Store a new row's earliest timestamp as first_seen (closes #112)

Each handler took the latest time per key in a batch and inserted it as
a new row's first_seen, so a key updated twice in one batch, seconds
apart, was stored first seen at the later time, and the IP lookup could
answer with a prefix first_seen after the route's last_seen.

The prefix handler now keeps the earliest time of all of the batch's
updates for each prefix. The AS, peer and peering handlers keep the
earliest and the latest time per key in database.SeenTimes, and
GetOrCreateASNBatch, UpdatePeerBatch and RecordPeering store a new row
first seen at the earliest and last seen at the latest. A stored row
still has only its last_seen set, to the latest.

Model: opus-5-5

internal/routewatch/app_integration_internal_test.go

@@ -152,7 +152,7 @@ func (m *mockStore) RecordAnnouncement(_ *database.Announcement) error {
 }
 
 // RecordPeering mock implementation
-func (m *mockStore) RecordPeering(asA, asB int, _ time.Time) error {
+func (m *mockStore) RecordPeering(asA, asB int, _ database.SeenTimes) error {
 	m.mu.Lock()
 	defer m.mu.Unlock()
 
@@ -396,16 +396,16 @@ func (m *mockStore) DeleteLiveRoutesOfPeer(_ string) (int, error) {
 }
 
 // GetOrCreateASNBatch mock implementation
-func (m *mockStore) GetOrCreateASNBatch(asns map[int]time.Time) error {
+func (m *mockStore) GetOrCreateASNBatch(asns map[int]database.SeenTimes) error {
 	m.mu.Lock()
 	defer m.mu.Unlock()
 
-	for number, timestamp := range asns {
+	for number, seen := range asns {
 		if _, exists := m.ASNs[number]; !exists {
 			m.ASNs[number] = &database.ASN{
 				ASN:       number,
-				FirstSeen: timestamp,
-				LastSeen:  timestamp,
+				FirstSeen: seen.First,
+				LastSeen:  seen.Last,
 			}
 			m.ASNCount++
 		}
@@ -766,7 +766,9 @@ const (
 	callClose      = "close"
 )
 
-func (s *closeOrderStore) GetOrCreateASNBatch(asns map[int]time.Time) error {
+func (s *closeOrderStore) GetOrCreateASNBatch(
+	asns map[int]database.SeenTimes,
+) error {
 	s.record(callWrite)
 
 	return s.Store.GetOrCreateASNBatch(asns)

internal/routewatch/ashandler.go

@@ -229,19 +229,21 @@ func (h *ASHandler) flushBatch() {
 		return
 	}
 
-	// Process ASNs first (deduped)
-	asnMap := make(map[int]time.Time)
+	// The earliest and the latest time the batch saw each AS. The batch holds
+	// several updates for an AS, from many peers, so a new AS is stored first
+	// seen at the earliest of them and last seen at the latest.
+	asns := make(map[int]database.SeenTimes)
 	for _, op := range batch {
-		if existing, ok := asnMap[op.number]; !ok || op.timestamp.After(existing) {
-			asnMap[op.number] = op.timestamp
-		}
+		seen := asns[op.number]
+		seen.Add(op.timestamp)
+		asns[op.number] = seen
 	}
 
 	// Process all ASNs in a single batch transaction
-	err := h.db.GetOrCreateASNBatch(asnMap)
+	err := h.db.GetOrCreateASNBatch(asns)
 	if err != nil {
 		h.logger.Error("Failed to process ASN batch",
-			"error", err, "count", len(asnMap))
+			"error", err, "count", len(asns))
 	}
 
 	h.mu.Lock()

