapshotReaderCloser(s.be.Snapshot())
+
+	// put the []byte snapshot of store into raft snapshot and return the merged snapshot with
+	// KV readCloser snapshot.
+	snapshot := raftpb.Snapshot{
+		Metadata: raftpb.SnapshotMetadata{
+			Index:     snapi,
+			Term:      snapt,
+			ConfState: confState,
+		},
+		Data: d,
+	}
+	m.Snapshot = snapshot
+
+	return *snap.NewMessage(m, rc)
+}
+
+func newSnapshotReaderCloser(snapshot backend.Snapshot) io.ReadCloser {
+	pr, pw := io.Pipe()
+	go func() {
+		n, err := snapshot.WriteTo(pw)
+		if err == nil {
+			plog.Infof("wrote database snapshot out [total bytes: %d]", n)
+		}
+		pw.CloseWithError(err)
+		snapshot.Close()
+	}()
+	return pr
+}

vendor/github.com/coreos/etcd/etcdserver/stats/leader.go

@@ -0,0 +1,123 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package stats
+
+import (
+	"encoding/json"
+	"math"
+	"sync"
+	"time"
+)
+
+// LeaderStats is used by the leader in an etcd cluster, and encapsulates
+// statistics about communication with its followers
+type LeaderStats struct {
+	// Leader is the ID of the leader in the etcd cluster.
+	// TODO(jonboulle): clarify that these are IDs, not names
+	Leader    string                    `json:"leader"`
+	Followers map[string]*FollowerStats `json:"followers"`
+
+	sync.Mutex
+}
+
+// NewLeaderStats generates a new LeaderStats with the given id as leader
+func NewLeaderStats(id string) *LeaderStats {
+	return &LeaderStats{
+		Leader:    id,
+		Followers: make(map[string]*FollowerStats),
+	}
+}
+
+func (ls *LeaderStats) JSON() []byte {
+	ls.Lock()
+	stats := *ls
+	ls.Unlock()
+	b, err := json.Marshal(stats)
+	// TODO(jonboulle): appropriate error handling?
+	if err != nil {
+		plog.Errorf("error marshalling leader stats (%v)", err)
+	}
+	return b
+}
+
+func (ls *LeaderStats) Follower(name string) *FollowerStats {
+	ls.Lock()
+	defer ls.Unlock()
+	fs, ok := ls.Followers[name]
+	if !ok {
+		fs = &FollowerStats{}
+		fs.Latency.Minimum = 1 << 63
+		ls.Followers[name] = fs
+	}
+	return fs
+}
+
+// FollowerStats encapsulates various statistics about a follower in an etcd cluster
+type FollowerStats struct {
+	Latency LatencyStats `json:"latency"`
+	Counts  CountsStats  `json:"counts"`
+
+	sync.Mutex
+}
+
+// LatencyStats encapsulates latency statistics.
+type LatencyStats struct {
+	Current           float64 `json:"current"`
+	Average           float64 `json:"average"`
+	averageSquare     float64
+	StandardDeviation float64 `json:"standardDeviation"`
+	Minimum           float64 `json:"minimum"`
+	Maximum           float64 `json:"maximum"`
+}
+
+// CountsStats encapsulates raft statistics.
+type CountsStats struct {
+	Fail    uint64 `json:"fail"`
+	Success uint64 `json:"success"`
+}
+
+// Succ updates the FollowerStats with a successful send
+func (fs *FollowerStats) Succ(d time.Duration) {
+	fs.Lock()
+	defer fs.Unlock()
+
+	total := float64(fs.Counts.Success) * fs.Latency.Average
+	totalSquare := float64(fs.Counts.Success) * fs.Latency.averageSquare
+
+	fs.Counts.Success++
+
+	fs.Latency.Current = float64(d) / (1000000.0)
+
+	if fs.Latency.Current > fs.Latency.Maximum {
+		fs.Latency.Maximum = fs.Latency.Current
+	}
+
+	if fs.Latency.Current < fs.Latency.Minimum {
+		fs.Latency.Minimum = fs.Latency.Current
+	}
+
+	fs.Latency.Average = (total + fs.Latency.Current) / float64(fs.Counts.Success)
+	fs.Latency.averageSquare = (totalSquare + fs.Latency.Current*fs.Latency.Current) / float64(fs.Counts.Success)
+
+	// sdv = sqrt(avg(x^2) - avg(x)^2)
+	fs.Latency.StandardDeviation = math.Sqrt(fs.Latency.averageSquare - fs.Latency.Average*fs.Latency.Average)
+}
+
+// Fail updates the FollowerStats with an unsuccessful send
+func (fs *FollowerStats) Fail() {
+	fs.Lock()
+	defer fs.Unlock()
+	fs.Counts.Fail++
+}

