stats: Add RPC event for blocking for a picker update

clientconn.go

@@ -349,7 +349,7 @@ func (cc *ClientConn) exitIdleMode() error {
 	cc.idlenessState = ccIdlenessStateExitingIdle
 	exitedIdle := false
 	if cc.blockingpicker == nil {
-		cc.blockingpicker = newPickerWrapper()
+		cc.blockingpicker = newPickerWrapper(cc.dopts.copts.StatsHandlers)
 	} else {
 		cc.blockingpicker.exitIdleMode()
 		exitedIdle = true

picker_wrapper.go

@@ -28,21 +28,26 @@ import (
 	"google.golang.org/grpc/internal/channelz"
 	istatus "google.golang.org/grpc/internal/status"
 	"google.golang.org/grpc/internal/transport"
+	"google.golang.org/grpc/stats"
 	"google.golang.org/grpc/status"
 )
 
 // pickerWrapper is a wrapper of balancer.Picker. It blocks on certain pick
 // actions and unblock when there's a picker update.
 type pickerWrapper struct {
-	mu         sync.Mutex
-	done       bool
-	idle       bool
-	blockingCh chan struct{}
-	picker     balancer.Picker
+	mu            sync.Mutex
+	done          bool
+	idle          bool
+	blockingCh    chan struct{}
+	picker        balancer.Picker
+	statsHandlers []stats.Handler // to record blocking picker calls
 }
 
-func newPickerWrapper() *pickerWrapper {
-	return &pickerWrapper{blockingCh: make(chan struct{})}
+func newPickerWrapper(statsHandlers []stats.Handler) *pickerWrapper {
+	return &pickerWrapper{
+		blockingCh:    make(chan struct{}),
+		statsHandlers: statsHandlers,
+	}
 }
 
 // updatePicker is called by UpdateBalancerState. It unblocks all blocked pick.
@@ -125,6 +130,9 @@ func (pw *pickerWrapper) pick(ctx context.Context, failfast bool, info balancer.
 					return nil, balancer.PickResult{}, status.Error(codes.Canceled, errStr)
 				}
 			case <-ch:
+				for _, sh := range pw.statsHandlers {
+					sh.HandleRPC(ctx, &stats.PickerUpdated{})
+				}
 			}
 			continue
 		}

picker_wrapper_test.go

@@ -66,7 +66,7 @@ func (p *testingPicker) Pick(info balancer.PickInfo) (balancer.PickResult, error
 }
 
 func (s) TestBlockingPickTimeout(t *testing.T) {
-	bp := newPickerWrapper()
+	bp := newPickerWrapper(nil)
 	ctx, cancel := context.WithTimeout(context.Background(), time.Millisecond)
 	defer cancel()
 	if _, _, err := bp.pick(ctx, true, balancer.PickInfo{}); status.Code(err) != codes.DeadlineExceeded {
@@ -75,7 +75,7 @@ func (s) TestBlockingPickTimeout(t *testing.T) {
 }
 
 func (s) TestBlockingPick(t *testing.T) {
-	bp := newPickerWrapper()
+	bp := newPickerWrapper(nil)
 	// All goroutines should block because picker is nil in bp.
 	var finishedCount uint64
 	for i := goroutineCount; i > 0; i-- {
@@ -94,7 +94,7 @@ func (s) TestBlockingPick(t *testing.T) {
 }
 
 func (s) TestBlockingPickNoSubAvailable(t *testing.T) {
-	bp := newPickerWrapper()
+	bp := newPickerWrapper(nil)
 	var finishedCount uint64
 	bp.updatePicker(&testingPicker{err: balancer.ErrNoSubConnAvailable, maxCalled: goroutineCount})
 	// All goroutines should block because picker returns no subConn available.
@@ -114,7 +114,7 @@ func (s) TestBlockingPickNoSubAvailable(t *testing.T) {
 }
 
 func (s) TestBlockingPickTransientWaitforready(t *testing.T) {
-	bp := newPickerWrapper()
+	bp := newPickerWrapper(nil)
 	bp.updatePicker(&testingPicker{err: balancer.ErrTransientFailure, maxCalled: goroutineCount})
 	var finishedCount uint64
 	// All goroutines should block because picker returns transientFailure and
@@ -135,7 +135,7 @@ func (s) TestBlockingPickTransientWaitforready(t *testing.T) {
 }
 
 func (s) TestBlockingPickSCNotReady(t *testing.T) {
-	bp := newPickerWrapper()
+	bp := newPickerWrapper(nil)
 	bp.updatePicker(&testingPicker{sc: testSCNotReady, maxCalled: goroutineCount})
 	var finishedCount uint64
 	// All goroutines should block because subConn is not ready.

