Relocation opencensus metrics to its own package and add more node stats

chain/sync.go

@@ -23,6 +23,7 @@ import (
 	"github.com/libp2p/go-libp2p-core/peer"
 	cbg "github.com/whyrusleeping/cbor-gen"
 	"github.com/whyrusleeping/pubsub"
+	"go.opencensus.io/stats"
 	"go.opencensus.io/trace"
 	"golang.org/x/xerrors"
 
@@ -38,6 +39,7 @@ import (
 	"github.com/filecoin-project/lotus/chain/store"
 	"github.com/filecoin-project/lotus/chain/types"
 	"github.com/filecoin-project/lotus/lib/sigs"
+	"github.com/filecoin-project/lotus/metrics"
 )
 
 var log = logging.Logger("chain")
@@ -394,6 +396,9 @@ func (syncer *Syncer) Sync(ctx context.Context, maybeHead *types.TipSet) error {
 		)
 	}
 
+	// Record current network chain height when sync is called
+	stats.Record(ctx, metrics.ChainHeight.M(int64(maybeHead.Height())))
+
 	if syncer.store.GetHeaviestTipSet().ParentWeight().GreaterThan(maybeHead.ParentWeight()) {
 		return nil
 	}
@@ -1038,6 +1043,8 @@ func (syncer *Syncer) syncMessagesAndCheckState(ctx context.Context, headers []*
 			return xerrors.Errorf("message processing failed: %w", err)
 		}
 
+		// Set current node sync height
+		stats.Record(ctx, metrics.ChainNodeHeight.M(int64(fts.TipSet().Height())))
 		ss.SetHeight(fts.TipSet().Height())
 
 		return nil

cmd/lotus/daemon.go

@@ -24,6 +24,7 @@ import (
 	"github.com/filecoin-project/lotus/chain/stmgr"
 	"github.com/filecoin-project/lotus/chain/store"
 	"github.com/filecoin-project/lotus/chain/vm"
+	"github.com/filecoin-project/lotus/metrics"
 	"github.com/filecoin-project/lotus/node"
 	"github.com/filecoin-project/lotus/node/modules"
 	"github.com/filecoin-project/lotus/node/modules/testing"
@@ -36,12 +37,6 @@ const (
 	preSealedSectorsFlag = "genesis-presealed-sectors"
 )
 
-var (
-	lotusInfo  = stats.Int64("info", "Arbitrary counter to tag lotus info to", stats.UnitDimensionless)
-	version, _ = tag.NewKey("version")
-	commit, _  = tag.NewKey("commit")
-)
-
 // DaemonCmd is the `go-lotus daemon` command
 var DaemonCmd = &cli.Command{
 	Name:  "daemon",
@@ -99,7 +94,7 @@ var DaemonCmd = &cli.Command{
 			defer pprof.StopCPUProfile()
 		}
 
-		ctx, _ := tag.New(context.Background(), tag.Insert(version, build.BuildVersion), tag.Insert(commit, build.CurrentCommit))
+		ctx, _ := tag.New(context.Background(), tag.Insert(metrics.Version, build.BuildVersion), tag.Insert(metrics.Commit, build.CurrentCommit))
 		{
 			dir, err := homedir.Expand(cctx.String("repo"))
 			if err != nil {
@@ -180,21 +175,15 @@ var DaemonCmd = &cli.Command{
 			return xerrors.Errorf("initializing node: %w", err)
 		}
 
-		// We are using this metric to tag info about lotus even though
-		// it doesn't contain any actual metrics
+		// Register all metric views
 		if err = view.Register(
-			&view.View{
-				Name:        "info",
-				Description: "Lotus node information",
-				Measure:     lotusInfo,
-				Aggregation: view.LastValue(),
-				TagKeys:     []tag.Key{version, commit},
-			},
+			metrics.DefaultViews...,
 		); err != nil {
 			log.Fatalf("Cannot register the view: %v", err)
 		}
+
 		// Set the metric to one so it is published to the exporter
-		stats.Record(ctx, lotusInfo.M(1))
+		stats.Record(ctx, metrics.LotusInfo.M(1))
 
 		endpoint, err := r.APIEndpoint()
 		if err != nil {

lib/jsonrpc/handler.go

@@ -9,6 +9,9 @@ import (
 	"io"
 	"reflect"
 
