Author SHA1 Message Date
sneak 431d1d0481 Identify a live route by prefix and peer (closes #55)
check / check (push) Canceled after 0s
A peer has at most one route for a prefix, but the live route tables were
keyed by prefix, origin AS and peer. An announcement with a new origin AS
added a second row instead of replacing the first, and a withdrawal sent in
the same message as announcements was given their origin AS and so missed
the withdrawn route.

The tables are now unique on prefix and peer. An announcement updates the
peer's row for the prefix, origin AS included, and a withdrawal deletes it
without looking at an origin AS. The origin-only delete statements, the
deletion's origin field and DeleteLiveRoute's origin parameter are removed.

Model: opus-5-5
2026-10-14 17:05:48 +00:00
15 changed files with 316 additions and 916 deletions
+1 -4
View File
@@ -141,10 +141,7 @@ routewatch/
- **Batched Writes**: All database operations are batched for performance - **Batched Writes**: All database operations are batched for performance
- **Backpressure**: Probabilistic message dropping when queues exceed 50% capacity - **Backpressure**: Probabilistic message dropping when queues exceed 50% capacity
- **Graceful Shutdown**: 60-second timeout, flushes all pending batches - **Graceful Shutdown**: 60-second timeout, flushes all pending batches
- **Reconnection**: Any end of the stream other than shutdown, including RIS Live - **Reconnection**: Exponential backoff (5s-320s) with reset after 30s of stable connection
ending the response normally, marks the feed disconnected and reconnects with
exponential backoff (5s-320s). The backoff resets after a stream that lasted
more than 30s. A stream with no line for 60s is dropped and reconnected.
- **IP Lookup**: the most specific live route is found by looking up the - **IP Lookup**: the most specific live route is found by looking up the
address's prefix at each mask length, longest first, in the prefix index address's prefix at each mask length, longest first, in the prefix index
(at most 25 lookups for IPv4, 113 for IPv6); prefixes are stored in the (at most 25 lookups for IPv4, 113 for IPv6); prefixes are stored in the
+3 -5
View File
@@ -128,7 +128,7 @@ func TestLiveCountsTrackWritesInRealtime(t *testing.T) {
// A withdrawal removes one route. // A withdrawal removes one route.
if err := db.DeleteLiveRouteBatch([]LiveRouteDeletion{ if err := db.DeleteLiveRouteBatch([]LiveRouteDeletion{
{Prefix: "203.0.113.0/24", OriginASN: 64501, PeerIP: "192.0.2.1", IPVersion: ipVersionV4}, {Prefix: "203.0.113.0/24", PeerIP: "192.0.2.1", IPVersion: ipVersionV4},
}); err != nil { }); err != nil {
t.Fatalf("DeleteLiveRouteBatch: %v", err) t.Fatalf("DeleteLiveRouteBatch: %v", err)
} }
@@ -327,13 +327,11 @@ func announceRoutes(db *Database, batch bool, routes []*LiveRoute) error {
} }
// withdrawRoutes removes routes in one DeleteLiveRouteBatch, or with one // withdrawRoutes removes routes in one DeleteLiveRouteBatch, or with one
// DeleteLiveRoute each. It names each route by prefix and peer only, with no // DeleteLiveRoute each.
