feat(coordinator): add new metric get_task_counter tracking prover info (#1235)

Co-authored-by: colinlyguo <colinlyguo@users.noreply.github.com>

common/version/version.go

@@ -5,7 +5,7 @@ import (
 	"runtime/debug"
 )
 
-var tag = "v4.3.86"
+var tag = "v4.3.87"
 
 var commit = func() string {
 	if info, ok := debug.ReadBuildInfo(); ok {

coordinator/internal/controller/api/get_task.go

@@ -6,6 +6,8 @@ import (
 
 	"github.com/gin-gonic/gin"
 	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
+	"github.com/scroll-tech/go-ethereum/log"
 	"github.com/scroll-tech/go-ethereum/params"
 	"gorm.io/gorm"
 
@@ -21,6 +23,8 @@ import (
 // GetTaskController the get prover task api controller
 type GetTaskController struct {
 	proverTasks map[message.ProofType]provertask.ProverTask
+
+	getTaskAccessCounter *prometheus.CounterVec
 }
 
 // NewGetTaskController create a get prover task controller
@@ -30,6 +34,10 @@ func NewGetTaskController(cfg *config.Config, chainCfg *params.ChainConfig, db *
 
 	ptc := &GetTaskController{
 		proverTasks: make(map[message.ProofType]provertask.ProverTask),
+		getTaskAccessCounter: promauto.With(reg).NewCounterVec(prometheus.CounterOpts{
+			Name: "coordinator_get_task_access_count",
+			Help: "Multi dimensions get task counter.",
+		}, []string{coordinatorType.LabelProverName, coordinatorType.LabelProverPublicKey, coordinatorType.LabelProverVersion}),
 	}
 
 	ptc.proverTasks[message.ProofTypeChunk] = chunkProverTask
@@ -38,6 +46,28 @@ func NewGetTaskController(cfg *config.Config, chainCfg *params.ChainConfig, db *
 	return ptc
 }
 
+func (ptc *GetTaskController) incGetTaskAccessCounter(ctx *gin.Context) error {
+	publicKey, publicKeyExist := ctx.Get(coordinatorType.PublicKey)
+	if !publicKeyExist {
+		return fmt.Errorf("get public key from context failed")
+	}
+	proverName, proverNameExist := ctx.Get(coordinatorType.ProverName)
+	if !proverNameExist {
+		return fmt.Errorf("get prover name from context failed")
+	}
+	proverVersion, proverVersionExist := ctx.Get(coordinatorType.ProverVersion)
+	if !proverVersionExist {
+		return fmt.Errorf("get prover version from context failed")
+	}
+
+	ptc.getTaskAccessCounter.With(prometheus.Labels{
+		coordinatorType.LabelProverPublicKey: publicKey.(string),
+		coordinatorType.LabelProverName:      proverName.(string),
+		coordinatorType.LabelProverVersion:   proverVersion.(string),
+	}).Inc()
+	return nil
+}
+
 // GetTasks get assigned chunk/batch task
 func (ptc *GetTaskController) GetTasks(ctx *gin.Context) {
 	var getTaskParameter coordinatorType.GetTaskParameter
@@ -55,6 +85,10 @@ func (ptc *GetTaskController) GetTasks(ctx *gin.Context) {
 		return
 	}
 
+	if err := ptc.incGetTaskAccessCounter(ctx); err != nil {
+		log.Warn("get_task access counter inc failed", "error", err.Error())
+	}
+
 	result, err := proverTask.Assign(ctx, &getTaskParameter)
 	if err != nil {
 		nerr := fmt.Errorf("return prover task err:%w", err)

coordinator/internal/logic/provertask/batch_prover_task.go

@@ -31,6 +31,7 @@ type BatchProverTask struct {
 
 	batchAttemptsExceedTotal prometheus.Counter
 	batchTaskGetTaskTotal    *prometheus.CounterVec
+	batchTaskGetTaskProver   *prometheus.CounterVec
 }
 