+	"github.com/filecoin-project/lotus/metrics"
+	"go.opencensus.io/stats"
+	"go.opencensus.io/tag"
 	"go.opencensus.io/trace"
 	"go.opencensus.io/trace/propagation"
 	"golang.org/x/xerrors"
@@ -151,12 +154,15 @@ func (handlers) getSpan(ctx context.Context, req request) (context.Context, *tra
 }
 
 func (h handlers) handle(ctx context.Context, req request, w func(func(io.Writer)), rpcError rpcErrFunc, done func(keepCtx bool), chOut chanOut) {
+	// Not sure if we need to sanitize the incoming req.Method or not.
 	ctx, span := h.getSpan(ctx, req)
+	ctx, _ = tag.New(context.Background(), tag.Insert(metrics.RPCMethod, req.Method))
 	defer span.End()
 
 	handler, ok := h[req.Method]
 	if !ok {
 		rpcError(w, &req, rpcMethodNotFound, fmt.Errorf("method '%s' not found", req.Method))
+		stats.Record(ctx, metrics.RPCInvalidMethod.M(1))
 		done(false)
 		return
 	}
@@ -201,6 +207,7 @@ func (h handlers) handle(ctx context.Context, req request, w func(func(io.Writer
 	if req.ID == nil {
 		return // notification
 	}
+	stats.Record(ctx, metrics.RPCRequestSuccess.M(1))
 
 	///////////////////
 
@@ -245,5 +252,6 @@ func (h handlers) handle(ctx context.Context, req request, w func(func(io.Writer
 			log.Error(err)
 			return
 		}
+		stats.Record(ctx, metrics.RPCResponseSuccess.M(1))
 	})
 }

metrics/metrics.go

@@ -0,0 +1,64 @@
+package metrics
+
+import (
+	"go.opencensus.io/stats"
+	"go.opencensus.io/stats/view"
+	"go.opencensus.io/tag"
+)
+
+// Global Tags
+var (
+	Version, _   = tag.NewKey("version")
+	Commit, _    = tag.NewKey("commit")
+	RPCMethod, _ = tag.NewKey("method")
+)
+
+// Measures
+var (
+	LotusInfo          = stats.Int64("info", "Arbitrary counter to tag lotus info to", stats.UnitDimensionless)
+	ChainHeight        = stats.Int64("chain/height", "Current Height of the chain", stats.UnitDimensionless)
+	ChainNodeHeight    = stats.Int64("chain/node_height", "Current Height of the node", stats.UnitDimensionless)
+	PeerCount          = stats.Int64("peer/count", "Current number of FIL peers", stats.UnitDimensionless)
+	RPCInvalidMethod   = stats.Int64("rpc/invalid_method", "Total number of invalid RPC methods called", stats.UnitDimensionless)
+	RPCRequestSuccess  = stats.Int64("rpc/request_success", "Total number of successful requests handled", stats.UnitDimensionless)
+	RPCResponseSuccess = stats.Int64("rpc/response_success", "Total number of succeessful responses handled", stats.UnitDimensionless)
+)
+
+// DefaultViews is an array of Consensus views for metric gathering purposes
+var DefaultViews = []*view.View{
+	&view.View{
+		Name:        "info",
+		Description: "Lotus node information",
+		Measure:     LotusInfo,
+		Aggregation: view.LastValue(),
+		TagKeys:     []tag.Key{Version, Commit},
+	},
+	&view.View{
+		Measure:     ChainHeight,
+		Aggregation: view.LastValue(),
+	},
+	&view.View{
+		Measure:     ChainNodeHeight,
+		Aggregation: view.LastValue(),
+	},
+	// All RPC related metrics should at the very least tag the RPCMethod
+	&view.View{
+		Measure:     RPCInvalidMethod,
+		Aggregation: view.Count(),
+		TagKeys:     []tag.Key{RPCMethod},
+	},
+	&view.View{
+		Measure:     RPCRequestSuccess,
+		Aggregation: view.Count(),
+		TagKeys:     []tag.Key{RPCMethod},
+	},
+	&view.View{
+		Measure:     RPCResponseSuccess,
+		Aggregation: view.Count(),
+		TagKeys:     []tag.Key{RPCMethod},
+	},
+	&view.View{
+		Measure:     PeerCount,
+		Aggregation: view.LastValue(),
+	},
+}

peermgr/peermgr.go

@@ -5,7 +5,9 @@ import (
 	"sync"
 	"time"
 
+	"github.com/filecoin-project/lotus/metrics"
 	"github.com/filecoin-project/lotus/node/modules/dtypes"
+	"go.opencensus.io/stats"
 	"go.uber.org/fx"
 
 	host "github.com/libp2p/go-libp2p-core/host"
@@ -115,6 +117,7 @@ func (pmgr *PeerMgr) Run(ctx context.Context) {
 			} else if pcount > pmgr.maxFilPeers {
 				log.Debug("peer count about threshold: %d > %d", pcount, pmgr.maxFilPeers)
 			}
+			stats.Record(ctx, metrics.PeerCount.M(int64(pmgr.getPeerCount())))
 		}
 	}
 }