// 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 { func withdrawRoutes(db *Database, batch bool, routes []*LiveRoute) error {
if !batch { if !batch {
for _, route := range routes { for _, route := range routes {
if err := db.DeleteLiveRoute(route.Prefix, 0, route.PeerIP); err != nil { if err := db.DeleteLiveRoute(route.Prefix, route.PeerIP); err != nil {
return err return err
} }
} }
+68 -268
View File
@@ -8,7 +8,6 @@ import (
"encoding/json" "encoding/json"
"errors" "errors"
"fmt" "fmt"
"math/rand/v2"
"net" "net"
"net/netip" "net/netip"
"os" "os"
@@ -62,17 +61,6 @@ const (
sqliteHardHeapLimitBytes = 1610612736 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 // Common errors
var ( var (
// ErrInvalidIP is returned when an IP address is malformed // ErrInvalidIP is returned when an IP address is malformed
@@ -240,21 +228,22 @@ func (d *Database) beginTx() (*loggingTx, error) {
} }
// A live-route upsert is an UPDATE followed, only when no row matched, by an // A live-route upsert is an UPDATE followed, only when no row matched, by an
// INSERT. The UPDATE's rows-affected count (1 for an existing key, 0 for a new // INSERT. A live route is keyed by prefix and peer, so an announcement replaces
// one) is what lets the in-memory route counters stay exact without a COUNT(*). // the peer's route for the prefix whatever its origin AS. The UPDATE's
// Callers hold the database write lock, so no other writer can insert the same // rows-affected count (1 for an existing key, 0 for a new one) is what lets the
// key between the two statements. The id column is set only on INSERT, so an // in-memory route counters stay exact without a COUNT(*). Callers hold the
// updated route keeps its original id, exactly as the previous ON CONFLICT // database write lock, so no other writer can insert the same key between the
// upsert did. // 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 ( const (
updateLiveRouteV4SQL = `UPDATE live_routes_v4 SET mask_length = ?, as_path = ?, next_hop = ?, updateLiveRouteV4SQL = `UPDATE live_routes_v4 SET mask_length = ?, origin_asn = ?, as_path = ?,
last_updated = ? WHERE prefix = ? AND origin_asn = ? AND peer_ip = ?` next_hop = ?, last_updated = ? WHERE prefix = ? AND peer_ip = ?`
insertLiveRouteV4SQL = `INSERT INTO live_routes_v4 (id, prefix, mask_length, origin_asn, peer_ip, insertLiveRouteV4SQL = `INSERT INTO live_routes_v4 (id, prefix, mask_length, origin_asn, peer_ip,
as_path, next_hop, last_updated) VALUES (?, ?, ?, ?, ?, ?, ?, ?)` as_path, next_hop, last_updated) VALUES (?, ?, ?, ?, ?, ?, ?, ?)`
updateLiveRouteV6SQL = `UPDATE live_routes_v6 SET mask_length = ?, as_path = ?, next_hop = ?, updateLiveRouteV6SQL = `UPDATE live_routes_v6 SET mask_length = ?, origin_asn = ?, as_path = ?,
last_updated = ? WHERE prefix = ? AND origin_asn = ? AND peer_ip = ?` next_hop = ?, last_updated = ? WHERE prefix = ? AND peer_ip = ?`
insertLiveRouteV6SQL = `INSERT INTO live_routes_v6 (id, prefix, mask_length, origin_asn, peer_ip, insertLiveRouteV6SQL = `INSERT INTO live_routes_v6 (id, prefix, mask_length, origin_asn, peer_ip,
as_path, next_hop, last_updated) VALUES (?, ?, ?, ?, ?, ?, ?, ?)` as_path, next_hop, last_updated) VALUES (?, ?, ?, ?, ?, ?, ?, ?)`
@@ -273,8 +262,8 @@ const (
// live route for its prefix. // live route for its prefix.
func upsertRouteRowV4(upd, ins, has *sql.Stmt, route *LiveRoute, pathJSON string) ( func upsertRouteRowV4(upd, ins, has *sql.Stmt, route *LiveRoute, pathJSON string) (
inserted, newPrefix bool, err error) { inserted, newPrefix bool, err error) {
res, err := upd.Exec(route.MaskLength, pathJSON, route.NextHop, route.LastUpdated, res, err := upd.Exec(route.MaskLength, route.OriginASN, pathJSON, route.NextHop, route.LastUpdated,
route.Prefix, route.OriginASN, route.PeerIP) route.Prefix, route.PeerIP)
if err != nil { if err != nil {
return false, false, err return false, false, err
} }
@@ -305,8 +294,8 @@ func upsertRouteRowV4(upd, ins, has *sql.Stmt, route *LiveRoute, pathJSON string
// live route for its prefix. // live route for its prefix.
