From 88fe924c3c73016aaa4d94f96666bc1696c561d2 Mon Sep 17 00:00:00 2001
From: Brian Ginsburg <7957636+bgins@users.noreply.github.com>
Date: Thu, 26 Sep 2024 10:47:50 -0700
Subject: [PATCH] feat: Add solver HTTP tracing (#388)

* chore: Add solver tracer

* feat: Add solver HTTP tracing

* chore: Add solver disable telemetry envvar

* chore: Change print error message to warn log
---
 cmd/lilypad/resource-provider.go |  5 ++---
 cmd/lilypad/solver.go            | 20 +++++++++++++-------
 docker/docker-compose.dev.yml    |  1 +
 go.mod                           |  7 ++++---
 go.sum                           |  8 ++++++++
 pkg/options/solver.go            | 17 ++++++++++++++---
 pkg/solver/controller.go         |  4 ++++
 pkg/solver/server.go             |  5 ++++-
 pkg/solver/solver.go             | 16 ++++++++++------
 9 files changed, 60 insertions(+), 23 deletions(-)

diff --git a/cmd/lilypad/resource-provider.go b/cmd/lilypad/resource-provider.go
index e7f36765..437cdd91 100644
--- a/cmd/lilypad/resource-provider.go
+++ b/cmd/lilypad/resource-provider.go
@@ -1,13 +1,12 @@
 package lilypad
 
 import (
-	"fmt"
-
 	"github.com/lilypad-tech/lilypad/pkg/executor/bacalhau"
 	optionsfactory "github.com/lilypad-tech/lilypad/pkg/options"
 	"github.com/lilypad-tech/lilypad/pkg/resourceprovider"
 	"github.com/lilypad-tech/lilypad/pkg/system"
 	"github.com/lilypad-tech/lilypad/pkg/web3"
+	"github.com/rs/zerolog/log"
 	"github.com/spf13/cobra"
 )
 
@@ -41,7 +40,7 @@ func runResourceProvider(cmd *cobra.Command, options resourceprovider.ResourcePr
 
 	telemetry, err := configureTelemetry(commandCtx.Ctx, system.ResourceProviderService, network, options.Telemetry, options.Web3)
 	if err != nil {
-		fmt.Printf("failed to setup opentelemetry: %s", err)
+		log.Warn().Msgf("failed to setup opentelemetry: %s", err)
 	}
 	commandCtx.Cm.RegisterCallbackWithContext(telemetry.Shutdown)
 	tracer := telemetry.TracerProvider.Tracer(system.GetOTelServiceName(system.ResourceProviderService))
diff --git a/cmd/lilypad/solver.go b/cmd/lilypad/solver.go
index ebe8e4cd..6c6e0704 100644
--- a/cmd/lilypad/solver.go
+++ b/cmd/lilypad/solver.go
@@ -6,8 +6,8 @@ import (
 	memorystore "github.com/lilypad-tech/lilypad/pkg/solver/store/memory"
 	"github.com/lilypad-tech/lilypad/pkg/system"
 	"github.com/lilypad-tech/lilypad/pkg/web3"
+	"github.com/rs/zerolog/log"
 	"github.com/spf13/cobra"
-	"go.opentelemetry.io/otel/trace/noop"
 )
 
 func newSolverCmd() *cobra.Command {
@@ -25,7 +25,7 @@ func newSolverCmd() *cobra.Command {
 			if err != nil {
 				return err
 			}
-			return runSolver(cmd, options)
+			return runSolver(cmd, options, network)
 		},
 	}
 
@@ -34,12 +34,18 @@ func newSolverCmd() *cobra.Command {
 	return solverCmd
 }
 
-func runSolver(cmd *cobra.Command, options solver.SolverOptions) error {
+func runSolver(cmd *cobra.Command, options solver.SolverOptions, network string) error {
 	commandCtx := system.NewCommandContext(cmd)
 	defer commandCtx.Cleanup()
 
