Merge PR #6399: SDK Telemetry

2020-06-16 11:11:02 -04:00
parent e2f336345d
commit 8ee7d1f403
11 changed files with 492 additions and 89 deletions
@@ -1,9 +1,11 @@
 package api

 import (
+	"fmt"
 	"net"
 	"net/http"
 	"os"
+	"strings"
 	"time"

 	"github.com/gorilla/handlers"
@@ -14,6 +16,8 @@ import (

 	"github.com/cosmos/cosmos-sdk/client"
 	"github.com/cosmos/cosmos-sdk/server/config"
+	"github.com/cosmos/cosmos-sdk/telemetry"
+	"github.com/cosmos/cosmos-sdk/types/rest"

 	// unnamed import of statik for swagger UI support
 	_ "github.com/cosmos/cosmos-sdk/client/docs/statik"
@@ -25,6 +29,7 @@ type Server struct {
 	ClientCtx client.Context

 	logger   log.Logger
+	metrics  *telemetry.Metrics
 	listener net.Listener
 }

@@ -40,18 +45,28 @@ func New(clientCtx client.Context) *Server {
 // JSON RPC server. Configuration options are provided via config.APIConfig
 // and are delegated to the Tendermint JSON RPC server. The process is
 // non-blocking, so an external signal handler must be used.
-func (s *Server) Start(cfg config.APIConfig) error {
-	if cfg.Swagger {
+func (s *Server) Start(cfg config.Config) error {
+	if cfg.API.Swagger {
 		s.registerSwaggerUI()
 	}

-	tmCfg := tmrpcserver.DefaultConfig()
-	tmCfg.MaxOpenConnections = int(cfg.MaxOpenConnections)
-	tmCfg.ReadTimeout = time.Duration(cfg.RPCReadTimeout) * time.Second
-	tmCfg.WriteTimeout = time.Duration(cfg.RPCWriteTimeout) * time.Second
-	tmCfg.MaxBodyBytes = int64(cfg.RPCMaxBodyBytes)
+	if cfg.Telemetry.Enabled {
+		m, err := telemetry.New(cfg.Telemetry)
+		if err != nil {
+			return err
+		}

-	listener, err := tmrpcserver.Listen(cfg.Address, tmCfg)
+		s.metrics = m
+		s.registerMetrics()
+	}
+
+	tmCfg := tmrpcserver.DefaultConfig()
+	tmCfg.MaxOpenConnections = int(cfg.API.MaxOpenConnections)
+	tmCfg.ReadTimeout = time.Duration(cfg.API.RPCReadTimeout) * time.Second
+	tmCfg.WriteTimeout = time.Duration(cfg.API.RPCWriteTimeout) * time.Second
+	tmCfg.MaxBodyBytes = int64(cfg.API.RPCMaxBodyBytes)
+
+	listener, err := tmrpcserver.Listen(cfg.API.Address, tmCfg)
 	if err != nil {
 		return err
 	}
@@ -59,7 +74,7 @@ func (s *Server) Start(cfg config.APIConfig) error {
 	s.listener = listener
 	var h http.Handler = s.Router

-	if cfg.EnableUnsafeCORS {
+	if cfg.API.EnableUnsafeCORS {
 		return tmrpcserver.Serve(s.listener, handlers.CORS()(h), s.logger, tmCfg)
 	}

@@ -75,3 +90,20 @@ func (s *Server) registerSwaggerUI() {
 	staticServer := http.FileServer(statikFS)
 	s.Router.PathPrefix("/").Handler(staticServer)
 }
+
+func (s *Server) registerMetrics() {
+	metricsHandler := func(w http.ResponseWriter, r *http.Request) {
+		format := strings.TrimSpace(r.FormValue("format"))
+
+		gr, err := s.metrics.Gather(format)
+		if err != nil {
+			rest.WriteErrorResponse(w, http.StatusBadRequest, fmt.Sprintf("failed to gather metrics: %s", err))
+			return
+		}
+
+		w.Header().Set("Content-Type", gr.ContentType)
+		_, _ = w.Write(gr.Metrics)
+	}
+
+	s.Router.HandleFunc("/metrics", metricsHandler).Methods("GET")
+}
@@ -4,7 +4,10 @@ import (
 	"fmt"
 	"strings"

+	"github.com/spf13/viper"
+
 	"github.com/cosmos/cosmos-sdk/store"
+	"github.com/cosmos/cosmos-sdk/telemetry"
 	sdk "github.com/cosmos/cosmos-sdk/types"
 )

@@ -75,7 +78,9 @@ type APIConfig struct {
 type Config struct {
 	BaseConfig `mapstructure:",squash"`