func upsertRouteRowV6(upd, ins, has *sql.Stmt, route *LiveRoute, pathJSON string) ( func upsertRouteRowV6(upd, ins, has *sql.Stmt, route *LiveRoute, pathJSON string) (
inserted, newPrefix bool, err error) { inserted, newPrefix bool, err error) {
res, err := upd.Exec(route.MaskLength, pathJSON, route.NextHop, route.LastUpdated, res, err := upd.Exec(route.MaskLength, route.OriginASN, pathJSON, route.NextHop, route.LastUpdated,
route.Prefix, route.OriginASN, route.PeerIP) route.Prefix, route.PeerIP)
if err != nil { if err != nil {
return false, false, err return false, false, err
} }
@@ -453,32 +442,18 @@ func (d *Database) DeleteLiveRouteBatch(deletions []LiveRouteDeletion) error {
} }
}() }()
// No longer need to separate deletions since we handle them in the loop below
// Prepare statements for both IPv4 and IPv6 tables // 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 = ?`) deleteV4, err := tx.Prepare(`DELETE FROM live_routes_v4 WHERE prefix = ? AND peer_ip = ?`)
if err != nil { if err != nil {
return fmt.Errorf("failed to prepare IPv4 delete with origin statement: %w", err) return fmt.Errorf("failed to prepare IPv4 delete statement: %w", err)
} }
defer func() { _ = stmtV4WithOrigin.Close() }() defer func() { _ = deleteV4.Close() }()
stmtV4WithoutOrigin, err := tx.Prepare(`DELETE FROM live_routes_v4 WHERE prefix = ? AND peer_ip = ?`) deleteV6, err := tx.Prepare(`DELETE FROM live_routes_v6 WHERE prefix = ? AND peer_ip = ?`)
if err != nil { if err != nil {
return fmt.Errorf("failed to prepare IPv4 delete without origin statement: %w", err) return fmt.Errorf("failed to prepare IPv6 delete statement: %w", err)
} }
defer func() { _ = stmtV4WithoutOrigin.Close() }() defer func() { _ = deleteV6.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) hasV4, err := tx.Prepare(prefixHasLiveRouteV4SQL)
if err != nil { if err != nil {
@@ -497,38 +472,18 @@ func (d *Database) DeleteLiveRouteBatch(deletions []LiveRouteDeletion) error {
// Mask lengths of the prefixes this batch leaves with no live route. // Mask lengths of the prefixes this batch leaves with no live route.
var gonePrefixMaskLengthsV4, gonePrefixMaskLengthsV6 []int var gonePrefixMaskLengthsV4, gonePrefixMaskLengthsV6 []int
for _, del := range deletions { for _, del := range deletions {
var stmt *sql.Stmt stmt := deleteV4
if del.IPVersion != ipVersionV4 {
// Select appropriate statement based on IP version and whether we have origin ASN stmt = deleteV6
//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 { if err != nil {
return fmt.Errorf("failed to delete route %s: %w", del.Prefix, err) return fmt.Errorf("failed to delete route %s: %w", del.Prefix, err)
} }
// A deletion with no origin ASN can remove several rows, so use the // The feed often withdraws a route that is not live, which deletes
// exact rows-affected count to keep the in-memory route counters right. // nothing, so the in-memory route counters use the rows-affected count.
affected, err := res.RowsAffected() affected, err := res.RowsAffected()
if err != nil { if err != nil {
return fmt.Errorf("failed to count deleted route %s: %w", del.Prefix, err) return fmt.Errorf("failed to count deleted route %s: %w", del.Prefix, err)
@@ -1294,9 +1249,8 @@ func (d *Database) UpsertLiveRoute(route *LiveRoute) error {
return nil return nil
} }
// DeleteLiveRoute deletes a live route // DeleteLiveRoute deletes the live route the peer has for the prefix
// If originASN is 0, deletes all routes for the prefix/peer combination func (d *Database) DeleteLiveRoute(prefix, peerIP string) error {
func (d *Database) DeleteLiveRoute(prefix string, originASN int, peerIP string) error {
d.lock("DeleteLiveRoute") d.lock("DeleteLiveRoute")
defer d.unlock() defer d.unlock()
@@ -1321,22 +1275,12 @@ func (d *Database) DeleteLiveRoute(prefix string, originASN int, peerIP string)
}() }()
// Literal per-table queries (rather than one formatted with the table name) // Literal per-table queries (rather than one formatted with the table name)
// so the delete carries no dynamically built SQL. A delete with no origin // so the delete carries no dynamically built SQL.
