Ensure everyone sees everyone else (#1275)

bootstrap/peer.go

@@ -124,8 +124,10 @@ func New(log *zap.Logger, full *identity.FullIdentity, db DB, config Config) (*P
 			return nil, errs.Combine(err, peer.Close())
 		}
 
+		peer.Transport = peer.Transport.WithObservers(peer.Kademlia.RoutingTable)
+
 		// TODO: reduce number of arguments
-		peer.Kademlia.Service, err = kademlia.NewService(peer.Log.Named("kademlia"), self, config.BootstrapNodes(), peer.Identity, config.Alpha, peer.Kademlia.RoutingTable)
+		peer.Kademlia.Service, err = kademlia.NewService(peer.Log.Named("kademlia"), self, config.BootstrapNodes(), peer.Transport, config.Alpha, peer.Kademlia.RoutingTable)
 		if err != nil {
 			return nil, errs.Combine(err, peer.Close())
 		}
@@ -148,7 +150,7 @@ func (peer *Peer) Run(ctx context.Context) error {
 		return ignoreCancel(peer.Kademlia.Service.Bootstrap(ctx))
 	})
 	group.Go(func() error {
-		return ignoreCancel(peer.Kademlia.Service.RunRefresh(ctx))
+		return ignoreCancel(peer.Kademlia.Service.Run(ctx))
 	})
 	group.Go(func() error {
 		// TODO: move the message into Server instead

internal/sync2/cycle.go

@@ -124,6 +124,8 @@ func (cycle *Cycle) Run(ctx context.Context, fn func(ctx context.Context) error)
 
 // Close closes all resources associated with it.
 func (cycle *Cycle) Close() {
+	cycle.Stop()
+	<-cycle.stop
 	close(cycle.control)
 }
 

internal/testplanet/planet.go

@@ -214,12 +214,54 @@ func (planet *Planet) Start(ctx context.Context) {
 
 	planet.started = true
 
+	planet.Bootstrap.Kademlia.Service.WaitForBootstrap()
+
 	for _, peer := range planet.Satellites {
 		peer.Kademlia.Service.WaitForBootstrap()
 	}
 	for _, peer := range planet.StorageNodes {
 		peer.Kademlia.Service.WaitForBootstrap()
 	}
+
+	planet.Reconnect(ctx)
+}
+
+// Reconnect reconnects all nodes with each other.
+func (planet *Planet) Reconnect(ctx context.Context) {
+	log := planet.log.Named("reconnect")
+
+	var group errgroup.Group
+
+	// TODO: instead of pinging try to use Lookups or natural discovery to ensure
+	// everyone finds everyone else
+
+	for _, storageNode := range planet.StorageNodes {
+		storageNode := storageNode
+		group.Go(func() error {
+			_, err := storageNode.Kademlia.Service.Ping(ctx, planet.Bootstrap.Local())
+			if err != nil {
+				log.Error("storage node did not find bootstrap", zap.Error(err))
+			}
+			return nil
+		})
+	}
+
+	for _, satellite := range planet.Satellites {
+		satellite := satellite
+		group.Go(func() error {
+			for _, storageNode := range planet.StorageNodes {
+				_, err := satellite.Kademlia.Service.Ping(ctx, storageNode.Local())
+				if err != nil {
+					log.Error("satellite did not find storage node", zap.Error(err))
+				}
+			}
+
+			satellite.Discovery.Service.Refresh.TriggerWait()
+			return nil
+		})
+	}
+
+	_ = group.Wait() // none of the goroutines return an error
 }
 
 // StopPeer stops a single peer in the planet
@@ -298,6 +340,7 @@ func (planet *Planet) newUplinks(prefix string, count, storageNodeCount int) ([]
 
 // newSatellites initializes satellites
 func (planet *Planet) newSatellites(count int) ([]*satellite.Peer, error) {
+	// TODO: move into separate file
 	var xs []*satellite.Peer
 	defer func() {
 		for _, x := range xs {
@@ -355,7 +398,6 @@ func (planet *Planet) newSatellites(count int) ([]*satellite.Peer, error) {
 				},
 			},
 			Overlay: overlay.Config{
-				RefreshInterval: 30 * time.Second,
 				Node: overlay.NodeSelectionConfig{
 					UptimeRatio:           0,
 					UptimeCount:           0,
@@ -425,6 +467,7 @@ func (planet *Planet) newSatellites(count int) ([]*satellite.Peer, error) {
 