vendor/github.com/coreos/etcd/etcdserver/stats/queue.go

@@ -0,0 +1,110 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package stats
+
+import (
+	"sync"
+	"time"
+)
+
+const (
+	queueCapacity = 200
+)
+
+// RequestStats represent the stats for a request.
+// It encapsulates the sending time and the size of the request.
+type RequestStats struct {
+	SendingTime time.Time
+	Size        int
+}
+
+type statsQueue struct {
+	items        [queueCapacity]*RequestStats
+	size         int
+	front        int
+	back         int
+	totalReqSize int
+	rwl          sync.RWMutex
+}
+
+func (q *statsQueue) Len() int {
+	return q.size
+}
+
+func (q *statsQueue) ReqSize() int {
+	return q.totalReqSize
+}
+
+// FrontAndBack gets the front and back elements in the queue
+// We must grab front and back together with the protection of the lock
+func (q *statsQueue) frontAndBack() (*RequestStats, *RequestStats) {
+	q.rwl.RLock()
+	defer q.rwl.RUnlock()
+	if q.size != 0 {
+		return q.items[q.front], q.items[q.back]
+	}
+	return nil, nil
+}
+
+// Insert function insert a RequestStats into the queue and update the records
+func (q *statsQueue) Insert(p *RequestStats) {
+	q.rwl.Lock()
+	defer q.rwl.Unlock()
+
+	q.back = (q.back + 1) % queueCapacity
+
+	if q.size == queueCapacity { //dequeue
+		q.totalReqSize -= q.items[q.front].Size
+		q.front = (q.back + 1) % queueCapacity
+	} else {
+		q.size++
+	}
+
+	q.items[q.back] = p
+	q.totalReqSize += q.items[q.back].Size
+
+}
+
+// Rate function returns the package rate and byte rate
+func (q *statsQueue) Rate() (float64, float64) {
+	front, back := q.frontAndBack()
+
+	if front == nil || back == nil {
+		return 0, 0
+	}
+
+	if time.Now().Sub(back.SendingTime) > time.Second {
+		q.Clear()
+		return 0, 0
+	}
+
+	sampleDuration := back.SendingTime.Sub(front.SendingTime)
+
+	pr := float64(q.Len()) / float64(sampleDuration) * float64(time.Second)
+
+	br := float64(q.ReqSize()) / float64(sampleDuration) * float64(time.Second)
+
+	return pr, br
+}
+
+// Clear function clear up the statsQueue
+func (q *statsQueue) Clear() {
+	q.rwl.Lock()
+	defer q.rwl.Unlock()
+	q.back = -1
+	q.front = 0
+	q.size = 0
+	q.totalReqSize = 0
+}

