Sample the prefix-length pages with bounded lookups, each prefix once (closes #66)

/prefixlength/{length} and /prefixlength6/{length} grouped every live route
of that length by prefix, shuffled all the prefixes and joined the routes
back, so a request grew with the number of routes and listed a prefix once
for each peer that shared its newest time.

Now a length with at most 5000 prefixes has them all read and shuffled, and
the page lists 500 of them. A longer one is sampled with random row-id
lookups on the mask length index, at most four per prefix asked for, filled
up from the shuffled prefixes if the lookups keep repeating. Each prefix is
shown once, with its newest route.

Model: opus-5-5

internal/database/counts_test.go

@@ -128,7 +128,7 @@ func TestLiveCountsTrackWritesInRealtime(t *testing.T) {
 
 	// A withdrawal removes one route.
 	if err := db.DeleteLiveRouteBatch([]LiveRouteDeletion{
-		{Prefix: "203.0.113.0/24", PeerIP: "192.0.2.1", IPVersion: ipVersionV4},
+		{Prefix: "203.0.113.0/24", OriginASN: 64501, PeerIP: "192.0.2.1", IPVersion: ipVersionV4},
 	}); err != nil {
 		t.Fatalf("DeleteLiveRouteBatch: %v", err)
 	}
@@ -327,11 +327,13 @@ func announceRoutes(db *Database, batch bool, routes []*LiveRoute) error {
 }
 
 // withdrawRoutes removes routes in one DeleteLiveRouteBatch, or with one
-// DeleteLiveRoute each.
+// DeleteLiveRoute each. It names each route by prefix and peer only, with no
+// origin ASN, as a withdrawal from the feed does when its message carries no AS
+// path.
 func withdrawRoutes(db *Database, batch bool, routes []*LiveRoute) error {
 	if !batch {
 		for _, route := range routes {
-			if err := db.DeleteLiveRoute(route.Prefix, route.PeerIP); err != nil {
+			if err := db.DeleteLiveRoute(route.Prefix, 0, route.PeerIP); err != nil {
 				return err
 			}
 		}

internal/database/database.go

@@ -8,6 +8,7 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
+	"math/rand/v2"
 	"net"
 	"net/netip"
 	"os"
@@ -61,6 +62,17 @@ const (
 	sqliteHardHeapLimitBytes = 1610612736
 )
 
+// Bounds on the work GetRandomPrefixesByLength does to pick its sample.
+const (
+	// maxPrefixesToShuffle is the most prefixes of one mask length that are
+	// read in full and shuffled to pick the sample. A mask length with more is
+	// sampled with random lookups.
+	maxPrefixesToShuffle = 5000
+	// sampleLookupsPerPrefix is the most random lookups made for each prefix
+	// asked for.
+	sampleLookupsPerPrefix = 4
+)
+
 // Common errors
 var (
 	// ErrInvalidIP is returned when an IP address is malformed
@@ -228,22 +240,21 @@ func (d *Database) beginTx() (*loggingTx, error) {
 }
 
 // A live-route upsert is an UPDATE followed, only when no row matched, by an
-// INSERT. A live route is keyed by prefix and peer, so an announcement replaces
-// the peer's route for the prefix whatever its origin AS. The UPDATE's
-// rows-affected count (1 for an existing key, 0 for a new one) is what lets the
-// in-memory route counters stay exact without a COUNT(*). Callers hold the
-// database write lock, so no other writer can insert the same key between the
-// two statements. The id column is set only on INSERT, so an updated route
-// keeps its original id, exactly as the previous ON CONFLICT upsert did.
+// INSERT. The UPDATE's rows-affected count (1 for an existing key, 0 for a new
+// one) is what lets the in-memory route counters stay exact without a COUNT(*).
+// Callers hold the database write lock, so no other writer can insert the same
+// key between the two statements. The id column is set only on INSERT, so an
+// updated route keeps its original id, exactly as the previous ON CONFLICT
+// upsert did.
 const (
-	updateLiveRouteV4SQL = `UPDATE live_routes_v4 SET mask_length = ?, origin_asn = ?, as_path = ?,
-		next_hop = ?, last_updated = ? WHERE prefix = ? AND peer_ip = ?`
+	updateLiveRouteV4SQL = `UPDATE live_routes_v4 SET mask_length = ?, as_path = ?, next_hop = ?,
+		last_updated = ? WHERE prefix = ? AND origin_asn = ? AND peer_ip = ?`
 
 	insertLiveRouteV4SQL = `INSERT INTO live_routes_v4 (id, prefix, mask_length, origin_asn, peer_ip,
 		as_path, next_hop, last_updated) VALUES (?, ?, ?, ?, ?, ?, ?, ?)`
 
-	updateLiveRouteV6SQL = `UPDATE live_routes_v6 SET mask_length = ?, origin_asn = ?, as_path = ?,
-		next_hop = ?, last_updated = ? WHERE prefix = ? AND peer_ip = ?`
+	updateLiveRouteV6SQL = `UPDATE live_routes_v6 SET mask_length = ?, as_path = ?, next_hop = ?,
+		last_updated = ? WHERE prefix = ? AND origin_asn = ? AND peer_ip = ?`
 
 	insertLiveRouteV6SQL = `INSERT INTO live_routes_v6 (id, prefix, mask_length, origin_asn, peer_ip,
 		as_path, next_hop, last_updated) VALUES (?, ?, ?, ?, ?, ?, ?, ?)`
@@ -262,8 +273,8 @@ const (
 // live route for its prefix.
 func upsertRouteRowV4(upd, ins, has *sql.Stmt, route *LiveRoute, pathJSON string) (
 	inserted, newPrefix bool, err error) {
-	res, err := upd.Exec(route.MaskLength, route.OriginASN, pathJSON, route.NextHop, route.LastUpdated,
-		route.Prefix, route.PeerIP)
+	res, err := upd.Exec(route.MaskLength, pathJSON, route.NextHop, route.LastUpdated,
+		route.Prefix, route.OriginASN, route.PeerIP)
 	if err != nil {
 		return false, false, err
 	}
@@ -294,8 +305,8 @@ func upsertRouteRowV4(upd, ins, has *sql.Stmt, route *LiveRoute, pathJSON string
 // live route for its prefix.
 func upsertRouteRowV6(upd, ins, has *sql.Stmt, route *LiveRoute, pathJSON string) (
 	inserted, newPrefix bool, err error) {
-	res, err := upd.Exec(route.MaskLength, route.OriginASN, pathJSON, route.NextHop, route.LastUpdated,
-		route.Prefix, route.PeerIP)
+	res, err := upd.Exec(route.MaskLength, pathJSON, route.NextHop, route.LastUpdated,
+		route.Prefix, route.OriginASN, route.PeerIP)
 	if err != nil {
 		return false, false, err
 	}
@@ -442,18 +453,32 @@ func (d *Database) DeleteLiveRouteBatch(deletions []LiveRouteDeletion) error {
 		}
 	}()
 