internal/routewatch/ashandler_internal_test.go

@@ -0,0 +1,44 @@
+package routewatch
+
+import (
+	"testing"
+
+	"git.eeqj.de/sneak/routewatch/internal/logger"
+)
+
+// TestASHandlerStoresTheEarliestAndLatestTimesOfANewAS handles two updates
+// whose path carries an AS the database does not hold yet, the second 5
+// seconds after the first, in one batch into a real database. The AS is stored
+// first seen at the time of the first update and last seen at the time of the
+// second.
+func TestASHandlerStoresTheEarliestAndLatestTimesOfANewAS(t *testing.T) {
+	t.Parallel()
+
+	db := newTestDatabase(t)
+
+	// Built without NewASHandler's flush goroutine, so the test flushes itself.
+	h := &ASHandler{db: db, logger: logger.New()}
+
+	first := peerAnnouncement(testPeerIP, testV4Prefix)
+	second := peerAnnouncement(testPeerIP, testV4Prefix)
+	second.ParsedTimestamp = first.ParsedTimestamp.Add(secondUpdateAfter)
+
+	h.HandleMessage(first)
+	h.HandleMessage(second)
+	h.flushBatch()
+
+	asn, _, err := db.GetASDetails(testASNA)
+	if err != nil {
+		t.Fatalf("GetASDetails(%d): %v", testASNA, err)
+	}
+
+	if !asn.FirstSeen.Equal(first.ParsedTimestamp) {
+		t.Errorf("first_seen of AS%d = %s, want %s", testASNA,
+			asn.FirstSeen, first.ParsedTimestamp)
+	}
+
+	if !asn.LastSeen.Equal(second.ParsedTimestamp) {
+		t.Errorf("last_seen of AS%d = %s, want %s", testASNA,
+			asn.LastSeen, second.ParsedTimestamp)
+	}
+}

internal/routewatch/batchwrite_internal_test.go

@@ -66,7 +66,7 @@ func (s *blockingStore) UpsertLiveRouteBatch(routes []*database.LiveRoute) error
 	return nil
 }
 
-func (s *blockingStore) GetOrCreateASNBatch(asns map[int]time.Time) error {
+func (s *blockingStore) GetOrCreateASNBatch(asns map[int]database.SeenTimes) error {
 	s.waitForRelease()
 	s.mu.Lock()
 	defer s.mu.Unlock()
@@ -96,10 +96,12 @@ func (s *blockingStore) DeleteLiveRoutesOfPeer(peerIP string) (int, error) {
 	return s.mockStore.DeleteLiveRoutesOfPeer(peerIP)
 }
 
-func (s *blockingStore) RecordPeering(asA, asB int, ts time.Time) error {
+func (s *blockingStore) RecordPeering(
+	asA, asB int, seen database.SeenTimes,
+) error {
 	s.waitForRelease()
 
-	return s.mockStore.RecordPeering(asA, asB, ts)
+	return s.mockStore.RecordPeering(asA, asB, seen)
 }
 
 func (s *blockingStore) waitForRelease() {

internal/routewatch/peerhandler.go

@@ -37,8 +37,8 @@ const (
 
 // PeerHandler tracks BGP peers from UPDATE messages using batched operations.
 // It maintains a queue of peer updates and periodically flushes them to the database
-// in batches to improve performance. The handler deduplicates peer updates within
-// each batch, keeping only the most recent update for each peer IP address.
+// in batches to improve performance. The handler writes one row per peer IP
+// address for each batch, with the earliest and the latest time of its updates.
 type PeerHandler struct {
 	db     database.Store
 	logger *logger.Logger
@@ -227,31 +227,28 @@ func (h *PeerHandler) flushBatch() {
 		return
 	}
 
-	// Deduplicate by peer IP, keeping the latest update for each peer
-	peerMap := make(map[string]peerUpdate)
+	// The earliest and the latest time the batch saw each peer, with the peer
+	// AS and message type of its latest update. The batch holds several
+	// updates for a peer, so a new peer is stored first seen at the earliest of
+	// them and last seen at the latest.
+	peers := make(map[string]database.PeerUpdate)
 	for _, update := range batch {
-		existing, ok := peerMap[update.peerIP]
-		if !ok || update.timestamp.After(existing.timestamp) {
-			peerMap[update.peerIP] = update
+		peer, ok := peers[update.peerIP]
+		if !ok || update.timestamp.After(peer.Seen.Last) {
+			peer.PeerIP = update.peerIP
+			peer.PeerASN = update.peerASN
+			peer.MessageType = update.messageType
 		}
-	}
 