vendor/github.com/coreos/etcd/etcdserver/stats/server.go

@@ -0,0 +1,150 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package stats
+
+import (
+	"encoding/json"
+	"log"
+	"sync"
+	"time"
+
+	"github.com/coreos/etcd/raft"
+)
+
+// ServerStats encapsulates various statistics about an EtcdServer and its
+// communication with other members of the cluster
+type ServerStats struct {
+	Name string `json:"name"`
+	// ID is the raft ID of the node.
+	// TODO(jonboulle): use ID instead of name?
+	ID        string         `json:"id"`
+	State     raft.StateType `json:"state"`
+	StartTime time.Time      `json:"startTime"`
+
+	LeaderInfo struct {
+		Name      string    `json:"leader"`
+		Uptime    string    `json:"uptime"`
+		StartTime time.Time `json:"startTime"`
+	} `json:"leaderInfo"`
+
+	RecvAppendRequestCnt uint64  `json:"recvAppendRequestCnt,"`
+	RecvingPkgRate       float64 `json:"recvPkgRate,omitempty"`
+	RecvingBandwidthRate float64 `json:"recvBandwidthRate,omitempty"`
+
+	SendAppendRequestCnt uint64  `json:"sendAppendRequestCnt"`
+	SendingPkgRate       float64 `json:"sendPkgRate,omitempty"`
+	SendingBandwidthRate float64 `json:"sendBandwidthRate,omitempty"`
+
+	sendRateQueue *statsQueue
+	recvRateQueue *statsQueue
+
+	sync.Mutex
+}
+
+func (ss *ServerStats) JSON() []byte {
+	ss.Lock()
+	stats := *ss
+	ss.Unlock()
+	stats.LeaderInfo.Uptime = time.Now().Sub(stats.LeaderInfo.StartTime).String()
+	stats.SendingPkgRate, stats.SendingBandwidthRate = stats.SendRates()
+	stats.RecvingPkgRate, stats.RecvingBandwidthRate = stats.RecvRates()
+	b, err := json.Marshal(stats)
+	// TODO(jonboulle): appropriate error handling?
+	if err != nil {
+		log.Printf("stats: error marshalling server stats: %v", err)
+	}
+	return b
+}
+
+// Initialize clears the statistics of ServerStats and resets its start time
+func (ss *ServerStats) Initialize() {
+	if ss == nil {
+		return
+	}
+	now := time.Now()
+	ss.StartTime = now
+	ss.LeaderInfo.StartTime = now
+	ss.sendRateQueue = &statsQueue{
+		back: -1,
+	}
+	ss.recvRateQueue = &statsQueue{
+		back: -1,
+	}
+}
+
+// RecvRates calculates and returns the rate of received append requests
+func (ss *ServerStats) RecvRates() (float64, float64) {
+	return ss.recvRateQueue.Rate()
+}
+
+// SendRates calculates and returns the rate of sent append requests
+func (ss *ServerStats) SendRates() (float64, float64) {
+	return ss.sendRateQueue.Rate()
+}
+
+// RecvAppendReq updates the ServerStats in response to an AppendRequest
+// from the given leader being received
+func (ss *ServerStats) RecvAppendReq(leader string, reqSize int) {
+	ss.Lock()
+	defer ss.Unlock()
+
+	now := time.Now()
+
+	ss.State = raft.StateFollower
+	if leader != ss.LeaderInfo.Name {
+		ss.LeaderInfo.Name = leader
+		ss.LeaderInfo.StartTime = now
+	}
+
+	ss.recvRateQueue.Insert(
+		&RequestStats{
+			SendingTime: now,
+			Size:        reqSize,
+		},
+	)
+	ss.RecvAppendRequestCnt++
+}
+
+// SendAppendReq updates the ServerStats in response to an AppendRequest
+// being sent by this server
+func (ss *ServerStats) SendAppendReq(reqSize int) {
+	ss.Lock()
+	defer ss.Unlock()
+
+	now := time.Now()
+
+	if ss.State != raft.StateLeader {
+		ss.State = raft.StateLeader
+		ss.LeaderInfo.Name = ss.ID
+		ss.LeaderInfo.StartTime = now
+	}
+
+	ss.sendRateQueue.Insert(
+		&RequestStats{
+			SendingTime: now,
+			Size:        reqSize,
+		},
+	)
+
+	ss.SendAppendRequestCnt++
+}
+
+func (ss *ServerStats) BecomeLeader() {
+	if ss.State != raft.StateLeader {
+		ss.State = raft.StateLeader
+		ss.LeaderInfo.Name = ss.ID
+		ss.LeaderInfo.StartTime = time.Now()
+	}
+}