-	// Prepare statements for both IPv4 and IPv6 tables
-	deleteV4, err := tx.Prepare(`DELETE FROM live_routes_v4 WHERE prefix = ? AND peer_ip = ?`)
-	if err != nil {
-		return fmt.Errorf("failed to prepare IPv4 delete statement: %w", err)
-	}
-	defer func() { _ = deleteV4.Close() }()
+	// No longer need to separate deletions since we handle them in the loop below
 
-	deleteV6, err := tx.Prepare(`DELETE FROM live_routes_v6 WHERE prefix = ? AND peer_ip = ?`)
+	// Prepare statements for both IPv4 and IPv6 tables
+	stmtV4WithOrigin, err := tx.Prepare(`DELETE FROM live_routes_v4 WHERE prefix = ? AND origin_asn = ? AND peer_ip = ?`)
 	if err != nil {
-		return fmt.Errorf("failed to prepare IPv6 delete statement: %w", err)
+		return fmt.Errorf("failed to prepare IPv4 delete with origin statement: %w", err)
 	}
-	defer func() { _ = deleteV6.Close() }()
+	defer func() { _ = stmtV4WithOrigin.Close() }()
+
+	stmtV4WithoutOrigin, err := tx.Prepare(`DELETE FROM live_routes_v4 WHERE prefix = ? AND peer_ip = ?`)
+	if err != nil {
+		return fmt.Errorf("failed to prepare IPv4 delete without origin statement: %w", err)
+	}
+	defer func() { _ = stmtV4WithoutOrigin.Close() }()
+
+	stmtV6WithOrigin, err := tx.Prepare(`DELETE FROM live_routes_v6 WHERE prefix = ? AND origin_asn = ? AND peer_ip = ?`)
+	if err != nil {
+		return fmt.Errorf("failed to prepare IPv6 delete with origin statement: %w", err)
+	}
+	defer func() { _ = stmtV6WithOrigin.Close() }()
+
+	stmtV6WithoutOrigin, err := tx.Prepare(`DELETE FROM live_routes_v6 WHERE prefix = ? AND peer_ip = ?`)
+	if err != nil {
+		return fmt.Errorf("failed to prepare IPv6 delete without origin statement: %w", err)
+	}
+	defer func() { _ = stmtV6WithoutOrigin.Close() }()
 
 	hasV4, err := tx.Prepare(prefixHasLiveRouteV4SQL)
 	if err != nil {
@@ -472,18 +497,38 @@ func (d *Database) DeleteLiveRouteBatch(deletions []LiveRouteDeletion) error {
 	// Mask lengths of the prefixes this batch leaves with no live route.
 	var gonePrefixMaskLengthsV4, gonePrefixMaskLengthsV6 []int
 	for _, del := range deletions {
-		stmt := deleteV4
-		if del.IPVersion != ipVersionV4 {
-			stmt = deleteV6
+		var stmt *sql.Stmt
+
+		// Select appropriate statement based on IP version and whether we have origin ASN
+		//nolint:nestif // Clear logic for selecting the right statement
+		if del.IPVersion == ipVersionV4 {
+			if del.OriginASN == 0 {
+				stmt = stmtV4WithoutOrigin
+			} else {
+				stmt = stmtV4WithOrigin
+			}
+		} else {
+			if del.OriginASN == 0 {
+				stmt = stmtV6WithoutOrigin
+			} else {
+				stmt = stmtV6WithOrigin
+			}
+		}
+
+		// Execute deletion
+		var res sql.Result
+		if del.OriginASN == 0 {
+			res, err = stmt.Exec(del.Prefix, del.PeerIP)
+		} else {
+			res, err = stmt.Exec(del.Prefix, del.OriginASN, del.PeerIP)
 		}
 
-		res, err := stmt.Exec(del.Prefix, del.PeerIP)
 		if err != nil {
 			return fmt.Errorf("failed to delete route %s: %w", del.Prefix, err)
 		}
 
-		// The feed often withdraws a route that is not live, which deletes
-		// nothing, so the in-memory route counters use the rows-affected count.
+		// A deletion with no origin ASN can remove several rows, so use the
+		// exact rows-affected count to keep the in-memory route counters right.
 		affected, err := res.RowsAffected()
 		if err != nil {
 			return fmt.Errorf("failed to count deleted route %s: %w", del.Prefix, err)
@@ -1249,8 +1294,9 @@ func (d *Database) UpsertLiveRoute(route *LiveRoute) error {
 	return nil
 }
 
-// DeleteLiveRoute deletes the live route the peer has for the prefix
-func (d *Database) DeleteLiveRoute(prefix, peerIP string) error {
+// DeleteLiveRoute deletes a live route
+// If originASN is 0, deletes all routes for the prefix/peer combination
+func (d *Database) DeleteLiveRoute(prefix string, originASN int, peerIP string) error {
 	d.lock("DeleteLiveRoute")
 	defer d.unlock()
 
@@ -1275,12 +1321,22 @@ func (d *Database) DeleteLiveRoute(prefix, peerIP string) error {
 	}()
 
 	// Literal per-table queries (rather than one formatted with the table name)
-	// so the delete carries no dynamically built SQL.
+	// so the delete carries no dynamically built SQL. A delete with no origin
+	// ASN can remove several rows.
 	var res sql.Result
-	if isV4 {
+	switch {
+	case isV4 && originASN == 0:
 		res, err = tx.Exec(`DELETE FROM live_routes_v4 WHERE prefix = ? AND peer_ip = ?`, prefix, peerIP)
-	} else {
+	case isV4:
+		res, err = tx.Exec(
+			`DELETE FROM live_routes_v4 WHERE prefix = ? AND origin_asn = ? AND peer_ip = ?`,
+			prefix, originASN, peerIP)
+	case originASN == 0:
 		res, err = tx.Exec(`DELETE FROM live_routes_v6 WHERE prefix = ? AND peer_ip = ?`, prefix, peerIP)
+	default:
+		res, err = tx.Exec(
+			`DELETE FROM live_routes_v6 WHERE prefix = ? AND origin_asn = ? AND peer_ip = ?`,
+			prefix, originASN, peerIP)
 	}
 	if err != nil {
 		return err
@@ -1660,45 +1716,183 @@ func (d *Database) GetRandomPrefixesByLength(maskLength, ipVersion, limit int) (
 	return d.GetRandomPrefixesByLengthContext(context.Background(), maskLength, ipVersion, limit)
 }
 
-// GetRandomPrefixesByLengthContext returns a random sample of prefixes with context support
+// GetRandomPrefixesByLengthContext returns limit prefixes with the given mask
+// length picked at random, or all of them when there are fewer, each once with
+// its most recently updated live route, for a limit up to
+// maxPrefixesToShuffle. However many routes there are, it reads at most
+// maxPrefixesToShuffle+1 prefixes and makes at most sampleLookupsPerPrefix
+// lookups for each prefix asked for.
 func (d *Database) GetRandomPrefixesByLengthContext(
 	ctx context.Context, maskLength, ipVersion, limit int) ([]LiveRoute, error) {
-	// Select unique prefixes with their most recent route information
 	tableName := "live_routes_v4"
 	if ipVersion == ipVersionV6 {
 		tableName = "live_routes_v6"
 	}
 