-	// Convert to database format
-	dbPeerMap := make(map[string]database.PeerUpdate)
-	for peerIP, update := range peerMap {
-		dbPeerMap[peerIP] = database.PeerUpdate{
-			PeerIP:      update.peerIP,
-			PeerASN:     update.peerASN,
-			MessageType: update.messageType,
-			Timestamp:   update.timestamp,
-		}
+		peer.Seen.Add(update.timestamp)
+		peers[update.peerIP] = peer
 	}
 
 	// Process all peers in a single batch transaction
-	err := h.db.UpdatePeerBatch(dbPeerMap)
+	err := h.db.UpdatePeerBatch(peers)
 	if err != nil {
 		h.logger.Error("Failed to process peer batch",
-			"error", err, "count", len(dbPeerMap))
+			"error", err, "count", len(peers))
 	}
 
 	h.mu.Lock()

internal/routewatch/peerhandler_internal_test.go

@@ -0,0 +1,59 @@
+package routewatch
+
+import (
+	"testing"
+
+	"git.eeqj.de/sneak/routewatch/internal/database"
+	"git.eeqj.de/sneak/routewatch/internal/logger"
+)
+
+// peerBatchStore keeps the peers of the last UpdatePeerBatch call. The
+// database has no read of the peer table, so the peer handler's test checks
+// what the handler writes.
+type peerBatchStore struct {
+	*mockStore
+
+	peers map[string]database.PeerUpdate
+}
+
+func (s *peerBatchStore) UpdatePeerBatch(
+	peers map[string]database.PeerUpdate,
+) error {
+	s.peers = peers
+
+	return nil
+}
+
+// TestPeerHandlerWritesTheEarliestAndLatestTimesOfANewPeer handles two updates
+// from one peer, the second 5 seconds after the first, in one batch. The peer
+// handler writes the peer first seen at the time of the first update and last
+// seen at the time of the second.
+func TestPeerHandlerWritesTheEarliestAndLatestTimesOfANewPeer(t *testing.T) {
+	t.Parallel()
+
+	store := &peerBatchStore{mockStore: newMockStore()}
+
+	// Built without NewPeerHandler's flush goroutine, so the test flushes
+	// itself.
+	h := &PeerHandler{db: store, logger: logger.New()}
+
+	first := peerAnnouncement(testPeerIP, testV4Prefix)
+	second := peerAnnouncement(testPeerIP, testV4Prefix)
+	second.ParsedTimestamp = first.ParsedTimestamp.Add(secondUpdateAfter)
+
+	h.HandleMessage(first)
+	h.HandleMessage(second)
+	h.flushBatch()
+
+	seen := store.peers[testPeerIP].Seen
+
+	if !seen.First.Equal(first.ParsedTimestamp) {
+		t.Errorf("first_seen of peer %s = %s, want %s", testPeerIP,
+			seen.First, first.ParsedTimestamp)
+	}
+
+	if !seen.Last.Equal(second.ParsedTimestamp) {
+		t.Errorf("last_seen of peer %s = %s, want %s", testPeerIP,
+			seen.Last, second.ParsedTimestamp)
+	}
+}

internal/routewatch/peeringhandler.go

@@ -42,10 +42,11 @@ type PeeringHandler struct {
 	db     database.Store
 	logger *logger.Logger
 