vendor/github.com/coreos/etcd/etcdserver/stats/stats.go

@@ -0,0 +1,32 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package stats defines a standard interface for etcd cluster statistics.
+package stats
+
+import "github.com/coreos/pkg/capnslog"
+
+var (
+	plog = capnslog.NewPackageLogger("github.com/coreos/etcd/etcdserver", "stats")
+)
+
+type Stats interface {
+	// SelfStats returns the struct representing statistics of this server
+	SelfStats() []byte
+	// LeaderStats returns the statistics of all followers in the cluster
+	// if this server is leader. Otherwise, nil is returned.
+	LeaderStats() []byte
+	// StoreStats returns statistics of the store backing this EtcdServer
+	StoreStats() []byte
+}

vendor/github.com/coreos/etcd/etcdserver/storage.go

@@ -0,0 +1,146 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	"io"
+	"os"
+	"path"
+
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/pkg/pbutil"
+	"github.com/coreos/etcd/pkg/types"
+	"github.com/coreos/etcd/raft/raftpb"
+	"github.com/coreos/etcd/snap"
+	"github.com/coreos/etcd/version"
+	"github.com/coreos/etcd/wal"
+	"github.com/coreos/etcd/wal/walpb"
+)
+
+type Storage interface {
+	// Save function saves ents and state to the underlying stable storage.
+	// Save MUST block until st and ents are on stable storage.
+	Save(st raftpb.HardState, ents []raftpb.Entry) error
+	// SaveSnap function saves snapshot to the underlying stable storage.
+	SaveSnap(snap raftpb.Snapshot) error
+	// DBFilePath returns the file path of database snapshot saved with given
+	// id.
+	DBFilePath(id uint64) (string, error)
+	// Close closes the Storage and performs finalization.
+	Close() error
+}
+
+type storage struct {
+	*wal.WAL
+	*snap.Snapshotter
+}
+
+func NewStorage(w *wal.WAL, s *snap.Snapshotter) Storage {
+	return &storage{w, s}
+}
+
+// SaveSnap saves the snapshot to disk and release the locked
+// wal files since they will not be used.
+func (st *storage) SaveSnap(snap raftpb.Snapshot) error {
+	walsnap := walpb.Snapshot{
+		Index: snap.Metadata.Index,
+		Term:  snap.Metadata.Term,
+	}
+	err := st.WAL.SaveSnapshot(walsnap)
+	if err != nil {
+		return err
+	}
+	err = st.Snapshotter.SaveSnap(snap)
+	if err != nil {
+		return err
+	}
+	err = st.WAL.ReleaseLockTo(snap.Metadata.Index)
+	if err != nil {
+		return err
+	}
+	return nil
+}
+
+func readWAL(waldir string, snap walpb.Snapshot) (w *wal.WAL, id, cid types.ID, st raftpb.HardState, ents []raftpb.Entry) {
+	var (
+		err       error
+		wmetadata []byte
+	)
+
+	repaired := false
+	for {
+		if w, err = wal.Open(waldir, snap); err != nil {
+			plog.Fatalf("open wal error: %v", err)
+		}
+		if wmetadata, st, ents, err = w.ReadAll(); err != nil {
+			w.Close()
+			// we can only repair ErrUnexpectedEOF and we never repair twice.
+			if repaired || err != io.ErrUnexpectedEOF {
+				plog.Fatalf("read wal error (%v) and cannot be repaired", err)
+			}
+			if !wal.Repair(waldir) {
+				plog.Fatalf("WAL error (%v) cannot be repaired", err)
+			} else {
+				plog.Infof("repaired WAL error (%v)", err)
+				repaired = true
+			}
+			continue
+		}
+		break
+	}
+	var metadata pb.Metadata
+	pbutil.MustUnmarshal(&metadata, wmetadata)
+	id = types.ID(metadata.NodeID)
+	cid = types.ID(metadata.ClusterID)
+	return
+}
+
+// upgradeDataDir converts an older version of the etcdServer data to the newest version.
+// It must ensure that, after upgrading, the most recent version is present.
+func upgradeDataDir(baseDataDir string, name string, ver version.DataDirVersion) error {
+	switch ver {
+	case version.DataDir2_0:
+		err := makeMemberDir(baseDataDir)
+		if err != nil {
+			return err
+		}
+		fallthrough
+	case version.DataDir2_0_1:
+		fallthrough
+	default:
+	}
+	return nil
+}
+
+func makeMemberDir(dir string) error {
+	membdir := path.Join(dir, "member")
+	_, err := os.Stat(membdir)
+	switch {
+	case err == nil:
+		return nil
+	case !os.IsNotExist(err):
+		return err
+	}
+	if err := os.MkdirAll(membdir, 0700); err != nil {
+		return err
+	}
+	names := []string{"snap", "wal"}
+	for _, name := range names {
+		if err := os.Rename(path.Join(dir, name), path.Join(membdir, name)); err != nil {
+			return err
+		}
+	}
+	return nil
+}