-	noopTracer := noop.NewTracerProvider().Tracer(system.GetOTelServiceName(system.SolverService))
-	web3SDK, err := web3.NewContractSDK(commandCtx.Ctx, options.Web3, noopTracer)
+	telemetry, err := configureTelemetry(commandCtx.Ctx, system.SolverService, network, options.Telemetry, options.Web3)
+	if err != nil {
+		log.Warn().Msgf("failed to setup opentelemetry: %s", err)
+	}
+	commandCtx.Cm.RegisterCallbackWithContext(telemetry.Shutdown)
+	tracer := telemetry.TracerProvider.Tracer(system.GetOTelServiceName(system.SolverService))
+
+	web3SDK, err := web3.NewContractSDK(commandCtx.Ctx, options.Web3, tracer)
 	if err != nil {
 		return err
 	}
@@ -49,12 +55,12 @@ func runSolver(cmd *cobra.Command, options solver.SolverOptions) error {
 		return err
 	}
 
-	solverService, err := solver.NewSolver(options, solverStore, web3SDK)
+	solverService, err := solver.NewSolver(options, solverStore, web3SDK, tracer)
 	if err != nil {
 		return err
 	}
 
-	solverErrors := solverService.Start(commandCtx.Ctx, commandCtx.Cm)
+	solverErrors := solverService.Start(commandCtx.Ctx, commandCtx.Cm, telemetry.TracerProvider)
 