// ASN can remove several rows.
var res sql.Result var res sql.Result
switch { if isV4 {
case isV4 && originASN == 0:
res, err = tx.Exec(`DELETE FROM live_routes_v4 WHERE prefix = ? AND peer_ip = ?`, prefix, peerIP) res, err = tx.Exec(`DELETE FROM live_routes_v4 WHERE prefix = ? AND peer_ip = ?`, prefix, peerIP)
case isV4: } else {
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) 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 { if err != nil {
return err return err
@@ -1644,17 +1588,16 @@ func (d *Database) GetASPeersContext(ctx context.Context, asn int) ([]ASPeer, er
} }
// GetPrefixDetails returns detailed information about a prefix // GetPrefixDetails returns detailed information about a prefix
func (d *Database) GetPrefixDetails(prefix string) ([]LiveRoute, map[int]ASN, error) { func (d *Database) GetPrefixDetails(prefix string) ([]LiveRoute, error) {
return d.GetPrefixDetailsContext(context.Background(), prefix) return d.GetPrefixDetailsContext(context.Background(), prefix)
} }
// GetPrefixDetailsContext returns the live routes for a prefix and, keyed by // GetPrefixDetailsContext returns detailed information about a prefix with context support
// AS number, the handle and description of each AS that originates it. func (d *Database) GetPrefixDetailsContext(ctx context.Context, prefix string) ([]LiveRoute, error) {
func (d *Database) GetPrefixDetailsContext(ctx context.Context, prefix string) ([]LiveRoute, map[int]ASN, error) {
// Determine if it's IPv4 or IPv6 by parsing the prefix // Determine if it's IPv4 or IPv6 by parsing the prefix
_, ipnet, err := net.ParseCIDR(prefix) _, ipnet, err := net.ParseCIDR(prefix)
if err != nil { if err != nil {
return nil, nil, fmt.Errorf("invalid prefix format: %w", err) return nil, fmt.Errorf("invalid prefix format: %w", err)
} }
tableName := "live_routes_v4" tableName := "live_routes_v4"
@@ -1676,12 +1619,11 @@ func (d *Database) GetPrefixDetailsContext(ctx context.Context, prefix string) (
rows, err := d.db.QueryContext(ctx, query, prefix) rows, err := d.db.QueryContext(ctx, query, prefix)
if err != nil { if err != nil {
return nil, nil, fmt.Errorf("failed to query prefix details: %w", err) return nil, fmt.Errorf("failed to query prefix details: %w", err)
} }
defer func() { _ = rows.Close() }() defer func() { _ = rows.Close() }()
var routes []LiveRoute var routes []LiveRoute
origins := make(map[int]ASN)
for rows.Next() { for rows.Next() {
var route LiveRoute var route LiveRoute
var pathJSON string var pathJSON string
@@ -1704,18 +1646,13 @@ func (d *Database) GetPrefixDetailsContext(ctx context.Context, prefix string) (
route.IPVersion = ipVersion route.IPVersion = ipVersion
route.MaskLength, _ = ipnet.Mask.Size() route.MaskLength, _ = ipnet.Mask.Size()
routes = append(routes, route) routes = append(routes, route)
origins[route.OriginASN] = ASN{
ASN: route.OriginASN,
Handle: handle.String,
Description: description.String,
}
} }
if len(routes) == 0 { if len(routes) == 0 {
return nil, nil, fmt.Errorf("%w: %s", ErrNoRoute, prefix) return nil, fmt.Errorf("%w: %s", ErrNoRoute, prefix)
} }
return routes, origins, nil return routes, nil
} }
// GetRandomPrefixesByLength returns a random sample of prefixes with the specified mask length // GetRandomPrefixesByLength returns a random sample of prefixes with the specified mask length
@@ -1723,183 +1660,45 @@ func (d *Database) GetRandomPrefixesByLength(maskLength, ipVersion, limit int) (
return d.GetRandomPrefixesByLengthContext(context.Background(), maskLength, ipVersion, limit) return d.GetRandomPrefixesByLengthContext(context.Background(), maskLength, ipVersion, limit)
} }
// GetRandomPrefixesByLengthContext returns limit prefixes with the given mask // GetRandomPrefixesByLengthContext returns a random sample of prefixes with context support
// 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( func (d *Database) GetRandomPrefixesByLengthContext(
ctx context.Context, maskLength, ipVersion, limit int) ([]LiveRoute, error) { ctx context.Context, maskLength, ipVersion, limit int) ([]LiveRoute, error) {
// Select unique prefixes with their most recent route information
tableName := "live_routes_v4" tableName := "live_routes_v4"
if ipVersion == ipVersionV6 { if ipVersion == ipVersionV6 {
tableName = "live_routes_v6" 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 //nolint:gosec // Table name is hardcoded based on IP version
query := fmt.Sprintf(` query := fmt.Sprintf(`
SELECT DISTINCT prefix FROM %s WITH unique_prefixes AS (
WHERE mask_length = ? SELECT prefix, MAX(last_updated) as max_updated
ORDER BY prefix FROM %s
LIMIT ? WHERE mask_length = ?