stats/opencensus/trace.go

@@ -99,9 +99,11 @@ func populateSpan(ctx context.Context, rs stats.RPCStats, ti *traceInfo) {
 			trace.BoolAttribute("Client", rs.Client),
 			trace.BoolAttribute("FailFast", rs.FailFast),
 		)
+	case *stats.PickerUpdated:
+		span.Annotate(nil, "Delayed LB pick complete")
 	case *stats.InPayload:
 		// message id - "must be calculated as two different counters starting
-		// from 1 one for sent messages and one for received messages."
+		// from one for sent messages and one for received messages."
 		mi := atomic.AddUint32(&ti.countRecvMsg, 1)
 		span.AddMessageReceiveEvent(int64(mi), int64(rs.Length), int64(rs.CompressedLength))
 	case *stats.OutPayload:

stats/stats.go

@@ -59,6 +59,16 @@ func (s *Begin) IsClient() bool { return s.Client }
 
 func (s *Begin) isRPCStats() {}
 
+// PickerUpdated represents an event that the picker finished picking if the
+// ClientConn blocks on picker picking while performing a RPC.
+type PickerUpdated struct{}
+
+// IsClient indicates if the stats information is from client side. Only Client
+// Side interfaces with a Picker, thus always returns true.
+func (*PickerUpdated) IsClient() bool { return true }
+
+func (*PickerUpdated) isRPCStats() {}
+
 // InPayload contains the information for an incoming payload.
 type InPayload struct {
 	// Client is true if this InPayload is from client side.

test/end2end_test.go

@@ -23,6 +23,7 @@ import (
 	"bytes"
 	"context"
 	"crypto/tls"
+	"encoding/json"
 	"errors"
 	"flag"
 	"fmt"
@@ -44,6 +45,8 @@ import (
 	"golang.org/x/net/http2"
 	"golang.org/x/net/http2/hpack"
 	"google.golang.org/grpc"
+	"google.golang.org/grpc/balancer"
+	"google.golang.org/grpc/balancer/roundrobin"
 	"google.golang.org/grpc/codes"
 	"google.golang.org/grpc/connectivity"
 	"google.golang.org/grpc/credentials"
@@ -62,6 +65,7 @@ import (
 	"google.golang.org/grpc/peer"
 	"google.golang.org/grpc/resolver"
 	"google.golang.org/grpc/resolver/manual"
+	"google.golang.org/grpc/serviceconfig"
 	"google.golang.org/grpc/stats"
 	"google.golang.org/grpc/status"
 	"google.golang.org/grpc/tap"
@@ -82,6 +86,7 @@ const defaultHealthService = "grpc.health.v1.Health"
 
 func init() {
 	channelz.TurnOn()
+	balancer.Register(blockingPickerBalancerBuilder{})
 }
 
