diff --git a/TODO.md b/TODO.md index d0b7920..688aa05 100644 --- a/TODO.md +++ b/TODO.md @@ -27,6 +27,11 @@ real 5 GiB limit (https://git.eeqj.de/sneak/routewatch/issues/3). # Completed Steps +- 2026-10-14: `/prefixlength/{length}` and `/prefixlength6/{length}` no + longer read every route of that length: a length with at most 500 + prefixes has all of them listed, a longer one is sampled with random + lookups on the mask length index, and each prefix is listed once, with + its newest route (closes #66) - 2026-10-03: `/api/v1/stats` serves the prefix distribution from memory, seeded at startup and adjusted on every live-route write, so a request no longer reads every live route (closes #30) diff --git a/internal/database/database.go b/internal/database/database.go index 5393f74..5701af0 100644 --- a/internal/database/database.go +++ b/internal/database/database.go @@ -8,6 +8,7 @@ import ( "encoding/json" "errors" "fmt" + "math/rand/v2" "net" "net/netip" "os" @@ -53,6 +54,13 @@ const ( sqliteHardHeapLimitBytes = 1610612736 ) +// sampleLookupsPerPrefix bounds the random lookups GetRandomPrefixesByLength +// makes when a mask length has more prefixes than it lists: at most this many +// for each prefix asked for. A lookup that lands on a prefix already picked is +// spent, so a sample can come back a little short when there are only a few +// more prefixes than asked for, or their routes are spread unevenly. +const sampleLookupsPerPrefix = 4 + // Common errors var ( // ErrInvalidIP is returned when an IP address is malformed @@ -1696,45 +1704,158 @@ 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 up to limit distinct prefixes with +// the given mask length, each with its most recently updated live route. Its +// work grows with limit, not with the number of routes: a mask length with at +// most limit prefixes has all of them listed, any other is sampled. 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 limit tells whether there are more prefixes + // to choose from than the page lists. + prefixes, err := d.firstPrefixesByLength(ctx, tableName, maskLength, limit+1) + if err != nil { + return nil, err + } + if len(prefixes) > limit { + prefixes, err = d.samplePrefixesByLength(ctx, tableName, maskLength, limit) + if err != nil { + return nil, err + } + } + + 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 up to 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. A prefix with more routes, or with more rows +// of other mask lengths just before its own, comes up more often: the page +// needs a varied sample, not an even one. +func (d *Database) samplePrefixesByLength( + ctx context.Context, tableName string, maskLength, limit int) ([]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) + 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) + } + if !picked[prefix] { + picked[prefix] = true + prefixes = append(prefixes, 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, @@ -1742,17 +1863,16 @@ 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) } diff --git a/internal/database/database_test.go b/internal/database/database_test.go index 82753a9..af408c9 100644 --- a/internal/database/database_test.go +++ b/internal/database/database_test.go @@ -45,6 +45,19 @@ 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" +) + // 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 @@ -317,3 +330,83 @@ 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) + } + } +} + +// TestGetRandomPrefixesByLengthSamplesManyPrefixes stores more /24 prefixes +// than the sample asks for, each learned from two peers in the same second, +// plus a /16, and checks that the sample lists as many prefixes as it asks +// for, each once and each a /24. +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 + for i := range sampledPrefixCount { + prefix := fmt.Sprintf("10.0.%d.0/24", i) + first := mkV4Route(t, prefix, 64500, ts) + second := mkV4Route(t, prefix, 64500, ts) + second.PeerIP = secondV4PeerIP + routes = append(routes, first, second) + } + routes = append(routes, mkV4Route(t, "10.1.0.0/16", 64501, ts)) + if err := db.UpsertLiveRouteBatch(routes); err != nil { + t.Fatalf("UpsertLiveRouteBatch: %v", err) + } + + maskLength := routes[0].MaskLength + sample, err := db.GetRandomPrefixesByLength(maskLength, ipVersionV4, sampleLimit) + if err != nil { + t.Fatalf("GetRandomPrefixesByLength: %v", err) + } + if len(sample) != sampleLimit { + t.Errorf("sample listed %d routes, want %d", len(sample), sampleLimit) + } + listed := make(map[string]bool) + for _, route := range sample { + if listed[route.Prefix] || route.MaskLength != maskLength { + t.Errorf("sample lists %s (mask %d) twice or at the wrong mask length", + route.Prefix, route.MaskLength) + } + listed[route.Prefix] = true + } +}