*: bump to v0.4.9

server: introduce /v2/migration/snapshot endpoint

CHANGELOG

@@ -1,3 +1,10 @@
+v0.4.6
+* Fix long-term timer leak (#900, #875, #868, #904)
+* Fix `Running` field in standby_info file (#881)
+* Add `quorum=true` query parameter for GET requests (#866, #883)
+* Add `Access-Control-Allow-Headers` header for CORS requests (#886)
+* Various documentation improvements (#907, #882)
+
 v0.4.5
 * Flush headers immediatly on `wait=true` requests (#877)
 * Add `ETCD_HTTP_READ_TIMEOUT` and `ETCD_HTTP_WRITE_TIMEOUT` (#880)

Documentation/api.md

@@ -13,6 +13,14 @@ This will bring up etcd listening on default ports (4001 for client communicatio
 The `-data-dir machine0` argument tells etcd to write machine configuration, logs and snapshots to the `./machine0/` directory.
 The `-name machine0` tells the rest of the cluster that this machine is named machine0.
 
+## Getting the etcd version
+
+The etcd version of a specific instance can be obtained from the `/version` endpoint.
+
+```sh
+curl -L http://127.0.0.1:4001/version
+```
+
 ## Key Space Operations
 
 The primary API of etcd is a hierarchical key space.
@@ -833,6 +841,8 @@ curl -L http://127.0.0.1:4001/v2/keys/afile -XPUT --data-urlencode value@afile.t
 
 ### Read Consistency
 
+#### Read from the Master
+
 Followers in a cluster can be behind the leader in their copy of the keyspace.
 If your application wants or needs the most up-to-date version of a key then it should ensure it reads from the current leader.
 By using the `consistent=true` flag in your GET requests, etcd will make sure you are talking to the current master.
@@ -843,6 +853,19 @@ The client is told the write was successful and the keyspace is updated.
 Meanwhile F2 has partitioned from the network and will have an out-of-date version of the keyspace until the partition resolves.
 Since F2 missed the most recent write, a client reading from F2 will have an out-of-date version of the keyspace.
 
+Implementation notes on `consistent=true`: If the leader you are talking to is
+partitioned it will be unable to determine if it is not currently the master.
+In a later version we will provide a mechanism to set an upperbound of time
+that the current master can be unable to contact the quorom and still serve
+reads.
+
+### Read Linearization
+
+If you want a read that is fully linearized you can use a `quorum=true` GET.
+The read will take a very similar path to a write and will have a similar
+speed. If you are unsure if you need this feature feel free to email etcd-dev
+for advice.
+
 ## Lock Module (*Deprecated and Removed*)
 
 The lock module is used to serialize access to resources used by clients.

Documentation/configuration.md

@@ -26,7 +26,7 @@ The full documentation is contained in the [API docs](https://github.com/coreos/
 
 ### Required
 
-* `-name` - The node name. Defaults to the hostname.
+* `-name` - The node name. Defaults to a UUID.
 