+	// Reading one prefix more than maxPrefixesToShuffle tells whether these
+	// are all the prefixes with this mask length.
+	prefixes, err := d.firstPrefixesByLength(ctx, tableName, maskLength, maxPrefixesToShuffle+1)
+	if err != nil {
+		return nil, err
+	}
+	rand.Shuffle(len(prefixes), func(i, j int) {
+		prefixes[i], prefixes[j] = prefixes[j], prefixes[i]
+	})
+	if len(prefixes) > maxPrefixesToShuffle {
+		prefixes, err = d.samplePrefixesByLength(ctx, tableName, maskLength, limit, prefixes)
+		if err != nil {
+			return nil, err
+		}
+	}
+	prefixes = prefixes[:min(limit, len(prefixes))]
+
+	return d.newestRouteOfEachPrefix(ctx, tableName, ipVersion, prefixes)
+}
+
+// firstPrefixesByLength returns the first n distinct prefixes, in text order,
+// with the given mask length. It reads only their entries in the mask length
+// and prefix index.
+func (d *Database) firstPrefixesByLength(
+	ctx context.Context, tableName string, maskLength, n int) ([]string, error) {
 	//nolint:gosec // Table name is hardcoded based on IP version
 	query := fmt.Sprintf(`
-		WITH unique_prefixes AS (
-			SELECT prefix, MAX(last_updated) as max_updated
-			FROM %s
-			WHERE mask_length = ?
-			GROUP BY prefix
-			ORDER BY RANDOM()
-			LIMIT ?
-		)
-		SELECT lr.prefix, lr.mask_length, lr.origin_asn, lr.as_path, 
-			lr.peer_ip, lr.last_updated
-		FROM %s lr
-		INNER JOIN unique_prefixes up ON lr.prefix = up.prefix AND lr.last_updated = up.max_updated
-		WHERE lr.mask_length = ?
-	`, tableName, tableName)
+		SELECT DISTINCT prefix FROM %s
+		WHERE mask_length = ?
+		ORDER BY prefix
+		LIMIT ?
+	`, tableName)
 
-	rows, err := d.db.QueryContext(ctx, query, maskLength, limit, maskLength)
+	rows, err := d.db.QueryContext(ctx, query, maskLength, n)
 	if err != nil {
-		return nil, fmt.Errorf("failed to query random prefixes: %w", err)
+		return nil, fmt.Errorf("failed to list prefixes: %w", err)
 	}
-	defer func() {
-		_ = rows.Close()
-	}()
+	defer func() { _ = rows.Close() }()
 
-	var routes []LiveRoute
+	var prefixes []string
 	for rows.Next() {
-		var route LiveRoute
+		var prefix string
+		if err := rows.Scan(&prefix); err != nil {
+			return nil, fmt.Errorf("failed to scan prefix: %w", err)
+		}
+		prefixes = append(prefixes, prefix)
+	}
+	if err := rows.Err(); err != nil {
+		return nil, fmt.Errorf("failed to list prefixes: %w", err)
+	}
+
+	return prefixes, nil
+}
+
+// samplePrefixesByLength picks limit distinct prefixes with the given mask
+// length at random, in at most sampleLookupsPerPrefix lookups for each prefix
+// asked for. ORDER BY RANDOM() would read every route with that mask length.
+// Instead each lookup draws a row id between those of the first and last route
+// with that mask length and takes the prefix of the first such route at or
+// after it. SQLite keeps the row id at the end of every index entry, so the
+// mask length index is ordered by mask length, then row id, and the lookup is
+// one search of it.
+//
+// The sample is uneven, as the page needs a varied sample, not an even one: a
+// route is picked in proportion to the row ids between it and the previous
+// route with that mask length, including the row ids of routes with other mask
+// lengths and of withdrawn routes. A route that follows a long run of
+// withdrawn routes is picked far more often than one that directly follows
+// another route with that mask length.
+//
+// Should the lookups keep landing on prefixes already picked, the rest are
+// taken from shuffled, prefixes of that length in random order.
+func (d *Database) samplePrefixesByLength(
+	ctx context.Context, tableName string, maskLength, limit int, shuffled []string) ([]string, error) {
+	//nolint:gosec // Table name is hardcoded based on IP version
+	rangeQuery := fmt.Sprintf(`
+		SELECT
+			(SELECT MIN(rowid) FROM %[1]s WHERE mask_length = ?),
+			(SELECT MAX(rowid) FROM %[1]s WHERE mask_length = ?)
+	`, tableName)
+
+	var first, last sql.NullInt64
+	if err := d.db.QueryRowContext(ctx, rangeQuery, maskLength, maskLength).Scan(&first, &last); err != nil {
+		return nil, fmt.Errorf("failed to read row id range: %w", err)
+	}
+	if !first.Valid {
+		// Every route with this mask length was withdrawn since they were listed.
+		return nil, nil
+	}
+
+	//nolint:gosec // Table name is hardcoded based on IP version
+	lookupQuery := fmt.Sprintf(`
+		SELECT prefix FROM %s
+		WHERE mask_length = ? AND rowid >= ?
+		ORDER BY rowid
+		LIMIT 1
+	`, tableName)
+
+	stmt, err := d.db.PrepareContext(ctx, lookupQuery)
+	if err != nil {
+		return nil, fmt.Errorf("failed to prepare prefix lookup: %w", err)
+	}
+	defer func() { _ = stmt.Close() }()
+
+	picked := make(map[string]bool, limit)
+	prefixes := make([]string, 0, limit)
+	pick := func(prefix string) {
+		if !picked[prefix] {
+			picked[prefix] = true
+			prefixes = append(prefixes, prefix)
+		}
+	}
+
+	for lookups := 0; lookups < limit*sampleLookupsPerPrefix && len(prefixes) < limit; lookups++ {
+		//nolint:gosec // The sample needs variety, not unpredictability.
+		rowID := first.Int64 + rand.Int64N(last.Int64-first.Int64+1)
+
+		var prefix string
+		err := stmt.QueryRowContext(ctx, maskLength, rowID).Scan(&prefix)
+		if errors.Is(err, sql.ErrNoRows) {
+			// The routes from rowID on were withdrawn since the range was read.
+			continue
+		}
+		if err != nil {
+			return nil, fmt.Errorf("failed to look up prefix: %w", err)
+		}
+		pick(prefix)
+	}
+
+	for _, prefix := range shuffled {
+		if len(prefixes) == limit {
+			break
+		}
+		pick(prefix)
+	}
+
+	return prefixes, nil
+}
+
+// newestRouteOfEachPrefix returns the most recently updated live route of each
+// prefix, one lookup on the prefix index per prefix. A prefix whose routes were
+// all withdrawn since it was picked is left out.
+func (d *Database) newestRouteOfEachPrefix(
+	ctx context.Context, tableName string, ipVersion int, prefixes []string) ([]LiveRoute, error) {
+	//nolint:gosec // Table name is hardcoded based on IP version
+	query := fmt.Sprintf(`
+		SELECT prefix, mask_length, origin_asn, as_path, peer_ip, last_updated
+		FROM %s
+		WHERE prefix = ?
+		ORDER BY last_updated DESC
+		LIMIT 1
+	`, tableName)
+
+	stmt, err := d.db.PrepareContext(ctx, query)
+	if err != nil {
+		return nil, fmt.Errorf("failed to prepare route query: %w", err)
+	}
+	defer func() { _ = stmt.Close() }()
+
+	routes := make([]LiveRoute, 0, len(prefixes))
+	for _, prefix := range prefixes {
+		route := LiveRoute{IPVersion: ipVersion}
 		var pathJSON string
-		err := rows.Scan(
+		err := stmt.QueryRowContext(ctx, prefix).Scan(
 			&route.Prefix,
 			&route.MaskLength,
 			&route.OriginASN,
@@ -1706,17 +1900,16 @@ func (d *Database) GetRandomPrefixesByLengthContext(
 			&route.PeerIP,
 			&route.LastUpdated,
 		)
-		// Set IP version based on which table we queried
-		route.IPVersion = ipVersion
-		if err != nil {
+		if errors.Is(err, sql.ErrNoRows) {
 			continue
 		}
+		if err != nil {
+			return nil, fmt.Errorf("failed to query route of %s: %w", prefix, err)
+		}
 
