add some supernode metrics

Signed-off-by: yeya24 <[email protected]>

dfdaemon/constant/constant.go

@@ -46,8 +46,8 @@ const (
 )
 
 const (
-	// Namespace is the prefix of the metrics' name of dragonfly
+	// Namespace is the prefix of the metricsutils' name of dragonfly
 	Namespace = "dragonfly"
-	// Subsystem represents metrics for dfdaemon
+	// Subsystem represents metricsutils for dfdaemon
 	Subsystem = "dfdaemon"
 )

dfdaemon/handler/root_handler.go

@@ -32,6 +32,6 @@ func New() *http.ServeMux {
 	s.HandleFunc("/args", getArgs)
 	s.HandleFunc("/env", getEnv)
 	s.HandleFunc("/debug/version", version.Handler)
-	s.HandleFunc("/metrics", promhttp.Handler().ServeHTTP)
+	s.HandleFunc("/metricsutils", promhttp.Handler().ServeHTTP)
 	return s
 }

docs/user_guide/metrics.md

@@ -9,6 +9,13 @@ This doc contains all the metrics that Dragonfly components currently support. N
 - dragonfly_supernode_http_request_duration_seconds{code, handler, method} - http request latency in seconds
 - dragonfly_supernode_http_request_size_bytes{code, handler, method} - http request size in bytes
 - dragonfly_supernode_http_response_size_bytes{code, handler, method} - http response size in bytes
+- dragonfly_supernode_peers{hostname} - dragonfly peers
+- dragonfly_supernode_tasks{taskid, cdnstatus} - dragonfly tasks
+- dragonfly_supernode_dfgettasks{taskid, callsystem} - dragonfly dfget tasks
+- dragonfly_supernode_daemon_dfgettasks{taskid, callsystem} - dragonfly current dfget tasks, which are called from dfdaemon
+- dragonfly_supernode_schedule_duration_milliseconds{taskid} - duration for task scheduling in milliseconds
+- dragonfly_supernode_trigger_cdn_total{} - total times of triggering cdn.
+- dragonfly_supernode_trigger_cdn_failed_total{} - total failed times of triggering cdn.
 
 ## Dfdaemon
 

go.mod

@@ -38,6 +38,7 @@ require (
 	github.com/stretchr/testify v1.2.2
 	github.com/valyala/fasthttp v1.3.0
 	github.com/willf/bitset v0.0.0-20190228212526-18bd95f470f9
+	golang.org/x/net v0.0.0-20190620200207-3b0461eec859 // indirect
 	gopkg.in/gcfg.v1 v1.2.3
 	gopkg.in/mgo.v2 v2.0.0-20160818020120-3f83fa500528 // indirect
 	gopkg.in/warnings.v0 v0.1.2

go.sum

@@ -182,6 +182,8 @@ golang.org/x/net v0.0.0-20181220203305-927f97764cc3/go.mod h1:mL1N/T3taQHkDXs73r
 golang.org/x/net v0.0.0-20190311183353-d8887717615a/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
 golang.org/x/net v0.0.0-20190522155817-f3200d17e092 h1:4QSRKanuywn15aTZvI/mIDEgPQpswuFndXpOj3rKEco=
 golang.org/x/net v0.0.0-20190522155817-f3200d17e092/go.mod h1:HSz+uSET+XFnRR8LxR5pz3Of3rY3CfYBVs4xY44aLks=
+golang.org/x/net v0.0.0-20190620200207-3b0461eec859 h1:R/3boaszxrf1GEUWTVDzSKVwLmSJpwZ1yqXm8j0v2QI=
+golang.org/x/net v0.0.0-20190620200207-3b0461eec859/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/oauth2 v0.0.0-20180821212333-d2e6202438be/go.mod h1:N/0e6XlmueqKjAGxoOufVs8QHGRruUQn6yWY3a++T0U=
 golang.org/x/sync v0.0.0-20180314180146-1d60e4601c6f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20181108010431-42b317875d0f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=

pkg/util/metrics_util.go → pkg/metricsutils/metrics_util.go

@@ -14,7 +14,7 @@
  * limitations under the License.
  */
 
-package util
+package metricsutils
 
 import (
 	"github.com/prometheus/client_golang/prometheus"
@@ -26,6 +26,7 @@ const (
 )
 