 ### Optional
 
@@ -34,7 +34,7 @@ The full documentation is contained in the [API docs](https://github.com/coreos/
 * `-discovery` - A URL to use for discovering the peer list. (i.e `"https://discovery.etcd.io/your-unique-key"`).
 * `-http-read-timeout` - The number of seconds before an HTTP read operation is timed out.
 * `-http-write-timeout` - The number of seconds before an HTTP write operation is timed out.
-* `-bind-addr` - The listening hostname for client communication. Defaults to advertised IP.
+* `-bind-addr` - The listening hostname for client communication. Defaults to 0.0.0.0 and the advertised port.
 * `-peers` - A comma separated list of peers in the cluster (i.e `"203.0.113.101:7001,203.0.113.102:7001"`).
 * `-peers-file` - The file path containing a comma separated list of peers in the cluster.
 * `-ca-file` - The path of the client CAFile. Enables client cert authentication when present.
@@ -47,7 +47,7 @@ The full documentation is contained in the [API docs](https://github.com/coreos/
 * `-max-result-buffer` - The max size of result buffer. Defaults to `1024`.
 * `-max-retry-attempts` - The max retry attempts when trying to join a cluster. Defaults to `3`.
 * `-peer-addr` - The advertised public hostname:port for server communication. Defaults to `127.0.0.1:7001`.
-* `-peer-bind-addr` - The listening hostname for server communication. Defaults to advertised IP.
+* `-peer-bind-addr` - The listening hostname for server communication. Defaults to 0.0.0.0 and the advertised peer port.
 * `-peer-ca-file` - The path of the CAFile. Enables client/peer cert authentication when present.
 * `-peer-cert-file` - The cert file of the server.
 * `-peer-key-file` - The key file of the server.
@@ -55,8 +55,8 @@ The full documentation is contained in the [API docs](https://github.com/coreos/
 * `-peer-heartbeat-interval` - The number of milliseconds in between heartbeat requests
 * `-snapshot=false` - Disable log snapshots. Defaults to `true`.
 * `-cluster-active-size` - The expected number of instances participating in the consensus protocol. Only applied if the etcd instance is the first peer in the cluster.
-* `-cluster-remove-delay` - The delay before one node is removed from the cluster since it cannot be connected at all. Only applied if the etcd instance is the first peer in the cluster.
-* `-cluster-sync-interval` - The interval between synchronization for standby-mode instance with the cluster. Only applied if the etcd instance is the first peer in the cluster.
+* `-cluster-remove-delay` - The number of seconds before one node is removed from the cluster since it cannot be connected at all. Only applied if the etcd instance is the first peer in the cluster.
+* `-cluster-sync-interval` - The number of seconds between synchronization for standby-mode instance with the cluster. Only applied if the etcd instance is the first peer in the cluster.
 * `-v` - Enable verbose logging. Defaults to `false`.
 * `-vv` - Enable very verbose logging. Defaults to `false`.
 * `-version` - Print the version and exit.

README.md

@@ -1,6 +1,6 @@
 # etcd
 
-README version 0.4.5
+README version 0.4.8
 
 A highly-available key value store for shared configuration and service discovery.
 etcd is inspired by [Apache ZooKeeper][zookeeper] and [doozer][doozer], with a focus on being:

config/config.go

@@ -86,7 +86,7 @@ type Config struct {
 		HeartbeatInterval int    `toml:"heartbeat_interval" env:"ETCD_PEER_HEARTBEAT_INTERVAL"`
 		ElectionTimeout   int    `toml:"election_timeout" env:"ETCD_PEER_ELECTION_TIMEOUT"`
 	}
-	strTrace     string `toml:"trace" env:"ETCD_TRACE"`
+	StrTrace     string `toml:"trace" env:"ETCD_TRACE"`
 	GraphiteHost string `toml:"graphite_host" env:"ETCD_GRAPHITE_HOST"`
 	Cluster      struct {
 		ActiveSize   int     `toml:"active_size" env:"ETCD_CLUSTER_ACTIVE_SIZE"`
@@ -275,7 +275,7 @@ func (c *Config) LoadFlags(arguments []string) error {
 	f.IntVar(&c.SnapshotCount, "snapshot-count", c.SnapshotCount, "")
 	f.StringVar(&c.CPUProfileFile, "cpuprofile", "", "")
 
-	f.StringVar(&c.strTrace, "trace", "", "")
+	f.StringVar(&c.StrTrace, "trace", "", "")
 	f.StringVar(&c.GraphiteHost, "graphite-host", "", "")
 
 	f.IntVar(&c.Cluster.ActiveSize, "cluster-active-size", c.Cluster.ActiveSize, "")
@@ -441,7 +441,7 @@ func (c *Config) MetricsBucketName() string {
 
 // Trace determines if any trace-level information should be emitted
 func (c *Config) Trace() bool {
-	return c.strTrace == "*"
+	return c.StrTrace == "*"
 }
 
 func (c *Config) ClusterConfig() *server.ClusterConfig {

http/cors.go

@@ -54,6 +54,7 @@ type CORSHandler struct {
 func (h *CORSHandler) addHeader(w http.ResponseWriter, origin string) {
 	w.Header().Add("Access-Control-Allow-Methods", "POST, GET, OPTIONS, PUT, DELETE")
 	w.Header().Add("Access-Control-Allow-Origin", origin)
+	w.Header().Add("Access-Control-Allow-Headers", "accept, content-type")
 }
 