-		// Decode AS path
 		if err := json.Unmarshal([]byte(pathJSON), &route.ASPath); err != nil {
 			route.ASPath = []int{}
 		}
-
 		routes = append(routes, route)
 	}
 

internal/database/database_test.go

@@ -6,6 +6,7 @@ import (
 	"errors"
 	"fmt"
 	"net/netip"
+	"slices"
 	"sync"
 	"testing"
 	"time"
@@ -45,6 +46,32 @@ const (
 	asnSecondBand = 100
 )
 
+// Parameters for the prefix-length sample tests.
+const (
+	// sampleLimit is how many prefixes the tests ask the sample for.
+	sampleLimit = 10
+	// sampledPrefixCount is how many prefixes the sampling test stores, more
+	// than sampleLimit so the sample has to choose among them.
+	sampledPrefixCount = 30
+	// secondV4PeerIP and secondV6PeerIP are the second peer each prefix is
+	// learned from, besides the one mkV4Route and mkV6Route set.
+	secondV4PeerIP = "192.0.2.2"
+	secondV6PeerIP = "2001:db8::2"
+	// largeSamplePrefixCount is how many prefixes the test of a mask length
+	// with more than maxPrefixesToShuffle prefixes stores: twice that, so about
+	// half of them come after the first maxPrefixesToShuffle+1 in text order.
+	largeSamplePrefixCount = 2 * maxPrefixesToShuffle
+	// largeSampleLimit is how many prefixes that test asks for. Each random
+	// lookup lands after the first maxPrefixesToShuffle+1 in text order about
+	// half the time, so none of them doing so is about as likely as
+	// largeSampleLimit coin tosses all coming up heads.
+	largeSampleLimit = 30
+	// liveRoutesV4Table and liveRoutesV6Table are the live route tables the
+	// tests pass to the sampling steps.
+	liveRoutesV4Table = "live_routes_v4"
+	liveRoutesV6Table = "live_routes_v6"
+)
+
 // TestGetIPInfoFindsMostSpecificLiveRoute stores nested live prefixes for both
 // families and checks that a lookup returns the most specific one covering the
 // address, ErrNoRoute when none covers it, and the next less specific prefix
@@ -122,8 +149,8 @@ func TestGetIPInfoFindsMostSpecificLiveRoute(t *testing.T) {
 	lookup("2001:db9::1", "")
 
