Write each handler batch from its flush goroutine, outside the mutex (closes #65)

The AS, peer and prefix handlers wrote their batch to the database while
holding the mutex HandleMessage needs, so while a write ran nothing was
taken off their queues and the streamer dropped updates.

Each handler now writes only from its flush goroutine: it takes the batch
under the mutex, leaves an empty one in its place, and writes after
releasing it. A full batch wakes the flush goroutine. While a write runs the
next batch may grow to ten batch sizes; at that size HandleMessage waits, so
the queue fills and the streamer drops as before. README "Memory" gives the
numbers and their cost.

Model: opus-5-5

internal/routewatch/peerhandler.go

@@ -23,6 +23,12 @@ const (
 	// peerBatchSize is the number of peer updates to batch together
 	peerBatchSize = 10000
 
+	// peerBatchMaxSize is the most peer updates the batch holds while the
+	// flush goroutine writes the previous one: ten batches, about 10 MiB
+	// (README "Memory"). At this size HandleMessage waits until the flush
+	// goroutine takes the batch.
+	peerBatchMaxSize = 100000
+
 	// peerBatchTimeout is the maximum time to wait before flushing a batch
 	peerBatchTimeout = 2 * time.Second
 )
@@ -35,12 +41,16 @@ type PeerHandler struct {
 	db     database.Store
 	logger *logger.Logger
 
-	// Batching
-	mu        sync.Mutex
-	peerBatch []peerUpdate
-	lastFlush time.Time
-	stopCh    chan struct{}
-	wg        sync.WaitGroup
+	// Batching. mu guards peerBatch only: the flush goroutine takes the batch
+	// under mu and writes it after releasing mu, so HandleMessage waits for
+	// a database write only while peerBatch is at peerBatchMaxSize.
+	mu          sync.Mutex
+	peerBatch   []peerUpdate
+	batchTaken  sync.Cond     // signalled when the flush goroutine takes peerBatch; its L is &mu
+	batchFullCh chan struct{} // wakes the flush goroutine when peerBatch is full
+	lastFlush   time.Time     // used only by the flush goroutine
+	stopCh      chan struct{}
+	wg          sync.WaitGroup
 }
 
 type peerUpdate struct {
@@ -56,12 +66,14 @@ type peerUpdate struct {
 // should be stopped by calling Stop when it is no longer needed.
 func NewPeerHandler(db database.Store, logger *logger.Logger) *PeerHandler {
 	h := &PeerHandler{
-		db:        db,
-		logger:    logger,
-		peerBatch: make([]peerUpdate, 0, peerBatchSize),
-		lastFlush: time.Now(),
-		stopCh:    make(chan struct{}),
+		db:          db,
+		logger:      logger,
+		peerBatch:   make([]peerUpdate, 0, peerBatchSize),
+		batchFullCh: make(chan struct{}, 1),
+		lastFlush:   time.Now(),
+		stopCh:      make(chan struct{}),
 	}
+	h.batchTaken.L = &h.mu
 
 	// Start the flush timer goroutine
 	h.wg.Add(1)
@@ -87,8 +99,8 @@ func (h *PeerHandler) QueueCapacity() int {
 
 // HandleMessage processes a RIS message to track peer information.
 // It extracts the peer IP address and ASN from the message and adds
-// the update to an internal batch. When the batch reaches peerBatchSize
-// or the batch timeout expires, the batch is flushed to the database.
+// the update to an internal batch. The flush goroutine writes the batch to the
+// database when it reaches peerBatchSize or the batch timeout expires.
 func (h *PeerHandler) HandleMessage(msg *ristypes.RISMessage) {
 	// Parse peer ASN from string
 	peerASN := 0
@@ -101,6 +113,13 @@ func (h *PeerHandler) HandleMessage(msg *ristypes.RISMessage) {
 	h.mu.Lock()
 	defer h.mu.Unlock()
 
+	// While the batch is at its maximum size, wait until the flush goroutine
+	// takes it. Meanwhile the handler takes nothing off its queue, which
+	// fills, and the streamer drops messages for it.
+	for len(h.peerBatch) >= peerBatchMaxSize {
+		h.batchTaken.Wait()
+	}
+
 	// Add to batch
 	h.peerBatch = append(h.peerBatch, peerUpdate{
 		peerIP:      msg.Peer,
@@ -109,13 +128,19 @@ func (h *PeerHandler) HandleMessage(msg *ristypes.RISMessage) {
 		timestamp:   msg.ParsedTimestamp,
 	})
 
-	// Check if we need to flush
+	// A full batch wakes the flush goroutine. If it is writing, the wake-up
+	// waits in batchFullCh and the batch keeps growing until the write ends
+	// or the batch reaches peerBatchMaxSize.
 	if len(h.peerBatch) >= peerBatchSize {
-		h.flushBatchLocked()
+		select {
+		case h.batchFullCh <- struct{}{}:
+		default: // already woken
+		}
 	}
 }
 
-// flushLoop runs in a goroutine and periodically flushes batches
+// flushLoop runs in a goroutine, the only one that writes the batch: it
+// flushes when the batch is full, periodically, and once more on Stop.
 func (h *PeerHandler) flushLoop() {
 	defer h.wg.Done()
 	ticker := time.NewTicker(peerBatchTimeout)
@@ -124,31 +149,47 @@ func (h *PeerHandler) flushLoop() {
 	for {
 		select {
 		case <-ticker.C:
-			h.mu.Lock()
 			if time.Since(h.lastFlush) >= peerBatchTimeout {
-				h.flushBatchLocked()
+				h.flushBatch()
 			}
-			h.mu.Unlock()
+		case <-h.batchFullCh:
+			h.flushBatch()
 		case <-h.stopCh:
 			// Final flush
-			h.mu.Lock()
-			h.flushBatchLocked()
-			h.mu.Unlock()
+			h.flushBatch()
 
 			return
 		}
 	}
 }
 
-// flushBatchLocked flushes the peer batch to the database (must be called with mutex held)
-func (h *PeerHandler) flushBatchLocked() {
+// takeBatch returns the pending updates and leaves an empty batch in their
+// place for HandleMessage to fill.
+func (h *PeerHandler) takeBatch() []peerUpdate {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+
 	if len(h.peerBatch) == 0 {
+		return nil
+	}
+	batch := h.peerBatch
+	h.peerBatch = make([]peerUpdate, 0, peerBatchSize)
+	h.batchTaken.Broadcast()
+
+	return batch
+}
+
+// flushBatch takes the pending batch and writes it to the database, without
+// holding the mutex during the write.
+func (h *PeerHandler) flushBatch() {
+	batch := h.takeBatch()
+	if len(batch) == 0 {
 		return
 	}
 
 	// Deduplicate by peer IP, keeping the latest update for each peer
 	peerMap := make(map[string]peerUpdate)
-	for _, update := range h.peerBatch {
+	for _, update := range batch {
 		if existing, ok := peerMap[update.peerIP]; !ok || update.timestamp.After(existing.timestamp) {
 			peerMap[update.peerIP] = update
 		}
@@ -170,8 +211,6 @@ func (h *PeerHandler) flushBatchLocked() {
 		h.logger.Error("Failed to process peer batch", "error", err, "count", len(dbPeerMap))
 	}
 
-	// Clear batch
-	h.peerBatch = h.peerBatch[:0]
 	h.lastFlush = time.Now()
 }