Merge pull request #801 from starius/sweepbatcher-wait

sweepbatcher: add options WithInitialDelay and WithPublishDelay

Author: starius

sweepbatcher/store.go

@@ -11,7 +11,6 @@ import (
 	"github.com/btcsuite/btcd/wire"
 	"github.com/lightninglabs/loop/loopdb"
 	"github.com/lightninglabs/loop/loopdb/sqlc"
-	"github.com/lightningnetwork/lnd/clock"
 	"github.com/lightningnetwork/lnd/lntypes"
 )
 
@@ -72,15 +71,13 @@ type SQLStore struct {
 	baseDb BaseDB
 
 	network *chaincfg.Params
-	clock   clock.Clock
 }
 
 // NewSQLStore creates a new SQLStore.
 func NewSQLStore(db BaseDB, network *chaincfg.Params) *SQLStore {
 	return &SQLStore{
 		baseDb:  db,
 		network: network,
-		clock:   clock.NewDefaultClock(),
 	}
 }
 

sweepbatcher/sweep_batch.go

@@ -24,6 +24,7 @@ import (
 	"github.com/lightninglabs/loop/swap"
 	sweeppkg "github.com/lightninglabs/loop/sweep"
 	"github.com/lightningnetwork/lnd/chainntnfs"
+	"github.com/lightningnetwork/lnd/clock"
 	"github.com/lightningnetwork/lnd/input"
 	"github.com/lightningnetwork/lnd/keychain"
 	"github.com/lightningnetwork/lnd/lnrpc/walletrpc"
@@ -134,8 +135,17 @@ type batchConfig struct {
 	// batchConfTarget is the confirmation target of the batch transaction.
 	batchConfTarget int32
 
-	// batchPublishDelay is the delay between receiving a new block and
-	// publishing the batch transaction.
+	// clock provides methods to work with time and timers.
+	clock clock.Clock
+
+	// initialDelay is the delay of first batch publishing after creation.
+	// It only affects newly created batches, not batches loaded from DB,
+	// so publishing does happen in case of a daemon restart (especially
+	// important in case of a crashloop).
+	initialDelay time.Duration
+
+	// batchPublishDelay is the delay between receiving a new block or
+	// initial delay completion and publishing the batch transaction.
 	batchPublishDelay time.Duration
 
 	// noBumping instructs sweepbatcher not to fee bump itself and rely on
@@ -507,6 +517,11 @@ func (b *batch) Wait() {
 	<-b.finished
 }
 
+// stillWaitingMsg is the format of the message printed if the batch is about
+// to publish, but initial delay has not ended yet.
+const stillWaitingMsg = "Skipping publishing, initial delay will end at " +
+	"%v, now is %v."
+
 // Run is the batch's main event loop.
 func (b *batch) Run(ctx context.Context) error {
 	runCtx, cancel := context.WithCancel(ctx)
@@ -527,6 +542,9 @@ func (b *batch) Run(ctx context.Context) error {
 		return fmt.Errorf("both musig2 signers provided")
 	}
 
+	// Cache clock variable.
+	clock := b.cfg.clock
+
 	blockChan, blockErrChan, err :=
 		b.chainNotifier.RegisterBlockEpochNtfn(runCtx)
 	if err != nil {
@@ -543,10 +561,25 @@ func (b *batch) Run(ctx context.Context) error {
 		}
 	}
 