 // newStorageNodes initializes storage nodes
 func (planet *Planet) newStorageNodes(count int) ([]*storagenode.Peer, error) {
+	// TODO: move into separate file
 	var xs []*storagenode.Peer
 	defer func() {
 		for _, x := range xs {
@@ -508,6 +551,7 @@ func (planet *Planet) newStorageNodes(count int) ([]*storagenode.Peer, error) {
 
 // newBootstrap initializes the bootstrap node
 func (planet *Planet) newBootstrap() (peer *bootstrap.Peer, err error) {
+	// TODO: move into separate file
 	defer func() {
 		planet.peers = append(planet.peers, closablePeer{peer: peer})
 	}()

internal/testplanet/uplink_test.go

@@ -6,7 +6,6 @@ package testplanet_test
 import (
 	"crypto/rand"
 	"testing"
-	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -25,7 +24,6 @@ func TestUploadDownload(t *testing.T) {
 	defer ctx.Check(planet.Shutdown)
 
 	planet.Start(ctx)
-	time.Sleep(2 * time.Second)
 
 	expectedData := make([]byte, 5*memory.MiB)
 	_, err = rand.Read(expectedData)

pkg/accounting/rollup/rollup_test.go

@@ -22,8 +22,6 @@ func TestQuery(t *testing.T) {
 		atRest := float64(5000)
 		bw := []int64{1000, 2000, 3000, 4000}
 
-		time.Sleep(time.Second * 2)
-
 		nodeData, bwTotals := createData(planet, atRest, bw)
 
 		// Set timestamp back by the number of days we want to save

pkg/audit/cursor_test.go

@@ -9,7 +9,6 @@ import (
 	"math/big"
 	"reflect"
 	"testing"
-	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -27,9 +26,6 @@ func TestAuditSegment(t *testing.T) {
 	testplanet.Run(t, testplanet.Config{
 		SatelliteCount: 1, StorageNodeCount: 4, UplinkCount: 1,
 	}, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) {
-		// we wait a second for all the nodes to complete bootstrapping off the satellite
-		time.Sleep(2 * time.Second)
-
 		type pathCount struct {
 			path  storj.Path
 			count int

pkg/discovery/service.go

@@ -10,7 +10,9 @@ import (
 
 	"github.com/zeebo/errs"
 	"go.uber.org/zap"
+	"golang.org/x/sync/errgroup"
 
+	"storj.io/storj/internal/sync2"
 	"storj.io/storj/pkg/kademlia"
 	"storj.io/storj/pkg/overlay"
 	"storj.io/storj/pkg/statdb"
@@ -38,69 +40,68 @@ type Discovery struct {
 	cache  *overlay.Cache
 	kad    *kademlia.Kademlia
 	statdb statdb.DB
-	config Config
 
 	// refreshOffset tracks the offset of the current refresh cycle
 	refreshOffset int64
+	refreshLimit  int
+
+	Refresh   sync2.Cycle
+	Graveyard sync2.Cycle
+	Discovery sync2.Cycle
 }
 
 // New returns a new discovery service.
 func New(logger *zap.Logger, ol *overlay.Cache, kad *kademlia.Kademlia, stat statdb.DB, config Config) *Discovery {
-	return &Discovery{
+	discovery := &Discovery{
 		log:    logger,
 		cache:  ol,
 		kad:    kad,
 		statdb: stat,
-		config: config,
 
 		refreshOffset: 0,
+		refreshLimit:  config.RefreshLimit,
 	}
-}
 
-// NewDiscovery Returns a new Discovery instance with cache, kad, and statdb loaded on
+	discovery.Refresh.SetInterval(config.RefreshInterval)
-func NewDiscovery(logger *zap.Logger, ol *overlay.Cache, kad *kademlia.Kademlia, stat statdb.DB, config Config) *Discovery {
+	discovery.Graveyard.SetInterval(config.GraveyardInterval)
-	return &Discovery{
+	discovery.Discovery.SetInterval(config.DiscoveryInterval)
-		log:    logger,
+
-		cache:  ol,
+	return discovery
-		kad:    kad,
-		statdb: stat,
-		config: config,
-	}
 }
 