 type s struct {
@@ -6362,3 +6367,165 @@ func (s) TestGlobalBinaryLoggingOptions(t *testing.T) {
 		t.Fatalf("want 8 server side binary logging events, got %v", ssbl.mml.events)
 	}
 }
+
+type statsHandlerRecordEvents struct {
+	mu sync.Mutex
+	s  []stats.RPCStats
+}
+
+func (*statsHandlerRecordEvents) TagRPC(ctx context.Context, _ *stats.RPCTagInfo) context.Context {
+	return ctx
+}
+func (h *statsHandlerRecordEvents) HandleRPC(_ context.Context, s stats.RPCStats) {
+	h.mu.Lock()
+	h.s = append(h.s, s)
+	h.mu.Unlock()
+}
+func (*statsHandlerRecordEvents) TagConn(ctx context.Context, _ *stats.ConnTagInfo) context.Context {
+	return ctx
+}
+func (*statsHandlerRecordEvents) HandleConn(context.Context, stats.ConnStats) {}
+
+type blockingPicker struct {
+	pickDone func()
+}
+
+func (bp *blockingPicker) Pick(pi balancer.PickInfo) (balancer.PickResult, error) {
+	defer bp.pickDone()
+	return balancer.PickResult{}, balancer.ErrNoSubConnAvailable
+}
+
+const name = "blockingPickerBalancer"
+
+type blockingPickerBalancerBuilder struct{}
+
+func (blockingPickerBalancerBuilder) Build(cc balancer.ClientConn, bOpts balancer.BuildOptions) balancer.Balancer {
+	b := &blockingPickerBalancer{
+		ClientConn: cc,
+	}
+	// round_robin child to complete balancer tree with a usable leaf policy and
+	// have RPCs actually work.
+	builder := balancer.Get(roundrobin.Name)
+	if builder == nil {
+		// Shouldn't happen, defensive programming. Registered from import of
+		// roundrobin package.
+		return nil
+	}
+	rr := builder.Build(b, bOpts)
+	if rr == nil {
+		// Shouldn't happen, defensive programming.
+		return nil
+	}
+	b.Balancer = rr
+	return b
+}
+
+func (blockingPickerBalancerBuilder) ParseConfig(json.RawMessage) (serviceconfig.LoadBalancingConfig, error) {
+	return &bpbConfig{}, nil
+}
+
+func (blockingPickerBalancerBuilder) Name() string {
+	return name
+}
+
+type bpbConfig struct {
+	serviceconfig.LoadBalancingConfig
+}
+
+type blockingPickerBalancer struct {
+	stateMu    sync.Mutex
+	childState balancer.State
+
+	blockingPickerDone *grpcsync.Event
+	// embed a ClientConn to wrap only UpdateState() operation
+	balancer.ClientConn
+	// embed a Balancer to wrap only UpdateClientConnState() operation
+	balancer.Balancer
+}
+
+func (bpb *blockingPickerBalancer) UpdateClientConnState(s balancer.ClientConnState) error {
+	bpb.blockingPickerDone = grpcsync.NewEvent()
+	err := bpb.Balancer.UpdateClientConnState(balancer.ClientConnState{
+		ResolverState: s.ResolverState,
+	})
+	bpb.ClientConn.UpdateState(balancer.State{
+		ConnectivityState: connectivity.Connecting,
+		Picker: &blockingPicker{
+			pickDone: func() {
+				bpb.blockingPickerDone.Fire()
+				bpb.stateMu.Lock()
+				cs := bpb.childState
+				bpb.stateMu.Unlock()
+				bpb.ClientConn.UpdateState(cs)
+			},
+		},
+	})
+	return err
+}
+
+func (bpb *blockingPickerBalancer) UpdateState(state balancer.State) {
+	bpb.stateMu.Lock()
+	bpb.childState = state
+	bpb.stateMu.Unlock()
+	if bpb.blockingPickerDone.HasFired() { // guard first one to get a picker sending ErrNoSubConnAvailable first
+		bpb.ClientConn.UpdateState(state) // after the first rr picker update, cc will trigger more, so actually forward these
+	}
+}
+
+// TestPickerBlockingStatsCall tests the emission of a stats handler call that
+// represents the picker had to block due to ErrNoSubConnAvailable being
+// returned from the first picker call.
+func (s) TestPickerBlockingStatsCall(t *testing.T) {
+	sh := &statsHandlerRecordEvents{}
+	ss := &stubserver.StubServer{
+		UnaryCallF: func(ctx context.Context, in *testpb.SimpleRequest) (*testpb.SimpleResponse, error) {
+			return &testpb.SimpleResponse{}, nil
+		},
+	}
+
+	if err := ss.StartServer(); err != nil {
+		t.Fatalf("Error starting endpoint server: %v", err)
+	}
+	defer ss.Stop()
+
+	lbCfgJSON := `{
+  		"loadBalancingConfig": [
+    		{
+      			"blockingPickerBalancer": {}
+    	}
+  	]
+	}`
+
+	sc := internal.ParseServiceConfig.(func(string) *serviceconfig.ParseResult)(lbCfgJSON)
+	mr := manual.NewBuilderWithScheme("pickerupdatedbalancer")
+	defer mr.Close()
+	print("ss.Address when setting: ", ss.Address)
+	mr.InitialState(resolver.State{
+		Addresses: []resolver.Address{
+			{Addr: ss.Address},
+		},
+		ServiceConfig: sc,
+	})
+
+	cc, err := grpc.Dial(mr.Scheme()+":///", grpc.WithResolvers(mr), grpc.WithStatsHandler(sh), grpc.WithTransportCredentials(insecure.NewCredentials()))
+	if err != nil {
+		t.Fatalf("grpc.Dial() failed: %v", err)
+	}
+	defer cc.Close()
+	ctx, cancel := context.WithTimeout(context.Background(), defaultTestTimeout)
+	defer cancel()
+	testServiceClient := testgrpc.NewTestServiceClient(cc)
+	if _, err := testServiceClient.UnaryCall(ctx, &testpb.SimpleRequest{}); err != nil {
+		t.Fatalf("Unexpected error from UnaryCall: %v", err)
+	}
+
+	var pickerUpdatedCount uint
+	for _, stat := range sh.s {
+		if _, ok := stat.(*stats.PickerUpdated); ok {
+			pickerUpdatedCount++
+		}
+	}
+	if pickerUpdatedCount == 0 {
+		t.Fatalf("sh.pickerUpdated count: %v, want: >0", pickerUpdatedCount)
+	}
+}

test/retry_test.go

@@ -474,6 +474,10 @@ func (*retryStatsHandler) TagRPC(ctx context.Context, _ *stats.RPCTagInfo) conte
 	return ctx
 }
 func (h *retryStatsHandler) HandleRPC(_ context.Context, s stats.RPCStats) {
+	// these calls come in nondeterministically - so can just ignore
+	if _, ok := s.(*stats.PickerUpdated); ok {
+		return
+	}
 	h.mu.Lock()
 	h.s = append(h.s, s)
 	h.mu.Unlock()