+	// skipBefore is the time before which we skip batch publishing.
+	// This is needed to facilitate better grouping of sweeps.
+	// For batches loaded from DB initialDelay should be 0.
+	skipBefore := clock.Now().Add(b.cfg.initialDelay)
+
+	// initialDelayChan is a timer which fires upon initial delay end.
+	// If initialDelay is 0, it does not fire to prevent race with
+	// blockChan which also fires immediately with current tip. Such a race
+	// may result in double publishing if batchPublishDelay is also 0.
+	var initialDelayChan <-chan time.Time
+	if b.cfg.initialDelay > 0 {
+		initialDelayChan = clock.TickAfter(b.cfg.initialDelay)
+	}
+
 	// We use a timer in order to not publish new transactions at the same
 	// time as the block epoch notification. This is done to prevent
 	// unnecessary transaction publishments when a spend is detected on that
-	// block.
+	// block. This timer starts after new block arrives or initialDelay
+	// completes.
 	var timerChan <-chan time.Time
 
 	b.log.Infof("started, primary %x, total sweeps %v",
@@ -557,20 +590,48 @@ func (b *batch) Run(ctx context.Context) error {
 		case <-b.callEnter:
 			<-b.callLeave
 
+		// blockChan provides immediately the current tip.
 		case height := <-blockChan:
 			b.log.Debugf("received block %v", height)
 
 			// Set the timer to publish the batch transaction after
 			// the configured delay.
-			timerChan = time.After(b.cfg.batchPublishDelay)
+			timerChan = clock.TickAfter(b.cfg.batchPublishDelay)
 			b.currentHeight = height
 
+		case <-initialDelayChan:
+			b.log.Debugf("initial delay of duration %v has ended",
+				b.cfg.initialDelay)
+
+			// Set the timer to publish the batch transaction after
+			// the configured delay.
+			timerChan = clock.TickAfter(b.cfg.batchPublishDelay)
+
 		case <-timerChan:
-			if b.state == Open {
-				err := b.publish(ctx)
-				if err != nil {
-					return err
-				}
+			// Check that batch is still open.
+			if b.state != Open {
+				b.log.Debugf("Skipping publishing, because the"+
+					"batch is not open (%v).", b.state)
+				continue
+			}
+
+			// If the batch became urgent, skipBefore is set to now.
+			if b.isUrgent(skipBefore) {
+				skipBefore = clock.Now()
+			}
+
+			// Check that the initial delay has ended. We have also
+			// batchPublishDelay on top of initialDelay, so if
+			// initialDelayChan has just fired, this check passes.
+			now := clock.Now()
+			if skipBefore.After(now) {
+				b.log.Debugf(stillWaitingMsg, skipBefore, now)
+				continue
+			}
+
+			err := b.publish(ctx)
+			if err != nil {
+				return err
 			}
 
 		case spend := <-b.spendChan:
@@ -604,6 +665,57 @@ func (b *batch) Run(ctx context.Context) error {
 	}
 }
 
+// timeout returns minimum timeout as block height among sweeps of the batch.
+// If the batch is empty, return -1.
+func (b *batch) timeout() int32 {
+	// Find minimum among sweeps' timeouts.
+	minTimeout := int32(-1)
+	for _, sweep := range b.sweeps {
+		if minTimeout == -1 || minTimeout > sweep.timeout {
+			minTimeout = sweep.timeout
+		}
+	}
+
+	return minTimeout
+}
+
+// isUrgent checks if the batch became urgent. This is determined by comparing
+// the remaining number of blocks until timeout to the initial delay remained,
+// given one block is 10 minutes.
+func (b *batch) isUrgent(skipBefore time.Time) bool {
+	timeout := b.timeout()
+	if timeout <= 0 {
+		b.log.Warnf("Method timeout() returned %v. Number of"+
+			" sweeps: %d. It may be an empty batch.",
+			timeout, len(b.sweeps))
+		return false
+	}
+
+	if b.currentHeight == 0 {
+		// currentHeight is not initiated yet.
+		return false
+	}
+
+	blocksToTimeout := timeout - b.currentHeight
+	const blockTime = 10 * time.Minute
+	timeBank := time.Duration(blocksToTimeout) * blockTime
+
+	// We want to have at least 2x as much time to be safe.
+	const safetyFactor = 2
+	remainingWaiting := skipBefore.Sub(b.cfg.clock.Now())
+
+	if timeBank >= safetyFactor*remainingWaiting {
+		// There is enough time, keep waiting.
+		return false
+	}
+
+	b.log.Debugf("cancelling waiting for urgent sweep (timeBank is %v, "+
+		"remainingWaiting is %v)", timeBank, remainingWaiting)
+
+	// Signal to the caller to cancel initialDelay.
+	return true
+}
+
 // publish creates and publishes the latest batch transaction to the network.
 func (b *batch) publish(ctx context.Context) error {
 	var (