 // Close closes resources
-func (discovery *Discovery) Close() error { return nil }
+func (discovery *Discovery) Close() error {
+	discovery.Refresh.Close()
+	discovery.Graveyard.Close()
+	discovery.Discovery.Close()
+	return nil
+}
 
 // Run runs the discovery service
 func (discovery *Discovery) Run(ctx context.Context) error {
-	refresh := time.NewTicker(discovery.config.RefreshInterval)
+	var group errgroup.Group
-	graveyard := time.NewTicker(discovery.config.GraveyardInterval)
+	discovery.Refresh.Start(ctx, &group, func(ctx context.Context) error {
-	discover := time.NewTicker(discovery.config.DiscoveryInterval)
+		err := discovery.refresh(ctx)
-	defer refresh.Stop()
+		if err != nil {
-	defer graveyard.Stop()
+			discovery.log.Error("error with cache refresh: ", zap.Error(err))
-	defer discover.Stop()
-
-	for {
-		select {
-		case <-refresh.C:
-			err := discovery.refresh(ctx)
-			if err != nil {
-				discovery.log.Error("error with cache refresh: ", zap.Error(err))
-			}
-		case <-discover.C:
-			err := discovery.discover(ctx)
-			if err != nil {
-				discovery.log.Error("error with cache discovery: ", zap.Error(err))
-			}
-		case <-graveyard.C:
-			err := discovery.searchGraveyard(ctx)
-			if err != nil {
-				discovery.log.Error("graveyard resurrection failed: ", zap.Error(err))
-			}
-		case <-ctx.Done():
-			return ctx.Err()
 		}
-	}
+		return nil
+	})
+	discovery.Graveyard.Start(ctx, &group, func(ctx context.Context) error {
+		err := discovery.searchGraveyard(ctx)
+		if err != nil {
+			discovery.log.Error("graveyard resurrection failed: ", zap.Error(err))
+		}
+		return nil
+	})
+	discovery.Discovery.Start(ctx, &group, func(ctx context.Context) error {
+		err := discovery.discover(ctx)
+		if err != nil {
+			discovery.log.Error("error with cache discovery: ", zap.Error(err))
+		}
+		return nil
+	})
+	return group.Wait()
 }
 
 // refresh updates the cache db with the current DHT.
@@ -114,7 +115,7 @@ func (discovery *Discovery) refresh(ctx context.Context) error {
 		}
 	}
 
-	list, more, err := discovery.cache.Paginate(ctx, discovery.refreshOffset, discovery.config.RefreshLimit)
+	list, more, err := discovery.cache.Paginate(ctx, discovery.refreshOffset, discovery.refreshLimit)
 	if err != nil {
 		return Error.Wrap(err)
 	}

pkg/discovery/service_test.go

@@ -5,7 +5,6 @@ package discovery_test
 
 import (
 	"testing"
-	"time"
 
 	"github.com/stretchr/testify/assert"
 
@@ -17,8 +16,6 @@ func TestCache_Refresh(t *testing.T) {
 	testplanet.Run(t, testplanet.Config{
 		SatelliteCount: 1, StorageNodeCount: 10, UplinkCount: 0,
 	}, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) {
-		time.Sleep(5 * time.Second)
-
 		satellite := planet.Satellites[0]
 		for _, storageNode := range planet.StorageNodes {
 			node, err := satellite.Overlay.Service.Get(ctx, storageNode.ID())

pkg/kademlia/dialer_test.go

@@ -6,7 +6,6 @@ package kademlia_test
 import (
 	"fmt"
 	"testing"
-	"time"
 
 	"github.com/zeebo/errs"
 	"go.uber.org/zap/zaptest"
@@ -23,7 +22,6 @@ func TestDialer(t *testing.T) {
 	testplanet.Run(t, testplanet.Config{
 		SatelliteCount: 1, StorageNodeCount: 4, UplinkCount: 3,
 	}, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) {
-		time.Sleep(2 * time.Second)
 		expectedKademliaEntries := 1 + len(planet.Satellites) + len(planet.StorageNodes)
 