 // NewBatchProverTask new a batch collector
@@ -58,6 +59,7 @@ func NewBatchProverTask(cfg *config.Config, chainCfg *params.ChainConfig, db *go
 			Name: "coordinator_batch_get_task_total",
 			Help: "Total number of batch get task.",
 		}, []string{"fork_name"}),
+		batchTaskGetTaskProver: newGetTaskCounterVec(promauto.With(reg), "batch"),
 	}
 	return bp
 }
@@ -180,6 +182,11 @@ func (bp *BatchProverTask) Assign(ctx *gin.Context, getTaskParameter *coordinato
 	}
 
 	bp.batchTaskGetTaskTotal.WithLabelValues(taskCtx.HardForkName).Inc()
+	bp.batchTaskGetTaskProver.With(prometheus.Labels{
+		coordinatorType.LabelProverName:      proverTask.ProverName,
+		coordinatorType.LabelProverPublicKey: proverTask.ProverPublicKey,
+		coordinatorType.LabelProverVersion:   proverTask.ProverVersion,
+	}).Inc()
 
 	return taskMsg, nil
 }

coordinator/internal/logic/provertask/chunk_prover_task.go

@@ -29,6 +29,7 @@ type ChunkProverTask struct {
 
 	chunkAttemptsExceedTotal prometheus.Counter
 	chunkTaskGetTaskTotal    *prometheus.CounterVec
+	chunkTaskGetTaskProver   *prometheus.CounterVec
 }
 
 // NewChunkProverTask new a chunk prover task
@@ -55,6 +56,7 @@ func NewChunkProverTask(cfg *config.Config, chainCfg *params.ChainConfig, db *go
 			Name: "coordinator_chunk_get_task_total",
 			Help: "Total number of chunk get task.",
 		}, []string{"fork_name"}),
+		chunkTaskGetTaskProver: newGetTaskCounterVec(promauto.With(reg), "chunk"),
 	}
 	return cp
 }
@@ -152,6 +154,11 @@ func (cp *ChunkProverTask) Assign(ctx *gin.Context, getTaskParameter *coordinato
 	}
 
 	cp.chunkTaskGetTaskTotal.WithLabelValues(taskCtx.HardForkName).Inc()
+	cp.chunkTaskGetTaskProver.With(prometheus.Labels{
+		coordinatorType.LabelProverName:      proverTask.ProverName,
+		coordinatorType.LabelProverPublicKey: proverTask.ProverPublicKey,
+		coordinatorType.LabelProverVersion:   proverTask.ProverVersion,
+	}).Inc()
 
 	return taskMsg, nil
 }

coordinator/internal/logic/provertask/prover_task.go

@@ -2,8 +2,11 @@ package provertask
 
 import (
 	"fmt"
+	"sync"
 
 	"github.com/gin-gonic/gin"
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
 	"github.com/scroll-tech/go-ethereum/log"
 	"gorm.io/gorm"
 
@@ -130,3 +133,22 @@ func (b *BaseProverTask) getHardForkNumberByName(forkName string) (uint64, error
 
 	return hardForkNumber, nil
 }
+
+var (
+	getTaskCounterInitOnce sync.Once
+	getTaskCounterVec      *prometheus.CounterVec = nil
+)
+
+func newGetTaskCounterVec(factory promauto.Factory, taskType string) *prometheus.CounterVec {
+	getTaskCounterInitOnce.Do(func() {
+		getTaskCounterVec = factory.NewCounterVec(prometheus.CounterOpts{
+			Name: "coordinator_get_task_count",
+			Help: "Multi dimensions get task counter.",
+		}, []string{"task_type",
+			coordinatorType.LabelProverName,
+			coordinatorType.LabelProverPublicKey,
+			coordinatorType.LabelProverVersion})
+	})
+
+	return getTaskCounterVec.MustCurryWith(prometheus.Labels{"task_type": taskType})
+}

coordinator/internal/types/metric.go

@@ -0,0 +1,10 @@
+package types
+
+var (
+	// LabelProverName label name for prover name; common label name using in prometheus metrics, same rule applies to below.
+	LabelProverName = "prover_name"
+	// LabelProverPublicKey label name for prover public key
+	LabelProverPublicKey = "prover_pubkey"
+	// LabelProverVersion label name for prover version
+	LabelProverVersion = "prover_version"
+)