 // ServeHTTP adds the correct CORS headers based on the origin and returns immediately

server/join_command.go

@@ -46,6 +46,7 @@ func (c *JoinCommand) NodeName() string {
 // applyJoin attempts to join a machine to the cluster.
 func applyJoin(c *JoinCommand, context raft.Context) (uint64, error) {
 	ps, _ := context.Server().Context().(*PeerServer)
+	ps.raftServer.FlushCommitIndex()
 	commitIndex := context.CommitIndex()
 
 	// Make sure we're not getting a cached value from the registry.

server/peer_server.go

@@ -772,9 +772,9 @@ func (s *PeerServer) startRoutine(f func()) {
 func (s *PeerServer) monitorSnapshot() {
 	for {
 		timer := time.NewTimer(s.snapConf.checkingInterval)
-		defer timer.Stop()
 		select {
 		case <-s.closeChan:
+			timer.Stop()
 			return
 		case <-timer.C:
 		}
@@ -807,6 +807,8 @@ func (s *PeerServer) monitorSync() {
 // monitorTimeoutThreshold groups timeout threshold events together and prints
 // them as a single log line.
 func (s *PeerServer) monitorTimeoutThreshold() {
+	ticker := time.NewTicker(ThresholdMonitorTimeout)
+	defer ticker.Stop()
 	for {
 		select {
 		case <-s.closeChan:
@@ -815,12 +817,10 @@ func (s *PeerServer) monitorTimeoutThreshold() {
 			log.Infof("%s: warning: heartbeat near election timeout: %v", s.Config.Name, value)
 		}
 
-		timer := time.NewTimer(ThresholdMonitorTimeout)
-		defer timer.Stop()
 		select {
 		case <-s.closeChan:
 			return
-		case <-timer.C:
+		case <-ticker.C:
 		}
 	}
 }
@@ -828,13 +828,13 @@ func (s *PeerServer) monitorTimeoutThreshold() {
 // monitorActiveSize has the leader periodically check the status of cluster
 // nodes and swaps them out for standbys as needed.
 func (s *PeerServer) monitorActiveSize() {
+	ticker := time.NewTicker(ActiveMonitorTimeout)
+	defer ticker.Stop()
 	for {
-		timer := time.NewTimer(ActiveMonitorTimeout)
-		defer timer.Stop()
 		select {
 		case <-s.closeChan:
 			return
-		case <-timer.C:
+		case <-ticker.C:
 		}
 
 		// Ignore while this peer is not a leader.
@@ -864,13 +864,13 @@ func (s *PeerServer) monitorActiveSize() {
 
 // monitorPeerActivity has the leader periodically for dead nodes and demotes them.
 func (s *PeerServer) monitorPeerActivity() {
+	ticker := time.NewTicker(PeerActivityMonitorTimeout)
+	defer ticker.Stop()
 	for {
-		timer := time.NewTimer(PeerActivityMonitorTimeout)
-		defer timer.Stop()
 		select {
 		case <-s.closeChan:
 			return
-		case <-timer.C:
+		case <-ticker.C:
 		}
 
 		// Ignore while this peer is not a leader.

server/release_version.go

@@ -1,3 +1,3 @@
 package server
 
-const ReleaseVersion = "0.4.5"
+const ReleaseVersion = "0.4.9"

server/server.go

@@ -3,8 +3,10 @@ package server
 import (
 	"encoding/json"
 	"fmt"
+	"io/ioutil"
 	"net/http"
 	"net/http/pprof"
+	"path"
 	"strings"
 	"time"
 
