From 722ec487df61f436f1565e2e8334ff726017756d Mon Sep 17 00:00:00 2001 From: Marek Siarkowicz Date: Thu, 27 Jan 2022 19:04:41 +0100 Subject: [PATCH] server: Split metrics and health code --- server/embed/etcd.go | 6 +- server/etcdmain/etcd.go | 2 +- server/etcdserver/api/etcdhttp/health.go | 195 ++++++++++++++++++ .../{metrics_test.go => health_test.go} | 2 +- server/etcdserver/api/etcdhttp/metrics.go | 177 +--------------- tests/framework/integration/cluster.go | 3 +- 6 files changed, 205 insertions(+), 180 deletions(-) create mode 100644 server/etcdserver/api/etcdhttp/health.go rename server/etcdserver/api/etcdhttp/{metrics_test.go => health_test.go} (98%) diff --git a/server/embed/etcd.go b/server/embed/etcd.go index becdc666f..cc8fef220 100644 --- a/server/embed/etcd.go +++ b/server/embed/etcd.go @@ -702,7 +702,8 @@ func (e *Etcd) serveClients() (err error) { // Start a client server goroutine for each listen address mux := http.NewServeMux() etcdhttp.HandleBasic(e.cfg.logger, mux, e.Server) - etcdhttp.HandleMetricsHealth(e.cfg.logger, mux, e.Server) + etcdhttp.HandleMetrics(mux) + etcdhttp.HandleHealth(e.cfg.logger, mux, e.Server) gopts := []grpc.ServerOption{} if e.cfg.GRPCKeepAliveMinTime > time.Duration(0) { @@ -735,7 +736,8 @@ func (e *Etcd) serveMetrics() (err error) { if len(e.cfg.ListenMetricsUrls) > 0 { metricsMux := http.NewServeMux() - etcdhttp.HandleMetricsHealth(e.cfg.logger, metricsMux, e.Server) + etcdhttp.HandleMetrics(metricsMux) + etcdhttp.HandleHealth(e.cfg.logger, metricsMux, e.Server) for _, murl := range e.cfg.ListenMetricsUrls { tlsInfo := &e.cfg.ClientTLSInfo diff --git a/server/etcdmain/etcd.go b/server/etcdmain/etcd.go index 69828c5fd..eb50558ff 100644 --- a/server/etcdmain/etcd.go +++ b/server/etcdmain/etcd.go @@ -422,7 +422,7 @@ func startProxy(cfg *config) error { go func() { lg.Info("v2 proxy started listening on client requests", zap.String("host", host)) mux := http.NewServeMux() - etcdhttp.HandlePrometheus(mux) // v2 proxy just uses the same port + etcdhttp.HandleMetrics(mux) // v2 proxy just uses the same port mux.Handle("/", ph) lg.Fatal("done serving", zap.Error(http.Serve(l, mux))) }() diff --git a/server/etcdserver/api/etcdhttp/health.go b/server/etcdserver/api/etcdhttp/health.go new file mode 100644 index 000000000..1d29d97be --- /dev/null +++ b/server/etcdserver/api/etcdhttp/health.go @@ -0,0 +1,195 @@ +// Copyright 2017 The etcd Authors +// +// 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 etcdhttp + +import ( + "context" + "encoding/json" + "fmt" + "net/http" + + "github.com/prometheus/client_golang/prometheus" + "go.etcd.io/etcd/api/v3/etcdserverpb" + pb "go.etcd.io/etcd/api/v3/etcdserverpb" + "go.etcd.io/etcd/client/pkg/v3/types" + "go.etcd.io/etcd/raft/v3" + "go.etcd.io/etcd/server/v3/auth" + "go.etcd.io/etcd/server/v3/config" + "go.uber.org/zap" +) + +const ( + PathHealth = "/health" + PathProxyHealth = "/proxy/health" +) + +type ServerHealth interface { + Alarms() []*pb.AlarmMember + Leader() types.ID + Range(context.Context, *pb.RangeRequest) (*pb.RangeResponse, error) + Config() config.ServerConfig +} + +// HandleHealth registers metrics and health handlers. it checks health by using v3 range request +// and its corresponding timeout. +func HandleHealth(lg *zap.Logger, mux *http.ServeMux, srv ServerHealth) { + mux.Handle(PathHealth, NewHealthHandler(lg, func(excludedAlarms AlarmSet, serializable bool) Health { + if h := checkAlarms(lg, srv, excludedAlarms); h.Health != "true" { + return h + } + if h := checkLeader(lg, srv, serializable); h.Health != "true" { + return h + } + return checkAPI(lg, srv, serializable) + })) +} + +// NewHealthHandler handles '/health' requests. +func NewHealthHandler(lg *zap.Logger, hfunc func(excludedAlarms AlarmSet, Serializable bool) Health) http.HandlerFunc { + return func(w http.ResponseWriter, r *http.Request) { + if r.Method != http.MethodGet { + w.Header().Set("Allow", http.MethodGet) + http.Error(w, "Method Not Allowed", http.StatusMethodNotAllowed) + lg.Warn("/health error", zap.Int("status-code", http.StatusMethodNotAllowed)) + return + } + excludedAlarms := getExcludedAlarms(r) + // Passing the query parameter "serializable=true" ensures that the + // health of the local etcd is checked vs the health of the cluster. + // This is useful for probes attempting to validate the liveness of + // the etcd process vs readiness of the cluster to serve requests. + serializableFlag := getSerializableFlag(r) + h := hfunc(excludedAlarms, serializableFlag) + defer func() { + if h.Health == "true" { + healthSuccess.Inc() + } else { + healthFailed.Inc() + } + }() + d, _ := json.Marshal(h) + if h.Health != "true" { + http.Error(w, string(d), http.StatusServiceUnavailable) + lg.Warn("/health error", zap.String("output", string(d)), zap.Int("status-code", http.StatusServiceUnavailable)) + return + } + w.WriteHeader(http.StatusOK) + w.Write(d) + lg.Debug("/health OK", zap.Int("status-code", http.StatusOK)) + } +} + +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 { + Health string `json:"health"` + Reason string `json:"reason"` +} + +type AlarmSet map[string]struct{} + +func getExcludedAlarms(r *http.Request) (alarms AlarmSet) { + alarms = make(map[string]struct{}, 2) + alms, found := r.URL.Query()["exclude"] + if found { + for _, alm := range alms { + if len(alm) == 0 { + continue + } + alarms[alm] = struct{}{} + } + } + return alarms +} + +func getSerializableFlag(r *http.Request) bool { + return r.URL.Query().Get("serializable") == "true" +} + +// TODO: etcdserver.ErrNoLeader in health API + +func checkAlarms(lg *zap.Logger, srv ServerHealth, excludedAlarms AlarmSet) Health { + h := Health{Health: "true"} + as := srv.Alarms() + if len(as) > 0 { + for _, v := range as { + alarmName := v.Alarm.String() + if _, found := excludedAlarms[alarmName]; found { + lg.Debug("/health excluded alarm", zap.String("alarm", v.String())) + continue + } + + h.Health = "false" + switch v.Alarm { + case etcdserverpb.AlarmType_NOSPACE: + h.Reason = "ALARM NOSPACE" + case etcdserverpb.AlarmType_CORRUPT: + h.Reason = "ALARM CORRUPT" + default: + h.Reason = "ALARM UNKNOWN" + } + lg.Warn("serving /health false due to an alarm", zap.String("alarm", v.String())) + return h + } + } + + return h +} + +func checkLeader(lg *zap.Logger, srv ServerHealth, serializable bool) Health { + h := Health{Health: "true"} + if !serializable && (uint64(srv.Leader()) == raft.None) { + h.Health = "false" + h.Reason = "RAFT NO LEADER" + lg.Warn("serving /health false; no leader") + } + return h +} + +func checkAPI(lg *zap.Logger, srv ServerHealth, serializable bool) Health { + h := Health{Health: "true"} + cfg := srv.Config() + ctx, cancel := context.WithTimeout(context.Background(), cfg.ReqTimeout()) + _, err := srv.Range(ctx, &etcdserverpb.RangeRequest{KeysOnly: true, Limit: 1, Serializable: serializable}) + cancel() + if err != nil && err != auth.ErrUserEmpty && err != auth.ErrPermissionDenied { + h.Health = "false" + h.Reason = fmt.Sprintf("RANGE ERROR:%s", err) + lg.Warn("serving /health false; Range fails", zap.Error(err)) + return h + } + lg.Debug("serving /health true") + return h +} diff --git a/server/etcdserver/api/etcdhttp/metrics_test.go b/server/etcdserver/api/etcdhttp/health_test.go similarity index 98% rename from server/etcdserver/api/etcdhttp/metrics_test.go rename to server/etcdserver/api/etcdhttp/health_test.go index 788a038aa..7cb30148b 100644 --- a/server/etcdserver/api/etcdhttp/metrics_test.go +++ b/server/etcdserver/api/etcdhttp/health_test.go @@ -137,7 +137,7 @@ func TestHealthHandler(t *testing.T) { for i, tt := range tests { t.Run(tt.name, func(t *testing.T) { mux := http.NewServeMux() - HandleMetricsHealth(zaptest.NewLogger(t), mux, &fakeHealthServer{ + HandleHealth(zaptest.NewLogger(t), mux, &fakeHealthServer{ fakeServer: fakeServer{alarms: tt.alarms}, health: tt.expectHealth, apiError: tt.apiError, diff --git a/server/etcdserver/api/etcdhttp/metrics.go b/server/etcdserver/api/etcdhttp/metrics.go index 5b84a03bc..bf7d4a4a4 100644 --- a/server/etcdserver/api/etcdhttp/metrics.go +++ b/server/etcdserver/api/etcdhttp/metrics.go @@ -15,190 +15,17 @@ package etcdhttp import ( - "context" - "encoding/json" - "fmt" "net/http" - "github.com/prometheus/client_golang/prometheus" "github.com/prometheus/client_golang/prometheus/promhttp" - "go.etcd.io/etcd/api/v3/etcdserverpb" - pb "go.etcd.io/etcd/api/v3/etcdserverpb" - "go.etcd.io/etcd/client/pkg/v3/types" - "go.etcd.io/etcd/raft/v3" - "go.etcd.io/etcd/server/v3/auth" - "go.etcd.io/etcd/server/v3/config" - "go.uber.org/zap" ) const ( PathMetrics = "/metrics" - PathHealth = "/health" PathProxyMetrics = "/proxy/metrics" - PathProxyHealth = "/proxy/health" ) -type ServerHealth interface { - Alarms() []*pb.AlarmMember - Leader() types.ID - Range(context.Context, *pb.RangeRequest) (*pb.RangeResponse, error) - Config() config.ServerConfig -} - -// HandleMetricsHealth registers metrics and health handlers. it checks health by using v3 range request -// and its corresponding timeout. -func HandleMetricsHealth(lg *zap.Logger, mux *http.ServeMux, srv ServerHealth) { - mux.Handle(PathMetrics, promhttp.Handler()) - mux.Handle(PathHealth, NewHealthHandler(lg, func(excludedAlarms AlarmSet, serializable bool) Health { - if h := checkAlarms(lg, srv, excludedAlarms); h.Health != "true" { - return h - } - if h := checkLeader(lg, srv, serializable); h.Health != "true" { - return h - } - return checkAPI(lg, srv, serializable) - })) -} - -// HandlePrometheus registers prometheus handler on '/metrics'. -func HandlePrometheus(mux *http.ServeMux) { +// HandleMetrics registers prometheus handler on '/metrics'. +func HandleMetrics(mux *http.ServeMux) { mux.Handle(PathMetrics, promhttp.Handler()) } - -// NewHealthHandler handles '/health' requests. -func NewHealthHandler(lg *zap.Logger, hfunc func(excludedAlarms AlarmSet, Serializable bool) Health) http.HandlerFunc { - return func(w http.ResponseWriter, r *http.Request) { - if r.Method != http.MethodGet { - w.Header().Set("Allow", http.MethodGet) - http.Error(w, "Method Not Allowed", http.StatusMethodNotAllowed) - lg.Warn("/health error", zap.Int("status-code", http.StatusMethodNotAllowed)) - return - } - excludedAlarms := getExcludedAlarms(r) - // Passing the query parameter "serializable=true" ensures that the - // health of the local etcd is checked vs the health of the