-	// In-memory AS path tracking
+	// In-memory AS path tracking: each path, JSON-encoded, with the earliest
+	// and the latest time an update carried it.
 	mu           sync.Mutex
-	asPaths      map[string]time.Time // key is JSON-encoded AS path
-	droppedPaths int                  // paths dropped because the map was full
+	asPaths      map[string]database.SeenTimes
+	droppedPaths int // paths dropped because the map was full
 	// writing is how many paths processPeerings took from asPaths and has
 	// not finished writing the peerings of.
 	writing int
@@ -64,7 +65,7 @@ func NewPeeringHandler(db database.Store, logger *logger.Logger) *PeeringHandler
 	h := &PeeringHandler{
 		db:      db,
 		logger:  logger,
-		asPaths: make(map[string]time.Time),
+		asPaths: make(map[string]database.SeenTimes),
 		stopCh:  make(chan context.Context, 1),
 		done:    make(chan struct{}),
 	}
@@ -118,16 +119,18 @@ func (h *PeeringHandler) HandleMessage(msg *ristypes.RISMessage) {
 	key := string(pathJSON)
 
 	h.mu.Lock()
-	if _, exists := h.asPaths[key]; exists {
-		// Already tracked: refresh its timestamp.
-		h.asPaths[key] = timestamp
-	} else if len(h.asPaths) >= maxTrackedPaths {
+	defer h.mu.Unlock()
+
+	seen, exists := h.asPaths[key]
+	if !exists && len(h.asPaths) >= maxTrackedPaths {
 		// Map is full; drop this new path and count it.
 		h.droppedPaths++
-	} else {
-		h.asPaths[key] = timestamp
+
+		return
 	}
-	h.mu.Unlock()
+
+	seen.Add(timestamp)
+	h.asPaths[key] = seen
 }
 
 // ProcessPeeringsNow triggers immediate processing of all accumulated AS
@@ -190,7 +193,7 @@ func (h *PeeringHandler) processPeerings() {
 	// released, so the map never grows past a single interval's traffic.
 	h.mu.Lock()
 	paths := h.asPaths
-	h.asPaths = make(map[string]time.Time)
+	h.asPaths = make(map[string]database.SeenTimes)
 	h.writing = len(paths)
 	dropped := h.droppedPaths
 	h.mu.Unlock()
@@ -204,9 +207,13 @@ func (h *PeeringHandler) processPeerings() {
 		low, high int
 	}
 
-	peerings := make(map[peeringKey]time.Time)
+	// The earliest and the latest time an update carried each peering. Several
+	// paths, each carried at several times, can hold a peering, so a new
+	// peering is stored first seen at the earliest of them and last seen at the
+	// latest.
+	peerings := make(map[peeringKey]database.SeenTimes)
 
-	for pathJSON, timestamp := range paths {
+	for pathJSON, pathSeen := range paths {
 		var path []int
 
 		err := json.Unmarshal([]byte(pathJSON), &path)
@@ -233,10 +240,11 @@ func (h *PeeringHandler) processPeerings() {
 			}
 
 			key := peeringKey{low: low, high: high}
-			// Update timestamp if this is newer
-			if existing, ok := peerings[key]; !ok || timestamp.After(existing) {
-				peerings[key] = timestamp
-			}
+
+			seen := peerings[key]
+			seen.Add(pathSeen.First)
+			seen.Add(pathSeen.Last)
+			peerings[key] = seen
 		}
 	}
 
@@ -244,8 +252,8 @@ func (h *PeeringHandler) processPeerings() {
 	start := time.Now()
 	successCount := 0
 
-	for key, ts := range peerings {
-		err := h.db.RecordPeering(key.low, key.high, ts)
+	for key, seen := range peerings {
+		err := h.db.RecordPeering(key.low, key.high, seen)
 		if err != nil {
 			h.logger.Error("Failed to record peering",
 				"as_a", key.low,

internal/routewatch/peeringhandler_internal_test.go

@@ -31,12 +31,14 @@ type recordingStore struct {
 	calls int
 }
 
-func (r *recordingStore) RecordPeering(asA, asB int, ts time.Time) error {
+func (r *recordingStore) RecordPeering(
+	asA, asB int, seen database.SeenTimes,
+) error {
 	r.mu.Lock()
 	r.calls++
 	r.mu.Unlock()
 