 	err = db.DeleteLiveRouteBatch([]LiveRouteDeletion{
-		{Prefix: "10.1.2.0/24", PeerIP: testPeerIP, IPVersion: ipVersionV4},
-		{Prefix: "2001:db8:1:2::/64", PeerIP: testPeerIP, IPVersion: ipVersionV6},
+		{Prefix: "10.1.2.0/24", OriginASN: 64502, PeerIP: testPeerIP, IPVersion: ipVersionV4},
+		{Prefix: "2001:db8:1:2::/64", OriginASN: 64502, PeerIP: testPeerIP, IPVersion: ipVersionV6},
 	})
 	if err != nil {
 		t.Fatalf("DeleteLiveRouteBatch: %v", err)
@@ -306,6 +333,7 @@ func TestVacuumReturnsFreePages(t *testing.T) {
 		routes = append(routes, route)
 		deletions = append(deletions, LiveRouteDeletion{
 			Prefix:    route.Prefix,
+			OriginASN: asn,
 			PeerIP:    route.PeerIP,
 			IPVersion: ipVersionV4,
 		})
@@ -388,3 +416,281 @@ func TestBatchWriteDuringCheckpoint(t *testing.T) {
 	cancel()
 	wg.Wait()
 }
+
+// TestGetRandomPrefixesByLengthListsPrefixOnce stores one prefix of each
+// family learned from two peers in the same second and checks that the
+// prefix-length sample lists each of them once.
+func TestGetRandomPrefixesByLengthListsPrefixOnce(t *testing.T) {
+	cfg := &config.Config{StateDir: t.TempDir()}
+
+	db, err := New(cfg, logger.New())
+	if err != nil {
+		t.Fatalf("failed to create database: %v", err)
+	}
+	defer func() { _ = db.Close() }()
+
+	ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
+	v4First := mkV4Route(t, "198.51.100.0/24", 64500, ts)
+	v4Second := mkV4Route(t, "198.51.100.0/24", 64500, ts)
+	v4Second.PeerIP = secondV4PeerIP
+	v6First := mkV6Route("2001:db8::/32", 64500, ts)
+	v6Second := mkV6Route("2001:db8::/32", 64500, ts)
+	v6Second.PeerIP = secondV6PeerIP
+	if err := db.UpsertLiveRouteBatch([]*LiveRoute{v4First, v4Second, v6First, v6Second}); err != nil {
+		t.Fatalf("UpsertLiveRouteBatch: %v", err)
+	}
+
+	for _, want := range []*LiveRoute{v4First, v6First} {
+		routes, err := db.GetRandomPrefixesByLength(want.MaskLength, want.IPVersion, sampleLimit)
+		if err != nil {
+			t.Fatalf("GetRandomPrefixesByLength(%d, %d): %v", want.MaskLength, want.IPVersion, err)
+		}
+		if len(routes) != 1 || routes[0].Prefix != want.Prefix {
+			t.Errorf("GetRandomPrefixesByLength(%d, %d) = %+v, want %s listed once",
+				want.MaskLength, want.IPVersion, routes, want.Prefix)
+		}
+	}
+}
+
+// TestGetRandomPrefixesByLengthListsNewestRoute stores a prefix learned from
+// two peers at different times, the older route first, and checks that the
+// prefix-length sample lists the newer route.
+func TestGetRandomPrefixesByLengthListsNewestRoute(t *testing.T) {
+	cfg := &config.Config{StateDir: t.TempDir()}
+
+	db, err := New(cfg, logger.New())
+	if err != nil {
+		t.Fatalf("failed to create database: %v", err)
+	}
+	defer func() { _ = db.Close() }()
+
+	older := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
+	olderRoute := mkV4Route(t, "198.51.100.0/24", 64500, older)
+	newerRoute := mkV4Route(t, "198.51.100.0/24", 64500, older.Add(time.Minute))
+	newerRoute.PeerIP = secondV4PeerIP
+	if err := db.UpsertLiveRouteBatch([]*LiveRoute{olderRoute, newerRoute}); err != nil {
+		t.Fatalf("UpsertLiveRouteBatch: %v", err)
+	}
+
+	routes, err := db.GetRandomPrefixesByLength(newerRoute.MaskLength, ipVersionV4, sampleLimit)
+	if err != nil {
+		t.Fatalf("GetRandomPrefixesByLength: %v", err)
+	}
+	if len(routes) != 1 || routes[0].PeerIP != newerRoute.PeerIP {
+		t.Errorf("GetRandomPrefixesByLength = %+v, want only the newer route, from %s",
+			routes, newerRoute.PeerIP)
+	}
+}
+
+// TestGetRandomPrefixesByLengthSamplesManyPrefixes stores more /24 prefixes
+// than the sample asks for, each followed by a /16 so that routes of another
+// mask length sit between the /24 rows, and each prefix learned from two peers
+// in the same second. It checks both the sample the pages get, picked from the
+// shuffled prefixes, and the random lookups used for a mask length with more
+// than maxPrefixesToShuffle prefixes.
+func TestGetRandomPrefixesByLengthSamplesManyPrefixes(t *testing.T) {
+	cfg := &config.Config{StateDir: t.TempDir()}
+
+	db, err := New(cfg, logger.New())
+	if err != nil {
+		t.Fatalf("failed to create database: %v", err)
+	}
+	defer func() { _ = db.Close() }()
+
+	ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
+	var routes []*LiveRoute
+	var stored []string
+	for i := range sampledPrefixCount {
+		prefix := fmt.Sprintf("10.0.%d.0/24", i)
+		for _, p := range []string{prefix, fmt.Sprintf("172.%d.0.0/16", i)} {
+			first := mkV4Route(t, p, 64500, ts)
+			second := mkV4Route(t, p, 64500, ts)
+			second.PeerIP = secondV4PeerIP
+			routes = append(routes, first, second)
+		}
+		stored = append(stored, prefix)
+	}
+	if err := db.UpsertLiveRouteBatch(routes); err != nil {
+		t.Fatalf("UpsertLiveRouteBatch: %v", err)
+	}
+	slices.Sort(stored)
+
+	maskLength := routes[0].MaskLength
+	sample, err := db.GetRandomPrefixesByLength(maskLength, ipVersionV4, sampleLimit)
+	if err != nil {
+		t.Fatalf("GetRandomPrefixesByLength: %v", err)
+	}
+	var listed []string
+	for _, route := range sample {
+		listed = append(listed, route.Prefix)
+	}
+	checkPrefixSample(t, "GetRandomPrefixesByLength", listed, stored)
+
+	looked, err := db.samplePrefixesByLength(
+		context.Background(), liveRoutesV4Table, maskLength, sampleLimit, nil)
+	if err != nil {
+		t.Fatalf("samplePrefixesByLength: %v", err)
+	}
+	checkPrefixSample(t, "samplePrefixesByLength", looked, stored)
+}
+
+// checkPrefixSample checks that sample lists sampleLimit prefixes, each once
+// and each one of stored, and that it is not just the first sampleLimit of
+// stored, which is in text order.
+func checkPrefixSample(t *testing.T, name string, sample, stored []string) {
+	t.Helper()
+
+	if len(sample) != sampleLimit {
+		t.Errorf("%s listed %d prefixes, want %d", name, len(sample), sampleLimit)
+	}
+	listed := make(map[string]bool)
+	for _, prefix := range sample {
+		if listed[prefix] || !slices.Contains(stored, prefix) {
+			t.Errorf("%s lists %s twice or it is not one of the /24s", name, prefix)
+		}
+		listed[prefix] = true
+	}
+	if slices.Equal(slices.Sorted(slices.Values(sample)), stored[:sampleLimit]) {
+		t.Errorf("%s lists just the first %d prefixes in text order", name, sampleLimit)
+	}
+}
+
+// TestGetRandomPrefixesByLengthSamplesPastFirstPrefixes stores more prefixes
+// of one mask length than maxPrefixesToShuffle. It checks that
+// firstPrefixesByLength reads just the first maxPrefixesToShuffle+1 in text
+// order, and that the sample lists the prefixes asked for, each once, some of
+// them after those first ones, which only the random lookups reach.
+func TestGetRandomPrefixesByLengthSamplesPastFirstPrefixes(t *testing.T) {
+	cfg := &config.Config{StateDir: t.TempDir()}
+
+	db, err := New(cfg, logger.New())
+	if err != nil {
+		t.Fatalf("failed to create database: %v", err)
+	}
+	defer func() { _ = db.Close() }()
+
+	ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
+	var routes []*LiveRoute
+	var stored []string
+	for i := range largeSamplePrefixCount {
+		route := mkV6Route(fmt.Sprintf("2001:%x::/32", i), 64500, ts)
+		routes = append(routes, route)
+		stored = append(stored, route.Prefix)
+	}
+	if err := db.UpsertLiveRouteBatch(routes); err != nil {
+		t.Fatalf("UpsertLiveRouteBatch: %v", err)
+	}
+	slices.Sort(stored)
+	firstPrefixes := stored[:maxPrefixesToShuffle+1]
+
+	maskLength := routes[0].MaskLength
+	read, err := db.firstPrefixesByLength(
+		context.Background(), liveRoutesV6Table, maskLength, maxPrefixesToShuffle+1)
+	if err != nil {
+		t.Fatalf("firstPrefixesByLength: %v", err)
+	}
+	if !slices.Equal(read, firstPrefixes) {
+		t.Errorf("firstPrefixesByLength read %d prefixes, want the first %d in text order",
+			len(read), len(firstPrefixes))
+	}
+
+	sample, err := db.GetRandomPrefixesByLength(maskLength, ipVersionV6, largeSampleLimit)
+	if err != nil {
+		t.Fatalf("GetRandomPrefixesByLength: %v", err)
+	}
+	if len(sample) != largeSampleLimit {
+		t.Errorf("GetRandomPrefixesByLength listed %d prefixes, want %d", len(sample), largeSampleLimit)
+	}
+	listed := make(map[string]bool)
+	pastFirstPrefixes := 0
+	for _, route := range sample {
+		if listed[route.Prefix] {
+			t.Errorf("GetRandomPrefixesByLength lists %s twice", route.Prefix)
+		}
+		listed[route.Prefix] = true
+		if !slices.Contains(firstPrefixes, route.Prefix) {
+			pastFirstPrefixes++
+		}
+	}
+	if pastFirstPrefixes == 0 {
+		t.Errorf("GetRandomPrefixesByLength lists only prefixes among the first %d in text order",
+			len(firstPrefixes))
+	}
+}
+
+// TestSamplePrefixesByLengthFillsUpFromShuffled stores a single prefix, so
+// every random lookup lands on it, and checks that the sample still lists
+// sampleLimit prefixes: the stored one, then the shuffled ones in order. The
+// stored prefix is the second shuffled one, so the fill-up must skip it.
+func TestSamplePrefixesByLengthFillsUpFromShuffled(t *testing.T) {
+	cfg := &config.Config{StateDir: t.TempDir()}
+
+	db, err := New(cfg, logger.New())
+	if err != nil {
+		t.Fatalf("failed to create database: %v", err)
+	}
+	defer func() { _ = db.Close() }()
+
+	var shuffled []string
+	for i := range sampledPrefixCount {
+		shuffled = append(shuffled, fmt.Sprintf("10.0.%d.0/24", i))
+	}
+	route := mkV4Route(t, shuffled[1], 64500, time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC))
+	if err := db.UpsertLiveRouteBatch([]*LiveRoute{route}); err != nil {
+		t.Fatalf("UpsertLiveRouteBatch: %v", err)
+	}
+
+	sample, err := db.samplePrefixesByLength(
+		context.Background(), liveRoutesV4Table, route.MaskLength, sampleLimit, shuffled)
+	if err != nil {
+		t.Fatalf("samplePrefixesByLength: %v", err)
+	}
+	want := append([]string{shuffled[1], shuffled[0]}, shuffled[2:sampleLimit]...)
+	if !slices.Equal(sample, want) {
+		t.Errorf("samplePrefixesByLength = %v, want %v", sample, want)
+	}
+}
+
+// TestGetRandomPrefixesByLengthSkipsWithdrawnRoutes withdraws a route after
+// storing it, as can happen between the steps of GetRandomPrefixesByLength. It
+// checks that the random lookups pick nothing for a mask length with no route
+// left, and that only the prefixes that still have a route get their newest
+// route listed.
+func TestGetRandomPrefixesByLengthSkipsWithdrawnRoutes(t *testing.T) {
+	cfg := &config.Config{StateDir: t.TempDir()}
+
+	db, err := New(cfg, logger.New())
+	if err != nil {
+		t.Fatalf("failed to create database: %v", err)
+	}
+	defer func() { _ = db.Close() }()
+
+	ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
+	kept := mkV4Route(t, "198.51.100.0/24", 64500, ts)
+	withdrawn := mkV4Route(t, "172.16.0.0/16", 64500, ts)
+	if err := db.UpsertLiveRouteBatch([]*LiveRoute{kept, withdrawn}); err != nil {
+		t.Fatalf("UpsertLiveRouteBatch: %v", err)
+	}
+	if err := db.DeleteLiveRoute(withdrawn.Prefix, withdrawn.OriginASN, withdrawn.PeerIP); err != nil {
+		t.Fatalf("DeleteLiveRoute: %v", err)
+	}
+
+	ctx := context.Background()
+	sample, err := db.samplePrefixesByLength(
+		ctx, liveRoutesV4Table, withdrawn.MaskLength, sampleLimit, []string{withdrawn.Prefix})
+	if err != nil {
+		t.Fatalf("samplePrefixesByLength: %v", err)
+	}
+	if len(sample) != 0 {
+		t.Errorf("samplePrefixesByLength = %v, want none for a mask length with no route left", sample)
+	}
+
+	routes, err := db.newestRouteOfEachPrefix(
+		ctx, liveRoutesV4Table, ipVersionV4, []string{withdrawn.Prefix, kept.Prefix})
+	if err != nil {
+		t.Fatalf("newestRouteOfEachPrefix: %v", err)
+	}
+	if len(routes) != 1 || routes[0].Prefix != kept.Prefix {
+		t.Errorf("newestRouteOfEachPrefix = %+v, want only the route of %s", routes, kept.Prefix)
+	}
+}