@@ -135,6 +137,7 @@ func (s *Server) installV2(r *mux.Router) {
 	s.handleFunc(r2, "/v2/stats/leader", s.GetLeaderStatsHandler).Methods("GET", "HEAD")
 	s.handleFunc(r2, "/v2/stats/store", s.GetStoreStatsHandler).Methods("GET", "HEAD")
 	s.handleFunc(r2, "/v2/speedTest", s.SpeedTestHandler).Methods("GET", "HEAD")
+	s.handleFunc(r2, "/v2/migration/snapshot", s.SnapshotHandler).Methods("GET")
 }
 
 func (s *Server) installMod(r *mux.Router) {
@@ -324,12 +327,8 @@ func (s *Server) GetLeaderStatsHandler(w http.ResponseWriter, req *http.Request)
 		return nil
 	}
 
-	leader := s.peerServer.RaftServer().Leader()
-	if leader == "" {
-		return etcdErr.NewError(300, "", s.Store().Index())
-	}
-	hostname, _ := s.registry.ClientURL(leader)
-	uhttp.Redirect(hostname, w, req)
+	w.WriteHeader(http.StatusForbidden)
+	w.Write([]byte("not current leader"))
 	return nil
 }
 
@@ -363,6 +362,41 @@ func (s *Server) SpeedTestHandler(w http.ResponseWriter, req *http.Request) erro
 	return nil
 }
 
+// SnapshotHandler forces etcd store to do a snapshot. If the disk parameter is set, the snapshot
+// will be written to disk at data-dir/index-migrate.snap. Or the snapshot will be returned as
+// http body.
+func (s *Server) SnapshotHandler(w http.ResponseWriter, req *http.Request) error {
+	data, err := s.Store().Save()
+	if err != nil {
+		http.Error(w, "failed to create snapshot: "+err.Error(), http.StatusInternalServerError)
+		log.Warn("Failed to create snapshot:" + err.Error())
+		return nil
+	}
+
+	disk := req.FormValue("disk")
+	if disk == "true" {
+		name := fmt.Sprintf("%d-migrate.snap", s.peerServer.RaftServer().CommitIndex())
+		err = ioutil.WriteFile(path.Join(s.peerServer.RaftServer().Path(), name), data, 0600)
+		if err != nil {
+			http.Error(w, "failed to save snapshot: "+err.Error(), http.StatusInternalServerError)
+			log.Warn("server: failed to save snapshot: " + err.Error())
+			return nil
+		}
+		log.Infof("server: saved snapshot file %s successfully", name)
+		return nil
+	}
+	if disk != "" && disk != "false" {
+		http.Error(w, "invalid parameter: disk="+disk, http.StatusBadRequest)
+		return nil
+	}
+
+	_, err = w.Write(data)
+	if err != nil {
+		log.Warnf("server: failed to write snapshot to %s: %v", req.RemoteAddr, err.Error())
+	}
+	return nil
+}
+
 // Retrieves metrics from bucket
 func (s *Server) GetMetricsHandler(w http.ResponseWriter, req *http.Request) error {
 	(*s.metrics).Dump(w)

server/standby_server.go

@@ -30,6 +30,7 @@ type StandbyServerConfig struct {
 }
 
 type standbyInfo struct {
+	// stay running in standby mode
 	Running      bool
 	Cluster      []*machineMessage
 	SyncInterval float64
@@ -78,12 +79,16 @@ func (s *StandbyServer) Start() {
 	s.removeNotify = make(chan bool)
 	s.closeChan = make(chan bool)
 
+	s.Running = true
+	if err := s.saveInfo(); err != nil {
+		log.Warnf("error saving cluster info for standby")
+	}
+
 	s.routineGroup.Add(1)
 	go func() {
 		defer s.routineGroup.Done()
 		s.monitorCluster()
 	}()
-	s.Running = true
 }
 
 // Stop stops the server gracefully.
@@ -97,11 +102,6 @@ func (s *StandbyServer) Stop() {
 
 	close(s.closeChan)
 	s.routineGroup.Wait()
-
-	if err := s.saveInfo(); err != nil {
-		log.Warnf("error saving cluster info for standby")
-	}
-	s.Running = false
 }
 