 		// TODO: also use satellites

pkg/kademlia/kademlia.go

@@ -47,22 +47,21 @@ type Kademlia struct {
 	routingTable   *RoutingTable
 	bootstrapNodes []pb.Node
 	dialer         *Dialer
-	identity       *identity.FullIdentity
 	lookups        sync2.WorkGroup
 
 	bootstrapFinished sync2.Fence
+
+	RefreshBuckets sync2.Cycle
 }
 
 // NewService returns a newly configured Kademlia instance
-func NewService(log *zap.Logger, self pb.Node, bootstrapNodes []pb.Node, identity *identity.FullIdentity, alpha int, rt *RoutingTable) (*Kademlia, error) {
+func NewService(log *zap.Logger, self pb.Node, bootstrapNodes []pb.Node, transport transport.Client, alpha int, rt *RoutingTable) (*Kademlia, error) {
 	k := &Kademlia{
 		log:            log,
 		alpha:          alpha,
 		routingTable:   rt,
 		bootstrapNodes: bootstrapNodes,
-		identity:       identity,
+		dialer:         NewDialer(log.Named("dialer"), transport),
-
-		dialer: NewDialer(log.Named("dialer"), transport.NewClient(identity, rt)),
 	}
 
 	return k, nil
@@ -283,25 +282,21 @@ func (k *Kademlia) Seen() []*pb.Node {
 	return nodes
 }
 
-// RunRefresh occasionally refreshes stale kad buckets
+// Run occasionally refreshes stale kad buckets
-func (k *Kademlia) RunRefresh(ctx context.Context) error {
+func (k *Kademlia) Run(ctx context.Context) error {
 	if !k.lookups.Start() {
 		return context.Canceled
 	}
 	defer k.lookups.Done()
 
-	ticker := time.NewTicker(5 * time.Minute)
+	k.RefreshBuckets.SetInterval(5 * time.Minute)
-	for {
+	return k.RefreshBuckets.Run(ctx, func(ctx context.Context) error {
-		if err := k.refresh(ctx, time.Minute); err != nil {
+		err := k.refresh(ctx, time.Minute)
+		if err != nil {
 			k.log.Warn("bucket refresh failed", zap.Error(err))
 		}
-		select {
+		return nil
-		case <-ticker.C:
+	})
-		case <-ctx.Done():
-			ticker.Stop()
-			return ctx.Err()
-		}
-	}
 }
 
 // refresh updates each Kademlia bucket not contacted in the last hour

pkg/kademlia/kademlia_test.go

@@ -29,6 +29,7 @@ import (
 	"storj.io/storj/pkg/identity"
 	"storj.io/storj/pkg/pb"
 	"storj.io/storj/pkg/storj"
+	"storj.io/storj/pkg/transport"
 	"storj.io/storj/storage/teststore"
 )
 
@@ -543,5 +544,5 @@ func newKademlia(log *zap.Logger, nodeType pb.NodeType, bootstrapNodes []pb.Node
 		return nil, BootstrapErr.Wrap(err)
 	}
 
-	return NewService(log, self, bootstrapNodes, identity, alpha, rt)
+	return NewService(log, self, bootstrapNodes, transport.NewClient(identity, rt), alpha, rt)
 }

pkg/metainfo/kvmetainfo/objects_test.go

@@ -10,7 +10,6 @@ import (
 	"io"
 	mathrand "math/rand"
 	"testing"
-	"time"
 
 	"github.com/stretchr/testify/assert"
 
@@ -120,9 +119,6 @@ func TestGetObject(t *testing.T) {
 
 func TestGetObjectStream(t *testing.T) {
 	runTest(t, func(ctx context.Context, planet *testplanet.Planet, db *kvmetainfo.DB, buckets buckets.Store, streams streams.Store) {
-		// we wait a second for all the nodes to complete bootstrapping off the satellite
-		time.Sleep(2 * time.Second)
-
 		data := make([]byte, 32*memory.KB)
 		_, err := rand.Read(data)
 		if !assert.NoError(t, err) {

pkg/miniogw/integration_test.go

@@ -83,8 +83,6 @@ func TestUploadDownload(t *testing.T) {
 
 	planet.Start(ctx)
 