internal/database/interface.go

@@ -55,7 +55,7 @@ type Store interface {
 	// Live route operations
 	UpsertLiveRoute(route *LiveRoute) error
 	UpsertLiveRouteBatch(routes []*LiveRoute) error
-	DeleteLiveRoute(prefix, peerIP string) error
+	DeleteLiveRoute(prefix string, originASN int, peerIP string) error
 	DeleteLiveRouteBatch(deletions []LiveRouteDeletion) error
 	GetPrefixDistribution() (ipv4 []PrefixDistribution, ipv6 []PrefixDistribution, err error)
 	GetPrefixDistributionContext(ctx context.Context) (ipv4 []PrefixDistribution, ipv6 []PrefixDistribution, err error)

internal/database/models.go

@@ -114,6 +114,7 @@ type IPInfo struct {
 // LiveRouteDeletion represents parameters for deleting a live route
 type LiveRouteDeletion struct {
 	Prefix    string
+	OriginASN int
 	PeerIP    string
 	IPVersion int
 }

internal/database/schema.sql

@@ -107,7 +107,7 @@ CREATE TABLE IF NOT EXISTS live_routes_v4 (
 	as_path TEXT NOT NULL, -- JSON array
 	next_hop TEXT NOT NULL,
 	last_updated DATETIME NOT NULL,
-	UNIQUE(prefix, peer_ip) -- a peer has at most one route for a prefix
+	UNIQUE(prefix, origin_asn, peer_ip)
 );
 
 -- IPv6 routing table maintained by PrefixHandler
@@ -120,7 +120,7 @@ CREATE TABLE IF NOT EXISTS live_routes_v6 (
 	as_path TEXT NOT NULL, -- JSON array
 	next_hop TEXT NOT NULL,
 	last_updated DATETIME NOT NULL,
-	UNIQUE(prefix, peer_ip) -- a peer has at most one route for a prefix
+	UNIQUE(prefix, origin_asn, peer_ip)
 );
 
 -- Indexes for live_routes_v4 table

internal/routewatch/app_integration_test.go

@@ -205,7 +205,7 @@ func (m *mockStore) UpsertLiveRoute(route *database.LiveRoute) error {
 }
 
 // DeleteLiveRoute mock implementation
-func (m *mockStore) DeleteLiveRoute(prefix, peerIP string) error {
+func (m *mockStore) DeleteLiveRoute(prefix string, originASN int, peerIP string) error {
 	// Simple mock - just return nil
 	return nil
 }

internal/routewatch/cli.go

@@ -20,6 +20,8 @@ const (
 	debugInterval = 60 * time.Second
 	// bytesPerMB is bytes per megabyte
 	bytesPerMB = 1024 * 1024
+	// failureExitCode is the exit status when RouteWatch stops on an error
+	failureExitCode = 1
 )
 
