Merge pull request #1675 from cfromknecht/peer-backoff-backoff

server: peer backoff backoff

peer.go

@@ -111,13 +111,12 @@ type peer struct {
 	addr        *lnwire.NetAddress
 	pubKeyBytes [33]byte
 
-	inbound bool
+	// startTime is the time this peer connection was successfully
+	// established. It will be zero for peers that did not successfuly
+	// Start().
+	startTime time.Time
 
-	// This mutex protects all the stats below it.
-	sync.RWMutex
-	timeConnected time.Time
-	lastSend      time.Time
-	lastRecv      time.Time
+	inbound bool
 
 	// sendQueue is the channel which is used to queue outgoing to be
 	// written onto the wire. Note that this channel is unbuffered.
@@ -303,6 +302,8 @@ func (p *peer) Start() error {
 		return fmt.Errorf("unable to load channels: %v", err)
 	}
 
+	p.startTime = time.Now()
+
 	p.wg.Add(5)
 	go p.queueHandler()
 	go p.writeHandler()
@@ -1977,6 +1978,8 @@ func (p *peer) sendInitMsg() error {
 // SendMessage sends a variadic number of message to remote peer. The first
 // argument denotes if the method should block until the message has been sent
 // to the remote peer.
+//
+// NOTE: Part of the lnpeer.Peer interface.
 func (p *peer) SendMessage(sync bool, msgs ...lnwire.Message) error {
 	// Add all incoming messages to the outgoing queue. A list of error
 	// chans is populated for each message if the caller requested a sync
@@ -2009,22 +2012,30 @@ func (p *peer) SendMessage(sync bool, msgs ...lnwire.Message) error {
 }
 
 // PubKey returns the pubkey of the peer in compressed serialized format.
+//
+// NOTE: Part of the lnpeer.Peer interface.
 func (p *peer) PubKey() [33]byte {
 	return p.pubKeyBytes
 }
 
 // IdentityKey returns the public key of the remote peer.
+//
+// NOTE: Part of the lnpeer.Peer interface.
 func (p *peer) IdentityKey() *btcec.PublicKey {
 	return p.addr.IdentityKey
 }
 
 // Address returns the network address of the remote peer.
+//
+// NOTE: Part of the lnpeer.Peer interface.
 func (p *peer) Address() net.Addr {
 	return p.addr.Address
 }
 
 // AddNewChannel adds a new channel to the peer. The channel should fail to be
 // added if the cancel channel is closed.
+//
+// NOTE: Part of the lnpeer.Peer interface.
 func (p *peer) AddNewChannel(channel *lnwallet.LightningChannel,
 	cancel <-chan struct{}) error {
 
@@ -2053,6 +2064,12 @@ func (p *peer) AddNewChannel(channel *lnwallet.LightningChannel,
 	return nil
 }
 
+// StartTime returns the time at which the connection was established if the
+// peer started successfully, and zero otherwise.
+func (p *peer) StartTime() time.Time {
+	return p.startTime
+}
+
 // TODO(roasbeef): make all start/stop mutexes a CAS
 
 // fetchLastChanUpdate returns a function which is able to retrieve the last

server.go

@@ -2018,7 +2018,7 @@ func (s *server) peerTerminationWatcher(p *peer) {
 			s.persistentConnReqs[pubStr], connReq)
 
 		// Record the computed backoff in the backoff map.
-		backoff := s.nextPeerBackoff(pubStr)
+		backoff := s.nextPeerBackoff(pubStr, p.StartTime())
 		s.persistentPeersBackoff[pubStr] = backoff
 
 		// Initialize a retry canceller for this peer if one does not
@@ -2055,7 +2055,9 @@ func (s *server) peerTerminationWatcher(p *peer) {
 // nextPeerBackoff computes the next backoff duration for a peer's pubkey using
 // exponential backoff. If no previous backoff was known, the default is
 // returned.
-func (s *server) nextPeerBackoff(pubStr string) time.Duration {
+func (s *server) nextPeerBackoff(pubStr string,
+	startTime time.Time) time.Duration {
+
 	// Now, determine the appropriate backoff to use for the retry.
 	backoff, ok := s.persistentPeersBackoff[pubStr]
 	if !ok {
@@ -2063,9 +2065,27 @@ func (s *server) nextPeerBackoff(pubStr string) time.Duration {
 		return defaultBackoff
 	}
 
-	// Otherwise, use a previous backoff to compute the
-	// subsequent randomized exponential backoff duration.
-	return computeNextBackoff(backoff)
+	// If the peer failed to start properly, we'll just use the previous
+	// backoff to compute the subsequent randomized exponential backoff
+	// duration. This will roughly double on average.
+	if startTime.IsZero() {
+		return computeNextBackoff(backoff)
+	}
+
+	// The peer succeeded in starting. We'll reduce the timeout duration
+	// by the length of the connection before applying randomized
+	// exponential backoff. We'll only apply this if:
+	//   backoff - connDuration > defaultBackoff
+	connDuration := time.Now().Sub(startTime)
+	relaxedBackoff := backoff - connDuration
+	if relaxedBackoff > defaultBackoff {
+		return computeNextBackoff(relaxedBackoff)
+	}
+
+	// Otherwise, backoff - connDuration <= defaultBackoff, meaning the
+	// connection lasted much longer than our previous backoff. To reward
+	// such good behavior, we'll reconnect after the default timeout.
+	return defaultBackoff
 }
 
 // shouldRequestGraphSync returns true if the servers deems it necessary that