-	API APIConfig `mapstructure:"api"`
+	// Telemetry defines the application telemetry configuration
+	Telemetry telemetry.Config `mapstructure:"telemetry"`
+	API       APIConfig        `mapstructure:"api"`
 }

 // SetMinGasPrices sets the validator's minimum gas prices.
@@ -115,6 +120,7 @@ func DefaultConfig() *Config {
 			PruningKeepEvery:     "0",
 			PruningSnapshotEvery: "0",
 		},
+		Telemetry: telemetry.Config{},
 		API: APIConfig{
 			Enable:             false,
 			Swagger:            false,
@@ -125,3 +131,34 @@ func DefaultConfig() *Config {
 		},
 	}
 }
+
+// GetConfig returns a fully parsed Config object.
+func GetConfig() Config {
+	return Config{
+		BaseConfig: BaseConfig{
+			MinGasPrices:         viper.GetString("minimum-gas-prices"),
+			InterBlockCache:      viper.GetBool("inter-block-cache"),
+			Pruning:              viper.GetString("pruning"),
+			PruningKeepEvery:     viper.GetString("pruning-keep-every"),
+			PruningSnapshotEvery: viper.GetString("pruning-snapshot-every"),
+			HaltHeight:           viper.GetUint64("halt-height"),
+			HaltTime:             viper.GetUint64("halt-time"),
+		},
+		Telemetry: telemetry.Config{
+			ServiceName:             viper.GetString("telemetry.service-name"),
+			Enabled:                 viper.GetBool("telemetry.enabled"),
+			EnableHostname:          viper.GetBool("telemetry.enable-hostname"),
+			EnableHostnameLabel:     viper.GetBool("telemetry.enable-hostname-label"),
+			EnableServiceLabel:      viper.GetBool("telemetry.enable-service-label"),
+			PrometheusRetentionTime: viper.GetInt64("telemetry.prometheus-retention-time"),
+		},
+		API: APIConfig{
+			Address:            viper.GetString("api.address"),
+			MaxOpenConnections: viper.GetUint("api.max-open-connections"),
+			RPCReadTimeout:     viper.GetUint("api.rpc-read-timeout"),
+			RPCWriteTimeout:    viper.GetUint("api.rpc-write-timeout"),
+			RPCMaxBodyBytes:    viper.GetUint("api.rpc-max-body-bytes"),
+			EnableUnsafeCORS:   viper.GetBool("api.enabled-unsafe-cors"),
+		},
+	}
+}
@@ -47,6 +47,32 @@ halt-time = {{ .BaseConfig.HaltTime }}
 # InterBlockCache enables inter-block caching.
 inter-block-cache = {{ .BaseConfig.InterBlockCache }}

+###############################################################################
+###                         Telemetry Configuration                         ###
+###############################################################################
+
+[telemetry]
+
+# Prefixed with keys to separate services
+service-name = "{{ .Telemetry.ServiceName }}"
+
+# Enabled enables the application telemetry functionality. When enabled,
+# an in-memory sink is also enabled by default. Operators may also enabled
+# other sinks such as Prometheus.
+enabled = {{ .Telemetry.Enabled }}
+
+# Enable prefixing gauge values with hostname
+enable-hostname = {{ .Telemetry.EnableHostname }}
+
+# Enable adding hostname to labels
+enable-hostname-label = {{ .Telemetry.EnableHostnameLabel }}
+
+# Enable adding service to labels
+enable-service-label = {{ .Telemetry.EnableServiceLabel }}
+
+# PrometheusRetentionTime, when positive, enables a Prometheus metrics sink.
+prometheus-retention-time = {{ .Telemetry.PrometheusRetentionTime }}
+
 ###############################################################################
 ###                           API Configuration                             ###
 ###############################################################################
@@ -209,18 +209,9 @@ func startInProcess(ctx *Context, cdc codec.JSONMarshaler, appCreator AppCreator
 			WithTrustNode(true)

 		apiSrv := api.New(ctx)
-		apiCfg := config.APIConfig{
-			Address:            viper.GetString("api.address"),
-			MaxOpenConnections: viper.GetUint("api.max-open-connections"),
-			RPCReadTimeout:     viper.GetUint("api.rpc-read-timeout"),
-			RPCWriteTimeout:    viper.GetUint("api.rpc-write-timeout"),
-			RPCMaxBodyBytes:    viper.GetUint("api.rpc-max-body-bytes"),
-			EnableUnsafeCORS:   viper.GetBool("api.enabled-unsafe-cors"),
-		}
-
 		app.RegisterAPIRoutes(apiSrv)

-		if err := apiSrv.Start(apiCfg); err != nil {
+		if err := apiSrv.Start(config.GetConfig()); err != nil {
 			return err
 		}
 	}