vendor/github.com/coreos/etcd/etcdserver/util.go

@@ -0,0 +1,42 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	"time"
+
+	"github.com/coreos/etcd/etcdserver/membership"
+	"github.com/coreos/etcd/pkg/types"
+	"github.com/coreos/etcd/rafthttp"
+)
+
+// isConnectedToQuorumSince checks whether the local member is connected to the
+// quorum of the cluster since the given time.
+func isConnectedToQuorumSince(transport rafthttp.Transporter, since time.Time, self types.ID, members []*membership.Member) bool {
+	var connectedNum int
+	for _, m := range members {
+		if m.ID == self || isConnectedSince(transport, since, m.ID) {
+			connectedNum++
+		}
+	}
+	return connectedNum >= (len(members)+1)/2
+}
+
+// isConnectedSince checks whether the local member is connected to the
+// remote member since the given time.
+func isConnectedSince(transport rafthttp.Transporter, since time.Time, remote types.ID) bool {
+	t := transport.ActiveSince(remote)
+	return !t.IsZero() && t.Before(since)
+}

vendor/github.com/coreos/etcd/etcdserver/v3demo_server.go

@@ -0,0 +1,273 @@
+// Copyright 2015 CoreOS, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package etcdserver
+
+import (
+	"time"
+
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"github.com/coreos/etcd/lease"
+	"github.com/coreos/etcd/lease/leasehttp"
+	dstorage "github.com/coreos/etcd/storage"
+	"golang.org/x/net/context"
+)
+
+const (
+	// the max request size that raft accepts.
+	// TODO: make this a flag? But we probably do not want to
+	// accept large request which might block raft stream. User
+	// specify a large value might end up with shooting in the foot.
+	maxRequestBytes = 1.5 * 1024 * 1024
+)
+
+type RaftKV interface {
+	Range(ctx context.Context, r *pb.RangeRequest) (*pb.RangeResponse, error)
+	Put(ctx context.Context, r *pb.PutRequest) (*pb.PutResponse, error)
+	DeleteRange(ctx context.Context, r *pb.DeleteRangeRequest) (*pb.DeleteRangeResponse, error)
+	Txn(ctx context.Context, r *pb.TxnRequest) (*pb.TxnResponse, error)
+	Compact(ctx context.Context, r *pb.CompactionRequest) (*pb.CompactionResponse, error)
+}
+
+type Lessor interface {
+	// LeaseGrant sends LeaseGrant request to raft and apply it after committed.
+	LeaseGrant(ctx context.Context, r *pb.LeaseGrantRequest) (*pb.LeaseGrantResponse, error)
+	// LeaseRevoke sends LeaseRevoke request to raft and apply it after committed.
+	LeaseRevoke(ctx context.Context, r *pb.LeaseRevokeRequest) (*pb.LeaseRevokeResponse, error)
+
+	// LeaseRenew renews the lease with given ID. The renewed TTL is returned. Or an error
+	// is returned.
+	LeaseRenew(id lease.LeaseID) (int64, error)
+}
+
+type Authenticator interface {
+	AuthEnable(ctx context.Context, r *pb.AuthEnableRequest) (*pb.AuthEnableResponse, error)
+	UserAdd(ctx context.Context, r *pb.AuthUserAddRequest) (*pb.AuthUserAddResponse, error)
+	UserDelete(ctx context.Context, r *pb.AuthUserDeleteRequest) (*pb.AuthUserDeleteResponse, error)
+	UserChangePassword(ctx context.Context, r *pb.AuthUserChangePasswordRequest) (*pb.AuthUserChangePasswordResponse, error)
+	RoleAdd(ctx context.Context, r *pb.AuthRoleAddRequest) (*pb.AuthRoleAddResponse, error)