sweepbatcher/sweep_batcher.go

@@ -16,6 +16,7 @@ import (
 	"github.com/lightninglabs/loop/loopdb"
 	"github.com/lightninglabs/loop/swap"
 	"github.com/lightninglabs/loop/utils"
+	"github.com/lightningnetwork/lnd/clock"
 	"github.com/lightningnetwork/lnd/input"
 	"github.com/lightningnetwork/lnd/lntypes"
 	"github.com/lightningnetwork/lnd/lnwallet/chainfee"
@@ -44,7 +45,7 @@ const (
 
 	// defaultTestnetPublishDelay is the default publish delay that is used
 	// for testnet.
-	defaultPublishDelay = 500 * time.Millisecond
+	defaultTestnetPublishDelay = 500 * time.Millisecond
 )
 
 type BatcherStore interface {
@@ -253,6 +254,23 @@ type Batcher struct {
 	// exit.
 	wg sync.WaitGroup
 
+	// clock provides methods to work with time and timers.
+	clock clock.Clock
+
+	// initialDelay is the delay of first batch publishing after creation.
+	// It only affects newly created batches, not batches loaded from DB,
+	// so publishing does happen in case of a daemon restart (especially
+	// important in case of a crashloop). If a sweep is about to expire
+	// (time until timeout is less that 2x initialDelay), then waiting is
+	// skipped.
+	initialDelay time.Duration
+
+	// publishDelay is the delay of batch publishing that is applied in the
+	// beginning, after the appearance of a new block in the network or
+	// after the end of initial delay. For batches recovered from DB this
+	// value is always 0s, regardless of this setting.
+	publishDelay time.Duration
+
 	// customFeeRate provides custom min fee rate per swap. The batch uses
 	// max of the fee rates of its swaps. In this mode confTarget is
 	// ignored and fee bumping by sweepbatcher is disabled.
@@ -267,6 +285,23 @@ type Batcher struct {
 
 // BatcherConfig holds batcher configuration.
 type BatcherConfig struct {
+	// clock provides methods to work with time and timers.
+	clock clock.Clock
+
+	// initialDelay is the delay of first batch publishing after creation.
+	// It only affects newly created batches, not batches loaded from DB,
+	// so publishing does happen in case of a daemon restart (especially
+	// important in case of a crashloop). If a sweep is about to expire
+	// (time until timeout is less that 2x initialDelay), then waiting is
+	// skipped.
+	initialDelay time.Duration
+
+	// publishDelay is the delay of batch publishing that is applied in the
+	// beginning, after the appearance of a new block in the network or
+	// after the end of initial delay. For batches recovered from DB this
+	// value is always 0s, regardless of this setting.
+	publishDelay time.Duration
+
 	// customFeeRate provides custom min fee rate per swap. The batch uses
 	// max of the fee rates of its swaps. In this mode confTarget is
 	// ignored and fee bumping by sweepbatcher is disabled.
@@ -282,6 +317,37 @@ type BatcherConfig struct {
 // BatcherOption configures batcher behaviour.
 type BatcherOption func(*BatcherConfig)
 