 	for {
 		select {
diff --git a/docker/docker-compose.dev.yml b/docker/docker-compose.dev.yml
index 97e5fa93..ebd74726 100644
--- a/docker/docker-compose.dev.yml
+++ b/docker/docker-compose.dev.yml
@@ -54,6 +54,7 @@ services:
       - SERVER_PORT=${SERVER_PORT}
       - SERVER_URL=${SERVER_URL}
       - WEB3_RPC_URL=${WEB3_RPC_URL}
+      - DISABLE_TELEMETRY=${DISABLE_TELEMETRY}
     ports:
       - 8080:8080
     healthcheck:
diff --git a/go.mod b/go.mod
index cef06d7f..96a9fbeb 100644
--- a/go.mod
+++ b/go.mod
@@ -21,11 +21,11 @@ require (
 	github.com/spf13/cobra v1.8.0
 	github.com/stretchr/testify v1.9.0
 	github.com/theckman/yacspin v0.13.12
-	go.opentelemetry.io/otel v1.28.0
+	go.opentelemetry.io/otel v1.30.0
 	go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.28.0
 	go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp v1.28.0
 	go.opentelemetry.io/otel/sdk v1.28.0
-	go.opentelemetry.io/otel/trace v1.28.0
+	go.opentelemetry.io/otel/trace v1.30.0
 	gorgonia.org/cu v0.9.7-0.20240623234718-3cd40db700e9
 	k8s.io/apimachinery v0.29.0
 )
@@ -268,12 +268,13 @@ require (
 	github.com/yashtewari/glob-intersection v0.2.0 // indirect
 	go.etcd.io/bbolt v1.3.8 // indirect
 	go.opencensus.io v0.24.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/github.com/gorilla/mux/otelmux v0.55.0 // indirect
 	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.47.0 // indirect
 	go.opentelemetry.io/otel/exporters/otlp/otlpmetric v0.40.0 // indirect
 	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v0.40.0 // indirect
 	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp v1.24.0 // indirect
 	go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.21.0 // indirect
-	go.opentelemetry.io/otel/metric v1.28.0 // indirect
+	go.opentelemetry.io/otel/metric v1.30.0 // indirect
 	go.opentelemetry.io/otel/sdk/metric v1.24.0 // indirect
 	go.opentelemetry.io/proto/otlp v1.3.1 // indirect
 	go.ptx.dk/multierrgroup v0.0.3 // indirect
diff --git a/go.sum b/go.sum
index 2cd6cf49..d49d274d 100644
--- a/go.sum
+++ b/go.sum
@@ -910,10 +910,14 @@ go.opencensus.io v0.22.4/go.mod h1:yxeiOL68Rb0Xd1ddK5vPZ/oVn4vY4Ynel7k9FzqtOIw=
 go.opencensus.io v0.22.5/go.mod h1:5pWMHQbX5EPX2/62yrJeAkowc+lfs/XD7Uxpq3pI6kk=
 go.opencensus.io v0.24.0 h1:y73uSU6J157QMP2kn2r30vwW1A2W2WFwSCGnAVxeaD0=
 go.opencensus.io v0.24.0/go.mod h1:vNK8G9p7aAivkbmorf4v+7Hgx+Zs0yY+0fOtgBfjQKo=
+go.opentelemetry.io/contrib/instrumentation/github.com/gorilla/mux/otelmux v0.55.0 h1:lRMfRoJnAaDWadgR58z6xyMIaH5UQ5SzFOhA+LfmUkA=
+go.opentelemetry.io/contrib/instrumentation/github.com/gorilla/mux/otelmux v0.55.0/go.mod h1:7kJADjE+s91WUhMkzN7qTnDf2aUJZPzeD7RKw/uB5Xg=
 go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.47.0 h1:sv9kVfal0MK0wBMCOGr+HeJm9v803BkJxGrk2au7j08=
 go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.47.0/go.mod h1:SK2UL73Zy1quvRPonmOmRDiWk1KBV3LyIeeIxcEApWw=
 go.opentelemetry.io/otel v1.28.0 h1:/SqNcYk+idO0CxKEUOtKQClMK/MimZihKYMruSMViUo=
 go.opentelemetry.io/otel v1.28.0/go.mod h1:q68ijF8Fc8CnMHKyzqL6akLO46ePnjkgfIMIjUIX9z4=
+go.opentelemetry.io/otel v1.30.0 h1:F2t8sK4qf1fAmY9ua4ohFS/K+FUuOPemHUIXHtktrts=
+go.opentelemetry.io/otel v1.30.0/go.mod h1:tFw4Br9b7fOS+uEao81PJjVMjW/5fvNCbpsDIXqP0pc=
 go.opentelemetry.io/otel/exporters/otlp/otlpmetric v0.40.0 h1:MZbjiZeMmn5wFMORhozpouGKDxj9POHTuU5UA8msBQk=
 go.opentelemetry.io/otel/exporters/otlp/otlpmetric v0.40.0/go.mod h1:C7tOYVCJmrDTCwxNny0MuUtnDIR3032vFHYke0F2ZrU=
 go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v0.40.0 h1:q3FNPi8FLQVjLlmV+WWHQfH9ZCCtQIS0O/+dn1+4cJ4=
@@ -928,12 +932,16 @@ go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp v1.28.0 h1:j9+03
 go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp v1.28.0/go.mod h1:Y5+XiUG4Emn1hTfciPzGPJaSI+RpDts6BnCIir0SLqk=
 go.opentelemetry.io/otel/metric v1.28.0 h1:f0HGvSl1KRAU1DLgLGFjrwVyismPlnuU6JD6bOeuA5Q=
 go.opentelemetry.io/otel/metric v1.28.0/go.mod h1:Fb1eVBFZmLVTMb6PPohq3TO9IIhUisDsbJoL/+uQW4s=
+go.opentelemetry.io/otel/metric v1.30.0 h1:4xNulvn9gjzo4hjg+wzIKG7iNFEaBMX00Qd4QIZs7+w=
+go.opentelemetry.io/otel/metric v1.30.0/go.mod h1:aXTfST94tswhWEb+5QjlSqG+cZlmyXy/u8jFpor3WqQ=
 go.opentelemetry.io/otel/sdk v1.28.0 h1:b9d7hIry8yZsgtbmM0DKyPWMMUMlK9NEKuIG4aBqWyE=
 go.opentelemetry.io/otel/sdk v1.28.0/go.mod h1:oYj7ClPUA7Iw3m+r7GeEjz0qckQRJK2B8zjcZEfu7Pg=
 go.opentelemetry.io/otel/sdk/metric v1.24.0 h1:yyMQrPzF+k88/DbH7o4FMAs80puqd+9osbiBrJrz/w8=
 go.opentelemetry.io/otel/sdk/metric v1.24.0/go.mod h1:I6Y5FjH6rvEnTTAYQz3Mmv2kl6Ek5IIrmwTLqMrrOE0=
 go.opentelemetry.io/otel/trace v1.28.0 h1:GhQ9cUuQGmNDd5BTCP2dAvv75RdMxEfTmYejp+lkx9g=
 go.opentelemetry.io/otel/trace v1.28.0/go.mod h1:jPyXzNPg6da9+38HEwElrQiHlVMTnVfM3/yv2OlIHaI=
+go.opentelemetry.io/otel/trace v1.30.0 h1:7UBkkYzeg3C7kQX8VAidWh2biiQbtAKjyIML8dQ9wmc=
+go.opentelemetry.io/otel/trace v1.30.0/go.mod h1:5EyKqTzzmyqB9bwtCCq6pDLktPK6fmGf/Dph+8VI02o=
 go.opentelemetry.io/proto/otlp v1.3.1 h1:TrMUixzpM0yuc/znrFTP9MMRh8trP93mkCiDVeXrui0=
 go.opentelemetry.io/proto/otlp v1.3.1/go.mod h1:0X1WI4de4ZsLrrJNLAQbFeLCm3T7yBkR0XqQ7niQU+8=
 go.ptx.dk/multierrgroup v0.0.3 h1:HNaevFVERSZ7/DyCAnOICWF86B8s+76QwdwOlgCPvQM=
diff --git a/pkg/options/solver.go b/pkg/options/solver.go
index 09595eb2..8f963ff8 100644
--- a/pkg/options/solver.go
+++ b/pkg/options/solver.go
@@ -8,9 +8,10 @@ import (
 