 // logDebugStats logs goroutine count and memory usage
@@ -83,6 +85,9 @@ func CLIEntry() {
 					go func() {
 						if err := rw.Run(ctx); err != nil {
 							logger.Error("RouteWatch error", "error", err)
+							if err := shutdowner.Shutdown(fx.ExitCode(failureExitCode)); err != nil {
+								logger.Error("Failed to shutdown gracefully", "error", err)
+							}
 						}
 					}()
 

internal/routewatch/prefixhandler.go

@@ -108,14 +108,14 @@ func (h *PrefixHandler) HandleMessage(msg *ristypes.RISMessage) {
 	// Withdrawals are queued before announcements, because the batch keeps
 	// each peer's last update for a prefix: a message that lists a prefix as
 	// both withdrawn and announced keeps the route (RFC 4271 section 4.3).
-	// A withdrawal names only the prefix and peer: the message's AS path, and
-	// so its origin AS, belongs to the announced prefixes.
 	for _, prefix := range msg.Withdrawals {
 		h.batch = append(h.batch, prefixUpdate{
 			prefix:      canonicalPrefix(prefix),
+			originASN:   originASN, // Use the originASN from path if available
 			peer:        msg.Peer,
 			messageType: "withdrawal",
 			timestamp:   timestamp,
+			path:        msg.Path,
 		})
 		// Record withdrawal in metrics
 		if h.metrics != nil {
@@ -221,6 +221,7 @@ func (h *PrefixHandler) flushBatchLocked() {
 			// Create deletion record for batch delete
 			routesToDelete = append(routesToDelete, database.LiveRouteDeletion{
 				Prefix:    update.prefix,
+				OriginASN: update.originASN,
 				PeerIP:    update.peer,
 				IPVersion: ipVersion,
 			})
@@ -424,24 +425,51 @@ func (h *PrefixHandler) processAnnouncementDirect(update prefixUpdate) {
 // processWithdrawalDirect handles removing a route directly without prefix table
 // nolint:unused // kept for potential future use
 func (h *PrefixHandler) processWithdrawalDirect(update prefixUpdate) {
-	if err := h.db.DeleteLiveRoute(update.prefix, update.peer); err != nil {
-		h.logger.Error("Failed to delete live route",
-			"prefix", update.prefix,
-			"peer", update.peer,
-			"error", err,
-		)
+	// For withdrawals, we need to delete the route from live_routes
+	if update.originASN > 0 {
+		if err := h.db.DeleteLiveRoute(update.prefix, update.originASN, update.peer); err != nil {
+			h.logger.Error("Failed to delete live route",
+				"prefix", update.prefix,
+				"origin_asn", update.originASN,
+				"peer", update.peer,
+				"error", err,
+			)
+		}
+	} else {
+		// If no origin ASN, just delete all routes for this prefix from this peer
+		if err := h.db.DeleteLiveRoute(update.prefix, 0, update.peer); err != nil {
+			h.logger.Error("Failed to delete live route (no origin ASN)",
+				"prefix", update.prefix,
+				"peer", update.peer,
+				"error", err,
+			)
+		}
 	}
 }
 
 // processWithdrawal handles removing a route from the live routing table
 // nolint:unused // kept for potential future use
 func (h *PrefixHandler) processWithdrawal(_ *database.Prefix, update prefixUpdate) {
-	if err := h.db.DeleteLiveRoute(update.prefix, update.peer); err != nil {
-		h.logger.Error("Failed to delete live route",
-			"prefix", update.prefix,
-			"peer", update.peer,
-			"error", err,
-		)
+	// For withdrawals, we need to delete the route from live_routes
+	// Since we have the origin ASN from the update, we can delete the specific route
+	if update.originASN > 0 {
+		if err := h.db.DeleteLiveRoute(update.prefix, update.originASN, update.peer); err != nil {
+			h.logger.Error("Failed to delete live route",
+				"prefix", update.prefix,
+				"origin_asn", update.originASN,
+				"peer", update.peer,
+				"error", err,
+			)
+		}
+	} else {
+		// If no origin ASN, just delete all routes for this prefix from this peer
+		if err := h.db.DeleteLiveRoute(update.prefix, 0, update.peer); err != nil {
+			h.logger.Error("Failed to delete live route (no origin ASN)",
+				"prefix", update.prefix,
+				"peer", update.peer,
+				"error", err,
+			)
+		}
 	}
 }
 

internal/routewatch/prefixhandler_test.go

@@ -182,139 +182,32 @@ func TestPrefixHandlerAppliesEachPeersLastUpdateInABatch(t *testing.T) {
 				t.Errorf("peers with a live route for %s = %v, want %v", prefix, gotPeers, tt.wantPeers)
 			}
 
