Update vendor package github.com/coreos/...

vendor/github.com/coreos/etcd/etcdserver/api/etcdhttp/BUILD

@@ -23,6 +23,7 @@ go_library(
         "//vendor/github.com/coreos/etcd/rafthttp:go_default_library",
         "//vendor/github.com/coreos/etcd/version:go_default_library",
         "//vendor/github.com/coreos/pkg/capnslog:go_default_library",
+        "//vendor/github.com/prometheus/client_golang/prometheus:go_default_library",
         "//vendor/github.com/prometheus/client_golang/prometheus/promhttp:go_default_library",
     ],
 )

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

@@ -24,6 +24,7 @@ import (
 	"github.com/coreos/etcd/etcdserver/etcdserverpb"
 	"github.com/coreos/etcd/raft"
 
+	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/client_golang/prometheus/promhttp"
 )
 
@@ -43,11 +44,6 @@ func HandlePrometheus(mux *http.ServeMux) {
 	mux.Handle(pathMetrics, promhttp.Handler())
 }
 
-// HandleHealth registers health handler on '/health'.
-func HandleHealth(mux *http.ServeMux, srv etcdserver.ServerV2) {
-	mux.Handle(PathHealth, NewHealthHandler(func() Health { return checkHealth(srv) }))
-}
-
 // NewHealthHandler handles '/health' requests.
 func NewHealthHandler(hfunc func() Health) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
@@ -67,6 +63,26 @@ func NewHealthHandler(hfunc func() Health) http.HandlerFunc {
 	}
 }
 
+var (
+	healthSuccess = prometheus.NewCounter(prometheus.CounterOpts{
+		Namespace: "etcd",
+		Subsystem: "server",
+		Name:      "health_success",
+		Help:      "The total number of successful health checks",
+	})
+	healthFailed = prometheus.NewCounter(prometheus.CounterOpts{
+		Namespace: "etcd",
+		Subsystem: "server",
+		Name:      "health_failures",
+		Help:      "The total number of failed health checks",
+	})
+)
+
+func init() {
+	prometheus.MustRegister(healthSuccess)
+	prometheus.MustRegister(healthFailed)
+}
+
 // Health defines etcd server health status.
 // TODO: remove manual parsing in etcdctl cluster-health
 type Health struct {
@@ -97,5 +113,11 @@ func checkHealth(srv etcdserver.ServerV2) Health {
 			h.Health = "false"
 		}
 	}
+
+	if h.Health == "true" {
+		healthSuccess.Inc()
+	} else {
+		healthFailed.Inc()
+	}
 	return h
 }

vendor/github.com/coreos/etcd/etcdserver/api/v3rpc/BUILD

@@ -37,15 +37,17 @@ go_library(
         "//vendor/github.com/coreos/etcd/version:go_default_library",
         "//vendor/github.com/coreos/pkg/capnslog:go_default_library",
         "//vendor/github.com/gogo/protobuf/proto:go_default_library",
+        "//vendor/github.com/grpc-ecosystem/go-grpc-middleware:go_default_library",
         "//vendor/github.com/grpc-ecosystem/go-grpc-prometheus:go_default_library",
         "//vendor/github.com/prometheus/client_golang/prometheus:go_default_library",
+        "//vendor/go.uber.org/zap:go_default_library",
         "//vendor/google.golang.org/grpc:go_default_library",
         "//vendor/google.golang.org/grpc/codes:go_default_library",
         "//vendor/google.golang.org/grpc/credentials:go_default_library",
-        "//vendor/google.golang.org/grpc/grpclog:go_default_library",
         "//vendor/google.golang.org/grpc/health:go_default_library",
         "//vendor/google.golang.org/grpc/health/grpc_health_v1:go_default_library",
         "//vendor/google.golang.org/grpc/metadata:go_default_library",
+        "//vendor/google.golang.org/grpc/peer:go_default_library",
         "//vendor/google.golang.org/grpc/status:go_default_library",
     ],
 )

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

@@ -16,18 +16,15 @@ package v3rpc
 
 import (
 	"crypto/tls"
-	"io/ioutil"
 	"math"
-	"os"
-	"sync"
 
 	"github.com/coreos/etcd/etcdserver"
 	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
 