 func NewSolverOptions() solver.SolverOptions {
 	options := solver.SolverOptions{
-		Server:   GetDefaultServerOptions(),
-		Web3:     GetDefaultWeb3Options(),
-		Services: GetDefaultServicesOptions(),
+		Server:    GetDefaultServerOptions(),
+		Web3:      GetDefaultWeb3Options(),
+		Services:  GetDefaultServicesOptions(),
+		Telemetry: GetDefaultTelemetryOptions(),
 	}
 	options.Web3.Service = system.SolverService
 	return options
@@ -20,6 +21,7 @@ func AddSolverCliFlags(cmd *cobra.Command, options *solver.SolverOptions) {
 	AddWeb3CliFlags(cmd, &options.Web3)
 	AddServerCliFlags(cmd, &options.Server)
 	AddServicesCliFlags(cmd, &options.Services)
+	AddTelemetryCliFlags(cmd, &options.Telemetry)
 }
 
 func CheckSolverOptions(options solver.SolverOptions) error {
@@ -31,6 +33,10 @@ func CheckSolverOptions(options solver.SolverOptions) error {
 	if err != nil {
 		return err
 	}
+	err = CheckTelemetryOptions(options.Telemetry)
+	if err != nil {
+		return err
+	}
 	return nil
 }
 
@@ -40,5 +46,10 @@ func ProcessSolverOptions(options solver.SolverOptions, network string) (solver.
 		return options, err
 	}
 	options.Web3 = newWeb3Options
+	newTelemetryOptions, err := ProcessTelemetryOptions(options.Telemetry, network)
+	if err != nil {
+		return options, err
+	}
+	options.Telemetry = newTelemetryOptions
 	return options, CheckSolverOptions(options)
 }
diff --git a/pkg/solver/controller.go b/pkg/solver/controller.go
index f0a1934b..30447eef 100644
--- a/pkg/solver/controller.go
+++ b/pkg/solver/controller.go
@@ -14,6 +14,7 @@ import (
 	"github.com/lilypad-tech/lilypad/pkg/web3/bindings/mediation"
 	"github.com/lilypad-tech/lilypad/pkg/web3/bindings/storage"
 	"github.com/rs/zerolog/log"
+	"go.opentelemetry.io/otel/trace"
 )
 
 // add an enum for various types of event
@@ -48,6 +49,7 @@ type SolverController struct {
 	solverEventSubs []func(SolverEvent)
 	options         SolverOptions
 	log             *system.ServiceLogger
+	tracer          trace.Tracer
 }
 