+}
+
+func (s *EtcdServer) Range(ctx context.Context, r *pb.RangeRequest) (*pb.RangeResponse, error) {
+	if r.Serializable {
+		return s.applyV3.Range(noTxn, r)
+	}
+
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{Range: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.RangeResponse), result.err
+}
+
+func (s *EtcdServer) Put(ctx context.Context, r *pb.PutRequest) (*pb.PutResponse, error) {
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{Put: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.PutResponse), result.err
+}
+
+func (s *EtcdServer) DeleteRange(ctx context.Context, r *pb.DeleteRangeRequest) (*pb.DeleteRangeResponse, error) {
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{DeleteRange: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.DeleteRangeResponse), result.err
+}
+
+func (s *EtcdServer) Txn(ctx context.Context, r *pb.TxnRequest) (*pb.TxnResponse, error) {
+	if isTxnSerializable(r) {
+		return s.applyV3.Txn(r)
+	}
+
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{Txn: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.TxnResponse), result.err
+}
+
+func isTxnSerializable(r *pb.TxnRequest) bool {
+	for _, u := range r.Success {
+		if r := u.GetRequestRange(); r == nil || !r.Serializable {
+			return false
+		}
+	}
+	for _, u := range r.Failure {
+		if r := u.GetRequestRange(); r == nil || !r.Serializable {
+			return false
+		}
+	}
+	return true
+}
+
+func (s *EtcdServer) Compact(ctx context.Context, r *pb.CompactionRequest) (*pb.CompactionResponse, error) {
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{Compaction: r})
+	if r.Physical && result.physc != nil {
+		<-result.physc
+		// The compaction is done deleting keys; the hash is now settled
+		// but the data is not necessarily committed. If there's a crash,
+		// the hash may revert to a hash prior to compaction completing
+		// if the compaction resumes. Force the finished compaction to
+		// commit so it won't resume following a crash.
+		s.be.ForceCommit()
+	}
+	if err != nil {
+		return nil, err
+	}
+	resp := result.resp.(*pb.CompactionResponse)
+	if resp == nil {
+		resp = &pb.CompactionResponse{}
+	}
+	if resp.Header == nil {
+		resp.Header = &pb.ResponseHeader{}
+	}
+	resp.Header.Revision = s.kv.Rev()
+	return resp, result.err
+}
+
+func (s *EtcdServer) LeaseGrant(ctx context.Context, r *pb.LeaseGrantRequest) (*pb.LeaseGrantResponse, error) {
+	// no id given? choose one
+	for r.ID == int64(lease.NoLease) {
+		// only use positive int64 id's
+		r.ID = int64(s.reqIDGen.Next() & ((1 << 63) - 1))
+	}
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{LeaseGrant: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.LeaseGrantResponse), result.err
+}
+
+func (s *EtcdServer) LeaseRevoke(ctx context.Context, r *pb.LeaseRevokeRequest) (*pb.LeaseRevokeResponse, error) {
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{LeaseRevoke: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.LeaseRevokeResponse), result.err
+}
+
+func (s *EtcdServer) LeaseRenew(id lease.LeaseID) (int64, error) {
+	ttl, err := s.lessor.Renew(id)
+	if err == nil {
+		return ttl, nil