-	time.Sleep(2 * time.Second)
-
 	// create identity for gateway
 	ca, err := testidentity.NewTestCA(ctx)
 	assert.NoError(t, err)

pkg/overlay/client_test.go

@@ -5,7 +5,6 @@ package overlay_test
 
 import (
 	"testing"
-	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -23,9 +22,6 @@ func TestChoose(t *testing.T) {
 	testplanet.Run(t, testplanet.Config{
 		SatelliteCount: 1, StorageNodeCount: 8, UplinkCount: 1,
 	}, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) {
-		// we wait a second for all the nodes to complete bootstrapping off the satellite
-		time.Sleep(2 * time.Second)
-
 		oc, err := planet.Uplinks[0].DialOverlay(planet.Satellites[0])
 		require.NoError(t, err)
 
@@ -63,9 +59,6 @@ func TestLookup(t *testing.T) {
 	testplanet.Run(t, testplanet.Config{
 		SatelliteCount: 1, StorageNodeCount: 4, UplinkCount: 1,
 	}, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) {
-		// we wait a second for all the nodes to complete bootstrapping off the satellite
-		time.Sleep(2 * time.Second)
-
 		oc, err := planet.Uplinks[0].DialOverlay(planet.Satellites[0])
 		require.NoError(t, err)
 
@@ -105,9 +98,6 @@ func TestBulkLookup(t *testing.T) {
 	testplanet.Run(t, testplanet.Config{
 		SatelliteCount: 1, StorageNodeCount: 4, UplinkCount: 1,
 	}, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) {
-		// we wait a second for all the nodes to complete bootstrapping off the satellite
-		time.Sleep(2 * time.Second)
-
 		oc, err := planet.Uplinks[0].DialOverlay(planet.Satellites[0])
 		require.NoError(t, err)
 
@@ -143,9 +133,6 @@ func TestBulkLookupV2(t *testing.T) {
 	testplanet.Run(t, testplanet.Config{
 		SatelliteCount: 1, StorageNodeCount: 4, UplinkCount: 1,
 	}, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) {
-		// we wait a second for all the nodes to complete bootstrapping off the satellite
-		time.Sleep(2 * time.Second)
-
 		oc, err := planet.Uplinks[0].DialOverlay(planet.Satellites[0])
 		require.NoError(t, err)
 

pkg/overlay/config.go

@@ -5,7 +5,6 @@ package overlay
 
 import (
 	"strings"
-	"time"
 
 	"github.com/zeebo/errs"
 	monkit "gopkg.in/spacemonkeygo/monkit.v2"
@@ -23,8 +22,7 @@ var (
 // Config is a configuration struct for everything you need to start the
 // Overlay cache responsibility.
 type Config struct {
-	RefreshInterval time.Duration `help:"the interval at which the cache refreshes itself in seconds" default:"1s"`
+	Node NodeSelectionConfig
-	Node            NodeSelectionConfig
 }
 
 // LookupConfig is a configuration struct for querying the overlay cache with one or more node IDs

pkg/overlay/server_test.go

@@ -5,7 +5,6 @@ package overlay_test
 
 import (
 	"testing"
-	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -22,9 +21,6 @@ func TestServer(t *testing.T) {
 	testplanet.Run(t, testplanet.Config{
 		SatelliteCount: 1, StorageNodeCount: 4, UplinkCount: 1,
 	}, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) {
-		// we wait a second for all the nodes to complete bootstrapping off the satellite
-		time.Sleep(2 * time.Second)
-
 		satellite := planet.Satellites[0]
 		server := satellite.Overlay.Endpoint
 		// TODO: handle cleanup
@@ -67,9 +63,6 @@ func TestNodeSelection(t *testing.T) {
 		var err error
 		satellite := planet.Satellites[0]
 