+	"github.com/grpc-ecosystem/go-grpc-middleware"
 	"github.com/grpc-ecosystem/go-grpc-prometheus"
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/credentials"
-	"google.golang.org/grpc/grpclog"
 	"google.golang.org/grpc/health"
 	healthpb "google.golang.org/grpc/health/grpc_health_v1"
 )
@@ -38,17 +35,21 @@ const (
 	maxSendBytes      = math.MaxInt32
 )
 
-// integration tests call this multiple times, which is racey in gRPC side
-var grpclogOnce sync.Once
-
 func Server(s *etcdserver.EtcdServer, tls *tls.Config, gopts ...grpc.ServerOption) *grpc.Server {
 	var opts []grpc.ServerOption
 	opts = append(opts, grpc.CustomCodec(&codec{}))
 	if tls != nil {
 		opts = append(opts, grpc.Creds(credentials.NewTLS(tls)))
 	}
-	opts = append(opts, grpc.UnaryInterceptor(newUnaryInterceptor(s)))
-	opts = append(opts, grpc.StreamInterceptor(newStreamInterceptor(s)))
+	opts = append(opts, grpc.UnaryInterceptor(grpc_middleware.ChainUnaryServer(
+		newLogUnaryInterceptor(s),
+		newUnaryInterceptor(s),
+		grpc_prometheus.UnaryServerInterceptor,
+	)))
+	opts = append(opts, grpc.StreamInterceptor(grpc_middleware.ChainStreamServer(
+		newStreamInterceptor(s),
+		grpc_prometheus.StreamServerInterceptor,
+	)))
 	opts = append(opts, grpc.MaxRecvMsgSize(int(s.Cfg.MaxRequestBytes+grpcOverheadBytes)))
 	opts = append(opts, grpc.MaxSendMsgSize(maxSendBytes))
 	opts = append(opts, grpc.MaxConcurrentStreams(maxStreams))
@@ -71,16 +72,5 @@ func Server(s *etcdserver.EtcdServer, tls *tls.Config, gopts ...grpc.ServerOptio
 	// set zero values for metrics registered for this grpc server
 	grpc_prometheus.Register(grpcServer)
 
-	grpclogOnce.Do(func() {
-		if s.Cfg.Debug {
-			grpc.EnableTracing = true
-			// enable info, warning, error
-			grpclog.SetLoggerV2(grpclog.NewLoggerV2(os.Stderr, os.Stderr, os.Stderr))
-		} else {
-			// only discard info
-			grpclog.SetLoggerV2(grpclog.NewLoggerV2(ioutil.Discard, os.Stderr, os.Stderr))
-		}
-	})
-
 	return grpcServer
 }

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

@@ -25,9 +25,11 @@ import (
 	"github.com/coreos/etcd/pkg/types"
 	"github.com/coreos/etcd/raft"
 
-	prometheus "github.com/grpc-ecosystem/go-grpc-prometheus"
+	pb "github.com/coreos/etcd/etcdserver/etcdserverpb"
+	"go.uber.org/zap"
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/metadata"
+	"google.golang.org/grpc/peer"
 )
 
 const (
@@ -40,7 +42,7 @@ type streamsMap struct {
 }
 
 func newUnaryInterceptor(s *etcdserver.EtcdServer) grpc.UnaryServerInterceptor {
-	return func(ctx context.Context, req interface{}, info *grpc.UnaryServerInfo, handler grpc.UnaryHandler) (resp interface{}, err error) {
+	return func(ctx context.Context, req interface{}, info *grpc.UnaryServerInfo, handler grpc.UnaryHandler) (interface{}, error) {
 		if !api.IsCapabilityEnabled(api.V3rpcCapability) {
 			return nil, rpctypes.ErrGRPCNotCapable
 		}
@@ -54,7 +56,124 @@ func newUnaryInterceptor(s *etcdserver.EtcdServer) grpc.UnaryServerInterceptor {
 			}
 		}
 