cluster. - // This is useful for probes attempting to validate the liveness of - // the etcd process vs readiness of the cluster to serve requests. - serializableFlag := getSerializableFlag(r) - h := hfunc(excludedAlarms, serializableFlag) - defer func() { - if h.Health == "true" { - healthSuccess.Inc() - } else { - healthFailed.Inc() - } - }() - d, _ := json.Marshal(h) - if h.Health != "true" { - http.Error(w, string(d), http.StatusServiceUnavailable) - lg.Warn("/health error", zap.String("output", string(d)), zap.Int("status-code", http.StatusServiceUnavailable)) - return - } - w.WriteHeader(http.StatusOK) - w.Write(d) - lg.Debug("/health OK", zap.Int("status-code", http.StatusOK)) - } -} - -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 { - Health string `json:"health"` - Reason string `json:"reason"` -} - -type AlarmSet map[string]struct{} - -func getExcludedAlarms(r *http.Request) (alarms AlarmSet) { - alarms = make(map[string]struct{}, 2) - alms, found := r.URL.Query()["exclude"] - if found { - for _, alm := range alms { - if len(alm) == 0 { - continue - } - alarms[alm] = struct{}{} - } - } - return alarms -} - -func getSerializableFlag(r *http.Request) bool { - return r.URL.Query().Get("serializable") == "true" -} - -// TODO: etcdserver.ErrNoLeader in health API - -func checkAlarms(lg *zap.Logger, srv ServerHealth, excludedAlarms AlarmSet) Health { - h := Health{Health: "true"} - as := srv.Alarms() - if len(as) > 0 { - for _, v := range as { - alarmName := v.Alarm.String() - if _, found := excludedAlarms[alarmName]; found { - lg.Debug("/health excluded alarm", zap.String("alarm", v.String())) - continue - } - - h.Health = "false" - switch v.Alarm { - case etcdserverpb.AlarmType_NOSPACE: - h.Reason = "ALARM NOSPACE" - case etcdserverpb.AlarmType_CORRUPT: - h.Reason = "ALARM CORRUPT" - default: - h.Reason = "ALARM UNKNOWN" - } - lg.Warn("serving /health false due to an alarm", zap.String("alarm", v.String())) - return h - } - } - - return h -} - -func checkLeader(lg *zap.Logger, srv ServerHealth, serializable bool) Health { - h := Health{Health: "true"} - if !serializable && (uint64(srv.Leader()) == raft.None) { - h.Health = "false" - h.Reason = "RAFT NO LEADER" - lg.Warn("serving /health false; no leader") - } - return h -} - -func checkAPI(lg *zap.Logger, srv ServerHealth, serializable bool) Health { - h := Health{Health: "true"} - cfg := srv.Config() - ctx, cancel := context.WithTimeout(context.Background(), cfg.ReqTimeout()) - _, err := srv.Range(ctx, &etcdserverpb.RangeRequest{KeysOnly: true, Limit: 1, Serializable: serializable}) - cancel() - if err != nil && err != auth.ErrUserEmpty && err != auth.ErrPermissionDenied { - h.Health = "false" - h.Reason = fmt.Sprintf("RANGE ERROR:%s", err) - lg.Warn("serving /health false; Range fails", zap.Error(err)) - return h - } - lg.Debug("serving /health true") - return h -} diff --git a/tests/framework/integration/cluster.go b/tests/framework/integration/cluster.go index 51c5525da..3dcb4c914 100644 --- a/tests/framework/integration/cluster.go +++ b/tests/framework/integration/cluster.go @@ -993,7 +993,8 @@ func (m *Member) Launch() error { for _, ln := range m.ClientListeners { handler := http.NewServeMux() etcdhttp.HandleBasic(m.Logger, handler, m.Server) - etcdhttp.HandleMetricsHealth(m.Logger, handler, m.Server) + etcdhttp.HandleMetrics(handler) + etcdhttp.HandleHealth(m.Logger, handler, m.Server) hs := &httptest.Server{ Listener: ln, Config: &http.Server{