tbcd: clean up peer and peermanager (#273)

* Start cleaning up peer and peermanager

* Rewrite a new peer manager

* Make new peer manager mostly work

* Please linter

* Add some prometheus crap

* bring back reset clause

* Make peer manager parallel

* This seems to make providing seeds work correctly

* Missing defer in mutex, thanks Joshua

* Remove debug

* Joshua suggestions

* Catch up to go 1.23 to get nifty new tcp dial options and fix tests

* Fix test

* Bring back ping all peers

* Deal with signals

* fix seeding, don't run test that connects to testnet3

* fix make race; do not use same host and port for each test

* Make it the same as bssd

* Oops, forgot to commit this

* sundry joshua nits

* Fix error

---------

Co-authored-by: ClaytonNorthey92 <clayton.northey@gmail.com>

cmd/tbcd/tbcd.go

@@ -141,11 +141,6 @@ func HandleSignals(ctx context.Context, cancel context.CancelFunc, callback func
 	os.Exit(2)
 }
 
-func init() {
-	version.Component = "tbcd"
-	welcome = "Hemi Tiny Bitcoin Daemon " + version.BuildInfo()
-}
-
 func _main() error {
 	// Parse configuration from environment
 	if err := config.Parse(cm); err != nil {
@@ -205,6 +200,11 @@ func _main() error {
 	return nil
 }
 
+func init() {
+	version.Component = "tbcd"
+	welcome = "Hemi Tiny Bitcoin Daemon " + version.BuildInfo()
+}
+
 func main() {
 	if len(os.Args) != 1 {
 		fmt.Fprintf(os.Stderr, "%v\n", welcome)
@@ -216,7 +216,7 @@ func main() {
 	}
 
 	if err := _main(); err != nil {
-		log.Errorf("%v", err)
+		fmt.Fprintf(os.Stderr, "%v\n", err)
 		os.Exit(1)
 	}
 }

go.mod

@@ -1,6 +1,6 @@
 module github.com/hemilabs/heminetwork
 
-go 1.22
+go 1.23
 
 toolchain go1.23.0
 

service/tbc/crawler.go

@@ -1309,7 +1309,7 @@ func (s *Server) SyncIndexersToHash(ctx context.Context, hash *chainhash.Hash) e
 			return
 		}
 		// get a random peer
-		p, err := s.randomPeer(ctx)
+		p, err := s.pm.Random()
 		if err != nil {
 			s.mtx.Unlock()
 			log.Errorf("sync indexers random peer: %v", err)

service/tbc/peer.go

@@ -39,6 +39,7 @@ type peer struct {
 	connected time.Time
 
 	address string
+	id      int
 
 	protocolVersion uint32
 	network         wire.BitcoinNet
@@ -47,19 +48,27 @@ type peer struct {
 	addrV2        bool
 }
 
-func NewPeer(network wire.BitcoinNet, address string) (*peer, error) {
-	// XXX parse address and return failure if it's wrong
+func NewPeer(network wire.BitcoinNet, id int, address string) (*peer, error) {
+	_, _, err := net.SplitHostPort(address)
+	if err != nil {
+		return nil, fmt.Errorf("%v: %w", address, err)
+	}
 	return &peer{
 		protocolVersion: wire.ProtocolVersion,
 		network:         network,
 		address:         address,
+		id:              id,
 	}, nil
 }
 
 func (p *peer) String() string {
 	return p.address
 }
 
+func (p *peer) Id() int {
+	return p.id
+}
+
 func (p *peer) write(timeout time.Duration, msg wire.Message) error {
 	p.mtx.Lock()
 	conn := p.conn
@@ -182,16 +191,20 @@ func (p *peer) connect(ctx context.Context) error {
 	p.mtx.Unlock()
 
 	d := net.Dialer{
-		Timeout:   5 * time.Second,
-		KeepAlive: 9 * time.Second,
+		Deadline: time.Now().Add(5 * time.Second),
+		KeepAliveConfig: net.KeepAliveConfig{
+			Enable:   true,
+			Idle:     7 * time.Second,
+			Interval: 7 * time.Second,
+			Count:    2,
+		},
 	}
 
 	log.Debugf("dialing %s", p.address)
 	conn, err := d.DialContext(ctx, "tcp", p.address)
 	if err != nil {
 		return fmt.Errorf("dial %v: %w", p.address, err)
 	}
-	log.Debugf("done dialing %s", p.address)
 
 	err = p.handshake(ctx, conn)
 	if err != nil {
@@ -212,9 +225,12 @@ func (p *peer) close() error {
 	defer log.Tracef("close exit")
 
 	p.mtx.Lock()
-	defer p.mtx.Unlock()
-	if p.conn != nil {
-		return p.conn.Close()
+	conn := p.conn
+	p.conn = nil
+	p.isDialing = true // mark not connected
+	p.mtx.Unlock()
+	if conn != nil {
+		return conn.Close()
 	}
 	return net.ErrClosed
 }