-		// we wait a second for all the nodes to complete bootstrapping off the satellite
-		time.Sleep(2 * time.Second)
-
 		// This sets a reputable audit count for a certain number of nodes.
 		for i, node := range planet.StorageNodes {
 			for k := 0; k < i; k++ {

pkg/transport/transport.go

@@ -36,6 +36,7 @@ type Client interface {
 	DialNode(ctx context.Context, node *pb.Node, opts ...grpc.DialOption) (*grpc.ClientConn, error)
 	DialAddress(ctx context.Context, address string, opts ...grpc.DialOption) (*grpc.ClientConn, error)
 	Identity() *identity.FullIdentity
+	WithObservers(obs ...Observer) *Transport
 }
 
 // Transport interface structure
@@ -109,6 +110,14 @@ func (transport *Transport) Identity() *identity.FullIdentity {
 	return transport.identity
 }
 
+// WithObservers returns a new transport including the listed observers.
+func (transport *Transport) WithObservers(obs ...Observer) *Transport {
+	tr := &Transport{identity: transport.identity}
+	tr.observers = append(tr.observers, transport.observers...)
+	tr.observers = append(tr.observers, obs...)
+	return tr
+}
+
 func alertFail(ctx context.Context, obs []Observer, node *pb.Node, err error) {
 	for _, o := range obs {
 		o.ConnFailure(ctx, node, err)

satellite/peer.go

@@ -198,6 +198,28 @@ func New(log *zap.Logger, full *identity.FullIdentity, db DB, config *Config) (*
 		}
 	}
 
+	{ // setup overlay
+		config := config.Overlay
+		peer.Overlay.Service = overlay.NewCache(peer.DB.OverlayCache(), peer.DB.StatDB())
+
+		// TODO: should overlay service be wired up to transport?
+
+		nodeSelectionConfig := &overlay.NodeSelectionConfig{
+			UptimeCount:           config.Node.UptimeCount,
+			UptimeRatio:           config.Node.UptimeRatio,
+			AuditSuccessRatio:     config.Node.AuditSuccessRatio,
+			AuditCount:            config.Node.AuditCount,
+			NewNodeAuditThreshold: config.Node.NewNodeAuditThreshold,
+			NewNodePercentage:     config.Node.NewNodePercentage,
+		}
+
+		peer.Overlay.Endpoint = overlay.NewServer(peer.Log.Named("overlay:endpoint"), peer.Overlay.Service, nodeSelectionConfig)
+		pb.RegisterOverlayServer(peer.Public.Server.GRPC(), peer.Overlay.Endpoint)
+
+		peer.Overlay.Inspector = overlay.NewInspector(peer.Overlay.Service)
+		pb.RegisterOverlayInspectorServer(peer.Public.Server.GRPC(), peer.Overlay.Inspector)
+	}
+
 	{ // setup kademlia
 		config := config.Kademlia
 		// TODO: move this setup logic into kademlia package
@@ -236,10 +258,12 @@ func New(log *zap.Logger, full *identity.FullIdentity, db DB, config *Config) (*
 			if err != nil {
 				return nil, errs.Combine(err, peer.Close())
 			}
+
+			peer.Transport = peer.Transport.WithObservers(peer.Kademlia.RoutingTable)
 		}
 
 		// TODO: reduce number of arguments
-		peer.Kademlia.Service, err = kademlia.NewService(peer.Log.Named("kademlia"), self, config.BootstrapNodes(), peer.Identity, config.Alpha, peer.Kademlia.RoutingTable)
+		peer.Kademlia.Service, err = kademlia.NewService(peer.Log.Named("kademlia"), self, config.BootstrapNodes(), peer.Transport, config.Alpha, peer.Kademlia.RoutingTable)
 		if err != nil {
 			return nil, errs.Combine(err, peer.Close())
 		}
@@ -251,26 +275,6 @@ func New(log *zap.Logger, full *identity.FullIdentity, db DB, config *Config) (*
 		pb.RegisterKadInspectorServer(peer.Public.Server.GRPC(), peer.Kademlia.Inspector)
 	}
 