`, tableName) 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)
rows, err := d.db.QueryContext(ctx, query, maskLength, n) rows, err := d.db.QueryContext(ctx, query, maskLength, limit, maskLength)
if err != nil { if err != nil {
return nil, fmt.Errorf("failed to list prefixes: %w", err) return nil, fmt.Errorf("failed to query random prefixes: %w", err)
} }
defer func() { _ = rows.Close() }() defer func() {
_ = rows.Close()
}()
var prefixes []string var routes []LiveRoute
for rows.Next() { for rows.Next() {
var prefix string var route LiveRoute
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 var pathJSON string
err := stmt.QueryRowContext(ctx, prefix).Scan( err := rows.Scan(
&route.Prefix, &route.Prefix,
&route.MaskLength, &route.MaskLength,
&route.OriginASN, &route.OriginASN,
@@ -1907,16 +1706,17 @@ func (d *Database) newestRouteOfEachPrefix(
&route.PeerIP, &route.PeerIP,
&route.LastUpdated, &route.LastUpdated,
) )
if errors.Is(err, sql.ErrNoRows) { // Set IP version based on which table we queried
route.IPVersion = ipVersion
if err != nil {
continue 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 { if err := json.Unmarshal([]byte(pathJSON), &route.ASPath); err != nil {
route.ASPath = []int{} route.ASPath = []int{}
} }
routes = append(routes, route) routes = append(routes, route)
} }
+37 -308
View File
@@ -6,7 +6,6 @@ import (
"errors" "errors"
"fmt" "fmt"
"net/netip" "net/netip"
"slices"
"sync" "sync"
"testing" "testing"
"time" "time"
@@ -14,6 +13,7 @@ import (
"git.eeqj.de/sneak/routewatch/internal/config" "git.eeqj.de/sneak/routewatch/internal/config"
"git.eeqj.de/sneak/routewatch/internal/logger" "git.eeqj.de/sneak/routewatch/internal/logger"
"github.com/google/uuid" "github.com/google/uuid"
"github.com/mattn/go-sqlite3"
) )
// tempStoreMemory is the PRAGMA temp_store value meaning "hold temp B-trees in // tempStoreMemory is the PRAGMA temp_store value meaning "hold temp B-trees in
@@ -46,32 +46,6 @@ const (
asnSecondBand = 100 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 // TestGetIPInfoFindsMostSpecificLiveRoute stores nested live prefixes for both
// families and checks that a lookup returns the most specific one covering the // 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 // address, ErrNoRoute when none covers it, and the next less specific prefix
@@ -149,8 +123,8 @@ func TestGetIPInfoFindsMostSpecificLiveRoute(t *testing.T) {
lookup("2001:db9::1", "") lookup("2001:db9::1", "")
err = db.DeleteLiveRouteBatch([]LiveRouteDeletion{ err = db.DeleteLiveRouteBatch([]LiveRouteDeletion{
{Prefix: "10.1.2.0/24", OriginASN: 64502, PeerIP: testPeerIP, IPVersion: ipVersionV4}, {Prefix: "10.1.2.0/24", PeerIP: testPeerIP, IPVersion: ipVersionV4},
{Prefix: "2001:db8:1:2::/64", OriginASN: 64502, PeerIP: testPeerIP, IPVersion: ipVersionV6}, {Prefix: "2001:db8:1:2::/64", PeerIP: testPeerIP, IPVersion: ipVersionV6},
}) })
if err != nil { if err != nil {
t.Fatalf("DeleteLiveRouteBatch: %v", err) t.Fatalf("DeleteLiveRouteBatch: %v", err)
@@ -232,6 +206,40 @@ func TestGetIPInfoIgnoresShortRoutes(t *testing.T) {
lookup("2001:db8::1", "2001:db8::/32", 64502) lookup("2001:db8::1", "2001:db8::/32", 64502)
} }
// TestLiveRouteTablesHoldOneRoutePerPrefixAndPeer checks that each live route
// table refuses a second route from a peer for a prefix, even with another
// origin AS. An upsert updates the peer's route before it would insert one, so
// only an insert straight into the table shows what the table allows.