-	return r.mockStore.RecordPeering(asA, asB, ts)
+	return r.mockStore.RecordPeering(asA, asB, seen)
 }
 
 func (r *recordingStore) callCount() int {
@@ -52,7 +54,7 @@ func newTestHandler(db database.Store) *PeeringHandler {
 	return &PeeringHandler{
 		db:      db,
 		logger:  logger.New(),
-		asPaths: make(map[string]time.Time),
+		asPaths: make(map[string]database.SeenTimes),
 	}
 }
 
@@ -220,10 +222,10 @@ func TestPeeringHandlerCapDropsAndCounts(t *testing.T) {
 	existing := pathKey(t, testASNA, testASNB)
 
 	h.mu.Lock()
-	h.asPaths[existing] = now
+	h.asPaths[existing] = database.SeenTimes{First: now, Last: now}
 
 	for i := 0; len(h.asPaths) < maxTrackedPaths; i++ {
-		h.asPaths[strconv.Itoa(i)] = now
+		h.asPaths[strconv.Itoa(i)] = database.SeenTimes{First: now, Last: now}
 	}
 	h.mu.Unlock()
 
@@ -253,11 +255,53 @@ func TestPeeringHandlerCapDropsAndCounts(t *testing.T) {
 	}
 
 	h.mu.Lock()
-	gotTS := h.asPaths[existing]
+	gotSeen := h.asPaths[existing]
 	h.mu.Unlock()
 
-	if !gotTS.Equal(refreshed) {
-		t.Fatalf("existing path timestamp not refreshed: want %v, got %v",
-			refreshed, gotTS)
+	if !gotSeen.Last.Equal(refreshed) {
+		t.Fatalf("existing path last seen not refreshed: want %v, got %v",
+			refreshed, gotSeen.Last)
+	}
+}
+
+// TestPeeringHandlerStoresTheEarliestAndLatestTimesOfANewPeering handles two
+// updates whose path carries a peering the database does not hold yet, the
+// second 5 seconds after the first, in one run into a real database. The
+// peering is stored first seen at the time of the first update and last seen
+// at the time of the second.
+func TestPeeringHandlerStoresTheEarliestAndLatestTimesOfANewPeering(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	db := newTestDatabase(t)
+	h := newTestHandler(db)
+
+	first := peerAnnouncement(testPeerIP, testV4Prefix)
+	second := peerAnnouncement(testPeerIP, testV4Prefix)
+	second.ParsedTimestamp = first.ParsedTimestamp.Add(secondUpdateAfter)
+
+	h.HandleMessage(first)
+	h.HandleMessage(second)
+	h.ProcessPeeringsNow()
+
+	peers, err := db.GetASPeers(testASNA)
+	if err != nil {
+		t.Fatalf("GetASPeers(%d): %v", testASNA, err)
+	}
+
+	if len(peers) != 1 || peers[0].ASN != testASNB {
+		t.Fatalf("peers of AS%d = %+v, want AS%d alone", testASNA, peers,
+			testASNB)
+	}
+
+	if !peers[0].FirstSeen.Equal(first.ParsedTimestamp) {
+		t.Errorf("first_seen of peering AS%d-AS%d = %s, want %s", testASNA,
+			testASNB, peers[0].FirstSeen, first.ParsedTimestamp)
+	}
+
+	if !peers[0].LastSeen.Equal(second.ParsedTimestamp) {
+		t.Errorf("last_seen of peering AS%d-AS%d = %s, want %s", testASNA,
+			testASNB, peers[0].LastSeen, second.ParsedTimestamp)
 	}
 }

internal/routewatch/prefixhandler.go

@@ -347,8 +347,19 @@ func (h *PrefixHandler) flushBatch() {
 
 	lastUpdates := make(map[prefixAndPeer]prefixUpdate)
 