-		return prometheus.UnaryServerInterceptor(ctx, req, info, handler)
+		return handler(ctx, req)
+	}
+}
+
+func newLogUnaryInterceptor(s *etcdserver.EtcdServer) grpc.UnaryServerInterceptor {
+	return func(ctx context.Context, req interface{}, info *grpc.UnaryServerInfo, handler grpc.UnaryHandler) (interface{}, error) {
+		startTime := time.Now()
+		resp, err := handler(ctx, req)
+		defer logUnaryRequestStats(ctx, nil, info, startTime, req, resp)
+		return resp, err
+	}
+}
+
+func logUnaryRequestStats(ctx context.Context, lg *zap.Logger, info *grpc.UnaryServerInfo, startTime time.Time, req interface{}, resp interface{}) {
+	duration := time.Since(startTime)
+	remote := "No remote client info."
+	peerInfo, ok := peer.FromContext(ctx)
+	if ok {
+		remote = peerInfo.Addr.String()
+	}
+	var responseType string = info.FullMethod
+	var reqCount, respCount int64
+	var reqSize, respSize int
+	var reqContent string
+	switch _resp := resp.(type) {
+	case *pb.RangeResponse:
+		_req, ok := req.(*pb.RangeRequest)
+		if ok {
+			reqCount = 0
+			reqSize = _req.Size()
+			reqContent = _req.String()
+		}
+		if _resp != nil {
+			respCount = _resp.GetCount()
+			respSize = _resp.Size()
+		}
+	case *pb.PutResponse:
+		_req, ok := req.(*pb.PutRequest)
+		if ok {
+			reqCount = 1
+			reqSize = _req.Size()
+			reqContent = pb.NewLoggablePutRequest(_req).String()
+			// redact value field from request content, see PR #9821
+		}
+		if _resp != nil {
+			respCount = 0
+			respSize = _resp.Size()
+		}
+	case *pb.DeleteRangeResponse:
+		_req, ok := req.(*pb.DeleteRangeRequest)
+		if ok {
+			reqCount = 0
+			reqSize = _req.Size()
+			reqContent = _req.String()
+		}
+		if _resp != nil {
+			respCount = _resp.GetDeleted()
+			respSize = _resp.Size()
+		}
+	case *pb.TxnResponse:
+		_req, ok := req.(*pb.TxnRequest)
+		if ok && _resp != nil {
+			if _resp.GetSucceeded() { // determine the 'actual' count and size of request based on success or failure
+				reqCount = int64(len(_req.GetSuccess()))
+				reqSize = 0
+				for _, r := range _req.GetSuccess() {
+					reqSize += r.Size()
+				}
+			} else {
+				reqCount = int64(len(_req.GetFailure()))
+				reqSize = 0
+				for _, r := range _req.GetFailure() {
+					reqSize += r.Size()
+				}
+			}
+			reqContent = pb.NewLoggableTxnRequest(_req).String()
+			// redact value field from request content, see PR #9821
+		}
+		if _resp != nil {
+			respCount = 0
+			respSize = _resp.Size()
+		}
+	default:
+		reqCount = -1
+		reqSize = -1
+		respCount = -1
+		respSize = -1
+	}
+
+	logGenericRequestStats(lg, startTime, duration, remote, responseType, reqCount, reqSize, respCount, respSize, reqContent)
+}
+
+func logGenericRequestStats(lg *zap.Logger, startTime time.Time, duration time.Duration, remote string, responseType string,
+	reqCount int64, reqSize int, respCount int64, respSize int, reqContent string) {
+	if lg == nil {
+		plog.Debugf("start time = %v, "+
+			"time spent = %v, "+
+			"remote = %s, "+
+			"response type = %s, "+
+			"request count = %d, "+
+			"request size = %d, "+
+			"response count = %d, "+
+			"response size = %d, "+
+			"request content = %s",
+			startTime, duration, remote, responseType, reqCount, reqSize, respCount, respSize, reqContent,
+		)
+	} else {
+		lg.Debug("request stats",
+			zap.Time("start time", startTime),
+			zap.Duration("time spent", duration),
+			zap.String("remote", remote),
+			zap.String("response type", responseType),
+			zap.Int64("request count", reqCount),
+			zap.Int("request size", reqSize),
+			zap.Int64("response count", respCount),
+			zap.Int("response size", respSize),
+			zap.String("request content", reqContent),
+		)
 	}
 }
 
@@ -90,7 +209,7 @@ func newStreamInterceptor(s *etcdserver.EtcdServer) grpc.StreamServerInterceptor
 			}
 		}
 
-		return prometheus.StreamServerInterceptor(srv, ss, info, handler)
+		return handler(srv, ss)
 	}
 }
 