 // the background "even if we have not heard of an event" loop
@@ -60,6 +62,7 @@ func NewSolverController(
 	web3SDK *web3.Web3SDK,
 	store store.SolverStore,
 	options SolverOptions,
+	tracer trace.Tracer,
 ) (*SolverController, error) {
 	controller := &SolverController{
 		web3SDK:    web3SDK,
@@ -67,6 +70,7 @@ func NewSolverController(
 		store:      store,
 		options:    options,
 		log:        system.NewServiceLogger(system.SolverService),
+		tracer:     tracer,
 	}
 	return controller, nil
 }
diff --git a/pkg/solver/server.go b/pkg/solver/server.go
index 16147b38..7e52ab70 100644
--- a/pkg/solver/server.go
+++ b/pkg/solver/server.go
@@ -18,6 +18,8 @@ import (
 	"github.com/lilypad-tech/lilypad/pkg/solver/store"
 	"github.com/lilypad-tech/lilypad/pkg/system"
 	"github.com/rs/zerolog/log"
+	"go.opentelemetry.io/contrib/instrumentation/github.com/gorilla/mux/otelmux"
+	"go.opentelemetry.io/otel/sdk/trace"
 )
 
 type solverServer struct {
@@ -56,12 +58,13 @@ func NewSolverServer(
  *
 */
 
-func (solverServer *solverServer) ListenAndServe(ctx context.Context, cm *system.CleanupManager) error {
+func (solverServer *solverServer) ListenAndServe(ctx context.Context, cm *system.CleanupManager, tracerProvider *trace.TracerProvider) error {
 	router := mux.NewRouter()
 
 	subrouter := router.PathPrefix(http.API_SUB_PATH).Subrouter()
 
 	subrouter.Use(http.CorsMiddleware)
+	subrouter.Use(otelmux.Middleware("solver", otelmux.WithTracerProvider(tracerProvider)))
 
 	subrouter.HandleFunc("/job_offers", http.GetHandler(solverServer.getJobOffers)).Methods("GET")
 	subrouter.HandleFunc("/job_offers", http.PostHandler(solverServer.addJobOffer)).Methods("POST")
diff --git a/pkg/solver/solver.go b/pkg/solver/solver.go
index fedd5efc..32e5b2fe 100644
--- a/pkg/solver/solver.go
+++ b/pkg/solver/solver.go
@@ -9,12 +9,15 @@ import (
 	"github.com/lilypad-tech/lilypad/pkg/system"
 	"github.com/lilypad-tech/lilypad/pkg/web3"
 	"github.com/rs/zerolog/log"
+	sdkTrace "go.opentelemetry.io/otel/sdk/trace"
+	"go.opentelemetry.io/otel/trace"
 )
 
 type SolverOptions struct {
-	Web3     web3.Web3Options
-	Server   http.ServerOptions
-	Services data.ServiceConfig
+	Web3      web3.Web3Options
+	Server    http.ServerOptions
+	Services  data.ServiceConfig
+	Telemetry system.TelemetryOptions
 }
 
 type Solver struct {
@@ -29,8 +32,9 @@ func NewSolver(
 	options SolverOptions,
 	store store.SolverStore,
 	web3SDK *web3.Web3SDK,
+	tracer trace.Tracer,
 ) (*Solver, error) {
-	controller, err := NewSolverController(web3SDK, store, options)
+	controller, err := NewSolverController(web3SDK, store, options, tracer)
 	if err != nil {
 		return nil, err
 	}
@@ -48,11 +52,11 @@ func NewSolver(
 	return solver, nil
 }
 
-func (solver *Solver) Start(ctx context.Context, cm *system.CleanupManager) chan error {
+func (solver *Solver) Start(ctx context.Context, cm *system.CleanupManager, tracerProvider *sdkTrace.TracerProvider) chan error {
 	errorChan := solver.controller.Start(ctx, cm)
 	log.Debug().Msgf("solver.server.ListenAndServe")
 	go func() {
-		err := solver.server.ListenAndServe(ctx, cm)
+		err := solver.server.ListenAndServe(ctx, cm, tracerProvider)
 		if err != nil {
 			errorChan <- err
 		}