 // RemoveNotify notifies the server is removed from standby mode and ready
@@ -178,13 +178,13 @@ func (s *StandbyServer) redirectRequests(w http.ResponseWriter, r *http.Request)
 // monitorCluster assumes that the machine has tried to join the cluster and
 // failed, so it waits for the interval at the beginning.
 func (s *StandbyServer) monitorCluster() {
+	ticker := time.NewTicker(time.Duration(int64(s.SyncInterval * float64(time.Second))))
+	defer ticker.Stop()
 	for {
-		timer := time.NewTimer(time.Duration(int64(s.SyncInterval * float64(time.Second))))
-		defer timer.Stop()
 		select {
 		case <-s.closeChan:
 			return
-		case <-timer.C:
+		case <-ticker.C:
 		}
 
 		if err := s.syncCluster(nil); err != nil {
@@ -204,6 +204,10 @@ func (s *StandbyServer) monitorCluster() {
 		}
 
 		log.Infof("join through leader %v", leader.PeerURL)
+		s.Running = false
+		if err := s.saveInfo(); err != nil {
+			log.Warnf("error saving cluster info for standby")
+		}
 		go func() {
 			s.Stop()
 			close(s.removeNotify)

server/v2/get_handler.go

@@ -17,6 +17,14 @@ func GetHandler(w http.ResponseWriter, req *http.Request, s Server) error {
 	vars := mux.Vars(req)
 	key := "/" + vars["key"]
 
+	recursive := (req.FormValue("recursive") == "true")
+	sort := (req.FormValue("sorted") == "true")
+
+	if req.FormValue("quorum") == "true" {
+		c := s.Store().CommandFactory().CreateGetCommand(key, recursive, sort)
+		return s.Dispatch(c, w, req)
+	}
+
 	// Help client to redirect the request to the current leader
 	if req.FormValue("consistent") == "true" && s.State() != raft.Leader {
 		leader := s.Leader()
@@ -35,8 +43,6 @@ func GetHandler(w http.ResponseWriter, req *http.Request, s Server) error {
 		return nil
 	}
 
-	recursive := (req.FormValue("recursive") == "true")
-	sort := (req.FormValue("sorted") == "true")
 	waitIndex := req.FormValue("waitIndex")
 	stream := (req.FormValue("stream") == "true")
 

store/command_factory.go

@@ -24,6 +24,7 @@ type CommandFactory interface {
 		prevIndex uint64, expireTime time.Time) raft.Command
 	CreateCompareAndDeleteCommand(key string, prevValue string, prevIndex uint64) raft.Command
 	CreateSyncCommand(now time.Time) raft.Command
+	CreateGetCommand(key string, recursive, sorted bool) raft.Command
 }
 
 // RegisterCommandFactory adds a command factory to the global registry.

store/v2/command_factory.go

@@ -89,3 +89,11 @@ func (f *CommandFactory) CreateSyncCommand(now time.Time) raft.Command {
 		Time: time.Now(),
 	}
 }
+
+func (f *CommandFactory) CreateGetCommand(key string, recursive, sorted bool) raft.Command {
+	return &GetCommand{
+		Key:       key,
+		Recursive: recursive,
+		Sorted:    sorted,
+	}
+}

store/v2/get_command.go

@@ -0,0 +1,35 @@
+package v2
+
+import (
+	"github.com/coreos/etcd/log"
+	"github.com/coreos/etcd/store"
+	"github.com/coreos/etcd/third_party/github.com/goraft/raft"
+)
+
+func init() {
+	raft.RegisterCommand(&GetCommand{})
+}
+
+// The GetCommand gets a key from the Store.
+type GetCommand struct {
+	Key       string `json:"key"`
+	Recursive bool   `json:"recursive"`
+	Sorted    bool   `json:sorted`
+}
+
+// The name of the get command in the log
+func (c *GetCommand) CommandName() string {
+	return "etcd:get"
+}
+
+// Get the key
+func (c *GetCommand) Apply(context raft.Context) (interface{}, error) {
+	s, _ := context.Server().StateMachine().(store.Store)
+	e, err := s.Get(c.Key, c.Recursive, c.Sorted)
+
+	if err != nil {
+		log.Debug(err)
+		return nil, err
+	}
+	return e, nil
+}