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

@@ -59,7 +59,7 @@ func (as *InternalRaftStringer) String() string {
 	case as.Request.Put != nil:
 		return fmt.Sprintf("header:<%s> put:<%s>",
 			as.Request.Header.String(),
-			newLoggablePutRequest(as.Request.Put).String(),
+			NewLoggablePutRequest(as.Request.Put).String(),
 		)
 	case as.Request.Txn != nil:
 		return fmt.Sprintf("header:<%s> txn:<%s>",
@@ -121,7 +121,7 @@ func newLoggableRequestOp(op *RequestOp) *requestOpStringer {
 func (as *requestOpStringer) String() string {
 	switch op := as.Op.Request.(type) {
 	case *RequestOp_RequestPut:
-		return fmt.Sprintf("request_put:<%s>", newLoggablePutRequest(op.RequestPut).String())
+		return fmt.Sprintf("request_put:<%s>", NewLoggablePutRequest(op.RequestPut).String())
 	case *RequestOp_RequestTxn:
 		return fmt.Sprintf("request_txn:<%s>", NewLoggableTxnRequest(op.RequestTxn).String())
 	default:
@@ -167,7 +167,7 @@ type loggablePutRequest struct {
 	IgnoreLease bool   `protobuf:"varint,6,opt,name=ignore_lease,proto3"`
 }
 
-func newLoggablePutRequest(request *PutRequest) *loggablePutRequest {
+func NewLoggablePutRequest(request *PutRequest) *loggablePutRequest {
 	return &loggablePutRequest{
 		request.Key,
 		len(request.Value),

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

@@ -90,6 +90,12 @@ var (
 		Name:      "slow_read_indexes_total",
 		Help:      "The total number of pending read indexes not in sync with leader's or timed out read index requests.",
 	})
+	readIndexFailed = prometheus.NewCounter(prometheus.CounterOpts{
+		Namespace: "etcd",
+		Subsystem: "server",
+		Name:      "read_indexes_failed_total",
+		Help:      "The total number of failed read indexes seen.",
+	})
 	quotaBackendBytes = prometheus.NewGauge(prometheus.GaugeOpts{
 		Namespace: "etcd",
 		Subsystem: "server",
@@ -110,6 +116,13 @@ var (
 		Help:      "Which Go version server is running with. 1 for 'server_go_version' label with current version.",
 	},
 		[]string{"server_go_version"})
+	serverID = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: "etcd",
+		Subsystem: "server",
+		Name:      "id",
+		Help:      "Server or member ID in hexadecimal format. 1 for 'server_id' label with current ID.",
+	},
+		[]string{"server_id"})
 )
 
 func init() {
@@ -124,9 +137,11 @@ func init() {
 	prometheus.MustRegister(proposalsFailed)
 	prometheus.MustRegister(leaseExpired)
 	prometheus.MustRegister(slowReadIndex)
+	prometheus.MustRegister(readIndexFailed)
 	prometheus.MustRegister(quotaBackendBytes)
 	prometheus.MustRegister(currentVersion)
 	prometheus.MustRegister(currentGoVersion)
+	prometheus.MustRegister(serverID)
 
 	currentVersion.With(prometheus.Labels{
 		"server_version": version.Version,

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

@@ -59,6 +59,7 @@ import (
 
 	"github.com/coreos/go-semver/semver"
 	"github.com/coreos/pkg/capnslog"
+	"github.com/prometheus/client_golang/prometheus"
 )
 
 const (
@@ -435,6 +436,7 @@ func NewServer(cfg ServerConfig) (srv *EtcdServer, err error) {
 		reqIDGen:      idutil.NewGenerator(uint16(id), time.Now()),
 		forceVersionC: make(chan struct{}),
 	}
+	serverID.With(prometheus.Labels{"server_id": id.String()}).Set(1)
 
 	srv.applyV2 = &applierV2store{store: srv.store, cluster: srv.cluster}
 

vendor/github.com/coreos/etcd/etcdserver/v3_server.go

@@ -634,6 +634,7 @@ func (s *EtcdServer) linearizableReadLoop() {
 				return
 			}
 			plog.Errorf("failed to get read index from raft: %v", err)
+			readIndexFailed.Inc()
 			nr.notify(err)
 			continue
 		}
@@ -659,7 +660,7 @@ func (s *EtcdServer) linearizableReadLoop() {
 				}
 
 			case <-time.After(s.Cfg.ReqTimeout()):
-				plog.Warningf("timed out waiting for read index response")
+				plog.Warningf("timed out waiting for read index response (local node might have slow network)")
 				nr.notify(ErrTimeout)
 				timeout = true
 				slowReadIndex.Inc()