+			stats, err := db.GetStatsContext(ctx)
+			if err != nil {
+				t.Fatalf("GetStatsContext: %v", err)
+			}
+			tableRoutesV4, _, err := db.GetLiveRouteCountsContext(ctx)
+			if err != nil {
+				t.Fatalf("GetLiveRouteCountsContext: %v", err)
+			}
+			if stats.IPv4Routes != len(tt.wantPeers) || tableRoutesV4 != len(tt.wantPeers) {
+				t.Errorf("IPv4 live routes: %d in memory, %d in the table, want %d",
+					stats.IPv4Routes, tableRoutesV4, len(tt.wantPeers))
+			}
+
 			var wantDistribution []database.PrefixDistribution
 			if len(tt.wantPeers) > 0 {
 				wantDistribution = []database.PrefixDistribution{{MaskLength: prefixMaskLength, Count: 1}}
 			}
-			assertIPv4Counts(t, db, len(tt.wantPeers), wantDistribution)
-		})
-	}
-}
-
-// TestPrefixHandlerKeepsOneRoutePerPrefixAndPeer runs updates from one peer
-// through the prefix handler into a real database, flushing each message as its
-// own batch. A peer has at most one route for a prefix: a later announcement
-// replaces it whatever its origin AS, and a withdrawal removes it whatever AS
-// the message's AS path ends in. The in-memory route count and prefix
-// distribution must match the route table.
-func TestPrefixHandlerKeepsOneRoutePerPrefixAndPeer(t *testing.T) {
-	const (
-		prefix      = "198.51.100.0/24"
-		otherPrefix = "203.0.113.0/24"
-		// The mask length of both prefixes.
-		prefixMaskLength = 24
-	)
-
-	ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
-	announce := func(path ristypes.ASPath, announced string) *ristypes.RISMessage {
-		return &ristypes.RISMessage{
-			Peer:            testPeerIP,
-			Path:            path,
-			ParsedTimestamp: ts,
-			Announcements:   []ristypes.RISAnnouncement{{NextHop: testPeerIP, Prefixes: []string{announced}}},
-		}
-	}
-	// One message that announces otherPrefix with origin AS testASNB and
-	// withdraws prefix.
-	announceOtherAndWithdraw := announce(ristypes.ASPath{testASNA, testASNB}, otherPrefix)
-	announceOtherAndWithdraw.Withdrawals = []string{prefix}
-
-	tests := []struct {
-		name     string
-		messages []*ristypes.RISMessage
-		// The origin AS of the peer's live route for each prefix. A prefix
-		// left out must have no live route.
-		wantOrigins map[string]int
-	}{
-		{
-			name: "the peer announces the prefix with a new origin AS",
-			messages: []*ristypes.RISMessage{
-				announce(ristypes.ASPath{testASNA}, prefix),
-				announce(ristypes.ASPath{testASNA, testASNB}, prefix),
-			},
-			wantOrigins: map[string]int{prefix: testASNB},
-		},
-		{
-			name: "one message announces a prefix with another origin AS and withdraws the prefix",
-			messages: []*ristypes.RISMessage{
-				announce(ristypes.ASPath{testASNA}, prefix),
-				announceOtherAndWithdraw,
-			},
-			wantOrigins: map[string]int{otherPrefix: testASNB},
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			db, err := database.New(&config.Config{StateDir: t.TempDir()}, logger.New())
+			queryDistributionV4, _, err := db.GetPrefixDistributionContext(ctx)
 			if err != nil {
-				t.Fatalf("failed to create database: %v", err)
+				t.Fatalf("GetPrefixDistributionContext: %v", err)
 			}
-			defer func() { _ = db.Close() }()
-
-			// Built without NewPrefixHandler's flush timer, so the test flushes itself.
-			h := &PrefixHandler{db: db, logger: logger.New()}
-			for _, msg := range tt.messages {
-				h.HandleMessage(msg)
-				h.mu.Lock()
-				h.flushBatchLocked()
-				h.mu.Unlock()
+			if !slices.Equal(stats.IPv4PrefixDistribution, wantDistribution) ||
+				!slices.Equal(queryDistributionV4, wantDistribution) {
+				t.Errorf("IPv4 prefix distribution: %v in memory, %v from the table, want %v",
+					stats.IPv4PrefixDistribution, queryDistributionV4, wantDistribution)
 			}
-
-			for _, p := range []string{prefix, otherPrefix} {
-				routes, err := db.GetPrefixDetailsContext(context.Background(), p)
-				if err != nil && !errors.Is(err, database.ErrNoRoute) {
-					t.Fatalf("GetPrefixDetailsContext(%s): %v", p, err)
-				}
-				var gotOrigins []int
-				for _, route := range routes {
-					gotOrigins = append(gotOrigins, route.OriginASN)
-				}
-				var wantOrigins []int
-				if origin, ok := tt.wantOrigins[p]; ok {
-					wantOrigins = []int{origin}
-				}
-				if !slices.Equal(gotOrigins, wantOrigins) {
-					t.Errorf("origin ASes of the live routes for %s = %v, want %v", p, gotOrigins, wantOrigins)
-				}
-			}
-
-			wantDistribution := []database.PrefixDistribution{
-				{MaskLength: prefixMaskLength, Count: len(tt.wantOrigins)},
-			}
-			assertIPv4Counts(t, db, len(tt.wantOrigins), wantDistribution)
 		})
 	}
 }
-
-// assertIPv4Counts checks that the IPv4 live route count and prefix
-// distribution are wantRoutes and wantDistribution, both as held in memory and
-// as read from the route table.
-func assertIPv4Counts(t *testing.T, db *database.Database, wantRoutes int,
-	wantDistribution []database.PrefixDistribution) {
-	t.Helper()
-
-	ctx := context.Background()
-	stats, err := db.GetStatsContext(ctx)
-	if err != nil {
-		t.Fatalf("GetStatsContext: %v", err)
-	}
-	tableRoutesV4, _, err := db.GetLiveRouteCountsContext(ctx)
-	if err != nil {
-		t.Fatalf("GetLiveRouteCountsContext: %v", err)
-	}
-	if stats.IPv4Routes != wantRoutes || tableRoutesV4 != wantRoutes {
-		t.Errorf("IPv4 live routes: %d in memory, %d in the table, want %d",
-			stats.IPv4Routes, tableRoutesV4, wantRoutes)
-	}
-
-	queryDistributionV4, _, err := db.GetPrefixDistributionContext(ctx)
-	if err != nil {
-		t.Fatalf("GetPrefixDistributionContext: %v", err)
-	}
-	if !slices.Equal(stats.IPv4PrefixDistribution, wantDistribution) ||
-		!slices.Equal(queryDistributionV4, wantDistribution) {
-		t.Errorf("IPv4 prefix distribution: %v in memory, %v from the table, want %v",
-			stats.IPv4PrefixDistribution, queryDistributionV4, wantDistribution)
-	}
-}

internal/server/server.go

@@ -3,6 +3,8 @@ package server
 
 import (
 	"context"
+	"fmt"
+	"net"
 	"net/http"
 	"strconv"
 	"time"
@@ -53,9 +55,12 @@ func New(db database.Store, streamer *streamer.Streamer, logger *logger.Logger,
 	return s
 }
 
-// Start starts the HTTP server
+// Start listens on the configured port and serves HTTP in the background.
+// It returns an error if it cannot listen, for example because the port is
+// already in use.
 func (s *Server) Start() error {
 	const (
+		network           = "tcp"
 		readHeaderTimeout = 40 * time.Second
 		readTimeout       = 60 * time.Second
 		writeTimeout      = 60 * time.Second
@@ -73,10 +78,15 @@ func (s *Server) Start() error {
 
 	s.logger.Info("Starting HTTP server", "port", s.port, "addr", s.srv.Addr)
 
-	// Start in goroutine but log when actually listening
+	listener, err := net.Listen(network, s.srv.Addr)
+	if err != nil {
+		return fmt.Errorf("starting HTTP server: %w", err)
+	}
+
+	s.logger.Info("HTTP server listening", "addr", listener.Addr().String())
+
 	go func() {
-		s.logger.Info("HTTP server listening", "addr", s.srv.Addr)
-		if err := s.srv.ListenAndServe(); err != nil && err != http.ErrServerClosed {
+		if err := s.srv.Serve(listener); err != nil && err != http.ErrServerClosed {
 			s.logger.Error("HTTP server error", "error", err)
 		}
 	}()

internal/server/server_test.go

@@ -0,0 +1,33 @@
+package server
+
+import (
+	"context"
+	"net"
+	"testing"
+
+	"git.eeqj.de/sneak/routewatch/internal/config"
+	"git.eeqj.de/sneak/routewatch/internal/logger"
+	"git.eeqj.de/sneak/routewatch/internal/metrics"
+	"git.eeqj.de/sneak/routewatch/internal/streamer"
+)
+
+// TestStartFailsWhenPortIsHeld holds a port with a listener of its own and
+// checks that Start on that port returns an error. Before the fix Start
+// listened in a goroutine, only logged the failure and returned nil, so the
+// daemon kept running without its web server.
+func TestStartFailsWhenPortIsHeld(t *testing.T) {
+	held, err := net.Listen("tcp", ":0")
+	if err != nil {
+		t.Fatalf("failed to open the test listener: %v", err)
+	}
+	defer func() { _ = held.Close() }()
+
+	port := held.Addr().(*net.TCPAddr).Port
+	log := logger.New()
+	s := New(nil, streamer.New(log, metrics.New()), log, &config.Config{Port: port})
+	defer func() { _ = s.Stop(context.Background()) }()
+
+	if err := s.Start(); err == nil {
+		t.Fatalf("Start returned nil while port %d was held by another listener", port)
+	}
+}