+	// Each prefix the batch names, announced or withdrawn, with the earliest
+	// timestamp among all of its updates, not only each peer's last one: the
+	// prefix tables store only first_seen, written when the prefix is new and
+	// never changed after.
+	prefixesToUpdate := make(map[string]time.Time)
+
 	for _, update := range batch {
 		lastUpdates[prefixAndPeer{update.prefix, update.peer}] = update
+
+		seen, exists := prefixesToUpdate[update.prefix]
+		if !exists || update.timestamp.Before(seen) {
+			prefixesToUpdate[update.prefix] = update.timestamp
+		}
 	}
 
 	// Collect routes to upsert and delete
@@ -357,16 +368,7 @@ func (h *PrefixHandler) flushBatch() {
 		routesToDelete []database.LiveRouteDeletion
 	)
 
-	// Collect unique prefixes to update
-	prefixesToUpdate := make(map[string]time.Time)
-
 	for _, update := range lastUpdates {
-		// Track prefix for both announcements and withdrawals
-		seen, exists := prefixesToUpdate[update.prefix]
-		if !exists || update.timestamp.After(seen) {
-			prefixesToUpdate[update.prefix] = update.timestamp
-		}
-
 		// An announcement is stored when its path gives an origin, or when it
 		// ends in an AS set of several members: that route is stored with
 		// origin 0 (none). An announcement with an empty path is not stored.

internal/routewatch/prefixhandler_internal_test.go

@@ -26,6 +26,10 @@ const (
 	testV6Prefix = "2001:db8:1::/48"
 )
 
+// secondUpdateAfter is how long after the first update the tests of first_seen
+// and last_seen send the second, in the same batch.
+const secondUpdateAfter = 5 * time.Second
+
 // TestPrefixHandlerStoresPrefixesTheIPLookupFinds runs announcements and
 // withdrawals through the prefix handler into a real database and looks the
 // addresses up. The prefixes are written the way the feed sends them: IPv6
@@ -83,6 +87,49 @@ func TestPrefixHandlerStoresPrefixesTheIPLookupFinds(t *testing.T) {
 	}
 }
 
+// TestPrefixHandlerStoresANewPrefixFirstSeenAtItsEarliestUpdate announces a
+// prefix the database does not hold yet twice in one batch, the second time 5
+// seconds after the first, through the prefix handler into a real database. The
+// prefix is stored first seen at the time of the first announcement, whether
+// the two come from two peers or from one, and the IP lookup answers with that
+// time.
+func TestPrefixHandlerStoresANewPrefixFirstSeenAtItsEarliestUpdate(
+	t *testing.T,
+) {
+	t.Parallel()
+
+	const otherPeerIP = "2001:db8:ffff::2"
+
+	tests := []struct {
+		name                  string
+		firstPeer, secondPeer string
+	}{
+		{name: "two peers", firstPeer: testPeerIP, secondPeer: otherPeerIP},
+		{name: "one peer", firstPeer: testPeerIP, secondPeer: testPeerIP},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			first := peerAnnouncement(tt.firstPeer, testV4Prefix)
+			second := peerAnnouncement(tt.secondPeer, testV4Prefix)
+			second.ParsedTimestamp = first.ParsedTimestamp.Add(secondUpdateAfter)
+			db := handleBatches(t, [][]*ristypes.RISMessage{{first, second}})
+
+			info, err := db.GetIPInfo("192.0.2.1")
+			if err != nil {
+				t.Fatalf("GetIPInfo(192.0.2.1): %v", err)
+			}
+
+			if !info.FirstSeen.Equal(first.ParsedTimestamp) {
+				t.Errorf("first_seen of %s = %s, want %s", testV4Prefix,
+					info.FirstSeen, first.ParsedTimestamp)
+			}
+		})
+	}
+}
+
 // TestPrefixHandlerAppliesEachPeersLastUpdateInABatch runs updates for one
 // prefix through the prefix handler into a real database. All updates carry the
 // same timestamp, as updates within one second of the feed do. Within a batch,