+// WithClock sets the clock used by sweepbatcher and its batches. It is needed
+// to manipulate time in tests.
+func WithClock(clock clock.Clock) BatcherOption {
+	return func(cfg *BatcherConfig) {
+		cfg.clock = clock
+	}
+}
+
+// WithInitialDelay instructs sweepbatcher to wait for the duration provided
+// after new batch creation before it is first published. This facilitates
+// better grouping. Defaults to 0s (no initial delay). If a sweep is about
+// to expire (time until timeout is less that 2x initialDelay), then waiting
+// is skipped.
+func WithInitialDelay(initialDelay time.Duration) BatcherOption {
+	return func(cfg *BatcherConfig) {
+		cfg.initialDelay = initialDelay
+	}
+}
+
+// WithPublishDelay sets the delay of batch publishing that is applied in the
+// beginning, after the appearance of a new block in the network or after the
+// end of initial delay (see WithInitialDelay). It is needed to prevent
+// unnecessary transaction publishments when a spend is detected on that block.
+// Default value depends on the network: 5 seconds in mainnet, 0.5s in testnet.
+// For batches recovered from DB this value is always 0s.
+func WithPublishDelay(publishDelay time.Duration) BatcherOption {
+	return func(cfg *BatcherConfig) {
+		cfg.publishDelay = publishDelay
+	}
+}
+
 // WithCustomFeeRate instructs sweepbatcher not to fee bump itself and rely on
 // external source of fee rates (FeeRateProvider). To apply a fee rate change,
 // the caller should re-add the sweep by calling AddSweep.
@@ -315,6 +381,11 @@ func NewBatcher(wallet lndclient.WalletKitClient,
 		opt(&cfg)
 	}
 
+	// If WithClock was not provided, use default clock.
+	if cfg.clock == nil {
+		cfg.clock = clock.NewDefaultClock()
+	}
+
 	if cfg.customMuSig2Signer != nil && musig2ServerSigner != nil {
 		panic("customMuSig2Signer must not be used with " +
 			"musig2ServerSigner")
@@ -334,6 +405,9 @@ func NewBatcher(wallet lndclient.WalletKitClient,
 		chainParams:        chainparams,
 		store:              store,
 		sweepStore:         sweepStore,
+		clock:              cfg.clock,
+		initialDelay:       cfg.initialDelay,
+		publishDelay:       cfg.publishDelay,
 		customFeeRate:      cfg.customFeeRate,
 		customMuSig2Signer: cfg.customMuSig2Signer,
 	}
@@ -536,8 +610,22 @@ func (b *Batcher) spinUpBatch(ctx context.Context) (*batch, error) {
 		cfg.batchPublishDelay = defaultMainnetPublishDelay
 
 	default:
-		cfg.batchPublishDelay = defaultPublishDelay
+		cfg.batchPublishDelay = defaultTestnetPublishDelay
+	}
+
+	if b.publishDelay != 0 {
+		if b.publishDelay < 0 {
+			return nil, fmt.Errorf("negative publishDelay: %v",
+				b.publishDelay)
+		}
+		cfg.batchPublishDelay = b.publishDelay
+	}
+
+	if b.initialDelay < 0 {
+		return nil, fmt.Errorf("negative initialDelay: %v",
+			b.initialDelay)
 	}
+	cfg.initialDelay = b.initialDelay
 
 	batchKit := b.newBatchKit()
 
@@ -626,6 +714,9 @@ func (b *Batcher) spinUpBatchFromDB(ctx context.Context, batch *batch) error {
 
 	cfg := b.newBatchConfig(batch.cfg.maxTimeoutDistance)
 
+	// Note that initialDelay and batchPublishDelay are 0 for batches
+	// recovered from DB so publishing happen in case of a daemon restart
+	// (especially important in case of a crashloop).
 	newBatch, err := NewBatchFromDB(cfg, batchKit)
 	if err != nil {
 		return fmt.Errorf("failed in NewBatchFromDB: %w", err)
@@ -934,6 +1025,7 @@ func (b *Batcher) newBatchConfig(maxTimeoutDistance int32) batchConfig {
 		maxTimeoutDistance: maxTimeoutDistance,
 		noBumping:          b.customFeeRate != nil,
 		customMuSig2Signer: b.customMuSig2Signer,
+		clock:              b.clock,
 	}
 }