-	{ // setup overlay
-		config := config.Overlay
-		peer.Overlay.Service = overlay.NewCache(peer.DB.OverlayCache(), peer.DB.StatDB())
-
-		nodeSelectionConfig := &overlay.NodeSelectionConfig{
-			UptimeCount:           config.Node.UptimeCount,
-			UptimeRatio:           config.Node.UptimeRatio,
-			AuditSuccessRatio:     config.Node.AuditSuccessRatio,
-			AuditCount:            config.Node.AuditCount,
-			NewNodeAuditThreshold: config.Node.NewNodeAuditThreshold,
-			NewNodePercentage:     config.Node.NewNodePercentage,
-		}
-
-		peer.Overlay.Endpoint = overlay.NewServer(peer.Log.Named("overlay:endpoint"), peer.Overlay.Service, nodeSelectionConfig)
-		pb.RegisterOverlayServer(peer.Public.Server.GRPC(), peer.Overlay.Endpoint)
-
-		peer.Overlay.Inspector = overlay.NewInspector(peer.Overlay.Service)
-		pb.RegisterOverlayInspectorServer(peer.Public.Server.GRPC(), peer.Overlay.Inspector)
-	}
-
 	{ // setup reputation
 		// TODO: find better structure with overlay
 		peer.Reputation.Inspector = statdb.NewInspector(peer.DB.StatDB())
@@ -385,7 +389,7 @@ func (peer *Peer) Run(ctx context.Context) error {
 		return ignoreCancel(peer.Kademlia.Service.Bootstrap(ctx))
 	})
 	group.Go(func() error {
-		return ignoreCancel(peer.Kademlia.Service.RunRefresh(ctx))
+		return ignoreCancel(peer.Kademlia.Service.Run(ctx))
 	})
 	group.Go(func() error {
 		return ignoreCancel(peer.Discovery.Service.Run(ctx))
@@ -464,13 +468,6 @@ func (peer *Peer) Close() error {
 		errlist.Add(peer.Discovery.Service.Close())
 	}
 
-	if peer.Overlay.Endpoint != nil {
-		errlist.Add(peer.Overlay.Endpoint.Close())
-	}
-	if peer.Overlay.Service != nil {
-		errlist.Add(peer.Overlay.Service.Close())
-	}
-
 	// TODO: add kademlia.Endpoint for consistency
 	if peer.Kademlia.Service != nil {
 		errlist.Add(peer.Kademlia.Service.Close())
@@ -479,6 +476,13 @@ func (peer *Peer) Close() error {
 		errlist.Add(peer.Kademlia.RoutingTable.Close())
 	}
 
+	if peer.Overlay.Endpoint != nil {
+		errlist.Add(peer.Overlay.Endpoint.Close())
+	}
+	if peer.Overlay.Service != nil {
+		errlist.Add(peer.Overlay.Service.Close())
+	}
+
 	if peer.Kademlia.ndb != nil || peer.Kademlia.kdb != nil {
 		errlist.Add(peer.Kademlia.kdb.Close())
 		errlist.Add(peer.Kademlia.ndb.Close())

storagenode/peer.go

@@ -142,8 +142,10 @@ func New(log *zap.Logger, full *identity.FullIdentity, db DB, config Config) (*P
 			return nil, errs.Combine(err, peer.Close())
 		}
 
+		peer.Transport = peer.Transport.WithObservers(peer.Kademlia.RoutingTable)
+
 		// TODO: reduce number of arguments
-		peer.Kademlia.Service, err = kademlia.NewService(peer.Log.Named("kademlia"), self, config.BootstrapNodes(), peer.Identity, config.Alpha, peer.Kademlia.RoutingTable)
+		peer.Kademlia.Service, err = kademlia.NewService(peer.Log.Named("kademlia"), self, config.BootstrapNodes(), peer.Transport, config.Alpha, peer.Kademlia.RoutingTable)
 		if err != nil {
 			return nil, errs.Combine(err, peer.Close())
 		}
@@ -191,7 +193,7 @@ func (peer *Peer) Run(ctx context.Context) error {
 		return ignoreCancel(peer.Kademlia.Service.Bootstrap(ctx))
 	})
 	group.Go(func() error {
-		return ignoreCancel(peer.Kademlia.Service.RunRefresh(ctx))
+		return ignoreCancel(peer.Kademlia.Service.Run(ctx))
 	})
 	group.Go(func() error {
 		return ignoreCancel(peer.Agreements.Sender.Run(ctx))