+	}
+	if err != lease.ErrNotPrimary {
+		return -1, err
+	}
+
+	// renewals don't go through raft; forward to leader manually
+	leader := s.cluster.Member(s.Leader())
+	for i := 0; i < 5 && leader == nil; i++ {
+		// wait an election
+		dur := time.Duration(s.cfg.ElectionTicks) * time.Duration(s.cfg.TickMs) * time.Millisecond
+		select {
+		case <-time.After(dur):
+			leader = s.cluster.Member(s.Leader())
+		case <-s.done:
+			return -1, ErrStopped
+		}
+	}
+	if leader == nil || len(leader.PeerURLs) == 0 {
+		return -1, ErrNoLeader
+	}
+
+	for _, url := range leader.PeerURLs {
+		lurl := url + "/leases"
+		ttl, err = leasehttp.RenewHTTP(id, lurl, s.peerRt, s.cfg.peerDialTimeout())
+		if err == nil {
+			break
+		}
+	}
+	return ttl, err
+}
+
+func (s *EtcdServer) Alarm(ctx context.Context, r *pb.AlarmRequest) (*pb.AlarmResponse, error) {
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{Alarm: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.AlarmResponse), result.err
+}
+
+func (s *EtcdServer) AuthEnable(ctx context.Context, r *pb.AuthEnableRequest) (*pb.AuthEnableResponse, error) {
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{AuthEnable: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.AuthEnableResponse), result.err
+}
+
+func (s *EtcdServer) UserAdd(ctx context.Context, r *pb.AuthUserAddRequest) (*pb.AuthUserAddResponse, error) {
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{AuthUserAdd: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.AuthUserAddResponse), result.err
+}
+
+func (s *EtcdServer) UserDelete(ctx context.Context, r *pb.AuthUserDeleteRequest) (*pb.AuthUserDeleteResponse, error) {
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{AuthUserDelete: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.AuthUserDeleteResponse), result.err
+}
+
+func (s *EtcdServer) UserChangePassword(ctx context.Context, r *pb.AuthUserChangePasswordRequest) (*pb.AuthUserChangePasswordResponse, error) {
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{AuthUserChangePassword: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.AuthUserChangePasswordResponse), result.err
+}
+
+func (s *EtcdServer) RoleAdd(ctx context.Context, r *pb.AuthRoleAddRequest) (*pb.AuthRoleAddResponse, error) {
+	result, err := s.processInternalRaftRequest(ctx, pb.InternalRaftRequest{AuthRoleAdd: r})
+	if err != nil {
+		return nil, err
+	}
+	return result.resp.(*pb.AuthRoleAddResponse), result.err
+}
+
+func (s *EtcdServer) processInternalRaftRequest(ctx context.Context, r pb.InternalRaftRequest) (*applyResult, error) {
+	r.ID = s.reqIDGen.Next()
+
+	data, err := r.Marshal()
+	if err != nil {
+		return nil, err
+	}
+
+	if len(data) > maxRequestBytes {
+		return nil, ErrRequestTooLarge
+	}
+
+	ch := s.w.Register(r.ID)
+
+	s.r.Propose(ctx, data)
+
+	select {
+	case x := <-ch:
+		return x.(*applyResult), nil
+	case <-ctx.Done():
+		s.w.Trigger(r.ID, nil) // GC wait
+		return nil, ctx.Err()
+	case <-s.done:
+		return nil, ErrStopped
+	}
+}
+
+// Watchable returns a watchable interface attached to the etcdserver.
+func (s *EtcdServer) Watchable() dstorage.Watchable { return s.KV() }