 // NewCounter will auto-register a Counter metric to prometheus default registry and return it.
+// TODO(yeya24): Stop using default registry, add registry as a parameter instead.
 func NewCounter(subsystem, name, help string, labels []string) *prometheus.CounterVec {
 	return promauto.NewCounterVec(
 		prometheus.CounterOpts{

pkg/timeutils/time_util.go

@@ -24,3 +24,8 @@ import (
 func GetCurrentTimeMillis() int64 {
 	return time.Now().UnixNano() / time.Millisecond.Nanoseconds()
 }
+
+// GetCurrentTimeMillisFloat returns the time in millis for now in float64 format.
+func GetCurrentTimeMillisFloat() float64 {
+	return float64(time.Now().UnixNano()) / float64(time.Millisecond.Nanoseconds())
+}

supernode/config/constants.go

@@ -70,8 +70,8 @@ const (
 )
 
 const (
-	// SubsystemSupernode represents metrics from supernode
+	// SubsystemSupernode represents metricsutils from supernode
 	SubsystemSupernode = "supernode"
-	// SubsystemDfget represents metrics from dfget
+	// SubsystemDfget represents metricsutils from dfget
 	SubsystemDfget = "dfget"
 )

supernode/daemon/mgr/dfgettask/manager.go

@@ -19,6 +19,8 @@ package dfgettask
 import (
 	"context"
 	"fmt"
+	"github.com/dragonflyoss/Dragonfly/pkg/metricsutils"
+	"github.com/dragonflyoss/Dragonfly/supernode/config"
 
 	"github.com/dragonflyoss/Dragonfly/apis/types"
 	"github.com/dragonflyoss/Dragonfly/pkg/errortypes"
@@ -28,21 +30,39 @@ import (
 	dutil "github.com/dragonflyoss/Dragonfly/supernode/daemon/util"
 
 	"github.com/pkg/errors"
+	"github.com/prometheus/client_golang/prometheus"
 )
 
 var _ mgr.DfgetTaskMgr = &Manager{}
 
+type metrics struct {
+	dfgetTasks       *prometheus.GaugeVec
+	dfgetTasksDaemon *prometheus.GaugeVec
+}
+
+func newMetrics() *metrics {
+	return &metrics{
+		dfgetTasks: metricsutils.NewGauge(config.SubsystemSupernode, "dfgettasks",
+			"The number of dfget tasks", []string{"taskid", "callsystem"}),
+
+		dfgetTasksDaemon: metricsutils.NewGauge(config.SubsystemSupernode, "daemon_dfgettasks",
+			"The number of dfget tasks from dfdaemon", []string{"taskid", "callsystem"}),
+	}
+}
+
 // Manager is an implementation of the interface of DfgetTaskMgr.
 type Manager struct {
 	dfgetTaskStore *dutil.Store
 	ptoc           *syncmap.SyncMap
+	metrics        *metrics
 }
 
 // NewManager returns a new Manager.
 func NewManager() (*Manager, error) {
 	return &Manager{
 		dfgetTaskStore: dutil.NewStore(),
 		ptoc:           syncmap.NewSyncMap(),
+		metrics:        newMetrics(),
 	}, nil
 }
 
@@ -73,6 +93,13 @@ func (dtm *Manager) Add(ctx context.Context, dfgetTask *types.DfGetTask) error {
 
 	dtm.ptoc.Add(generatePeerKey(dfgetTask.PeerID, dfgetTask.TaskID), dfgetTask.CID)
 	dtm.dfgetTaskStore.Put(key, dfgetTask)
+
+	if dfgetTask.Dfdaemon {
+		dtm.metrics.dfgetTasksDaemon.WithLabelValues(dfgetTask.TaskID, dfgetTask.CallSystem).Inc()
+	} else {
+		dtm.metrics.dfgetTasks.WithLabelValues(dfgetTask.TaskID, dfgetTask.CallSystem).Inc()
+	}
+
 	return nil
 }
 
@@ -103,7 +130,11 @@ func (dtm *Manager) Delete(ctx context.Context, clientID, taskID string) error {
 		return err
 	}
 	dtm.ptoc.Delete(generatePeerKey(dfgetTask.PeerID, dfgetTask.TaskID))
-
+	if dfgetTask.Dfdaemon {
+		dtm.metrics.dfgetTasksDaemon.WithLabelValues(dfgetTask.TaskID, dfgetTask.CallSystem).Dec()
+	} else {
+		dtm.metrics.dfgetTasks.WithLabelValues(dfgetTask.TaskID, dfgetTask.CallSystem).Dec()
+	}
 	return dtm.dfgetTaskStore.Delete(key)
 }