func TestLiveRouteTablesHoldOneRoutePerPrefixAndPeer(t *testing.T) {
db, err := New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("failed to create database: %v", err)
}
defer func() { _ = db.Close() }()
const otherOriginASN = 64501
ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
for _, tt := range []struct {
route *LiveRoute
insertSQL string
}{
{route: mkV4Route(t, "198.51.100.0/24", 64500, ts), insertSQL: insertLiveRouteV4SQL},
{route: mkV6Route("2001:db8::/32", 64500, ts), insertSQL: insertLiveRouteV6SQL},
} {
if err := db.UpsertLiveRoute(tt.route); err != nil {
t.Fatalf("UpsertLiveRoute(%s): %v", tt.route.Prefix, err)
}
_, err := db.db.Exec(tt.insertSQL, uuid.NewString(), tt.route.Prefix, tt.route.MaskLength,
otherOriginASN, tt.route.PeerIP, "[]", tt.route.NextHop, ts)
var sqliteErr sqlite3.Error
if !errors.As(err, &sqliteErr) || sqliteErr.ExtendedCode != sqlite3.ErrConstraintUnique {
t.Errorf("second route from peer %s for %s: error %v, want a UNIQUE constraint error",
tt.route.PeerIP, tt.route.Prefix, err)
}
}
}
// TestConnectionPoolPragmas holds several pooled connections open at once and // TestConnectionPoolPragmas holds several pooled connections open at once and
// checks each one carries the per-connection settings from the DSN, plus the // checks each one carries the per-connection settings from the DSN, plus the
// process-wide hard heap limit, and sees the new file with auto_vacuum // process-wide hard heap limit, and sees the new file with auto_vacuum
@@ -333,7 +341,6 @@ func TestVacuumReturnsFreePages(t *testing.T) {
routes = append(routes, route) routes = append(routes, route)
deletions = append(deletions, LiveRouteDeletion{ deletions = append(deletions, LiveRouteDeletion{
Prefix: route.Prefix, Prefix: route.Prefix,
OriginASN: asn,
PeerIP: route.PeerIP, PeerIP: route.PeerIP,
IPVersion: ipVersionV4, IPVersion: ipVersionV4,
}) })
@@ -416,281 +423,3 @@ func TestBatchWriteDuringCheckpoint(t *testing.T) {
cancel() cancel()
wg.Wait() 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)
}
}
+3 -3
View File
@@ -55,7 +55,7 @@ type Store interface {
// Live route operations // Live route operations
UpsertLiveRoute(route *LiveRoute) error UpsertLiveRoute(route *LiveRoute) error
UpsertLiveRouteBatch(routes []*LiveRoute) error UpsertLiveRouteBatch(routes []*LiveRoute) error
DeleteLiveRoute(prefix string, originASN int, peerIP string) error DeleteLiveRoute(prefix, peerIP string) error
DeleteLiveRouteBatch(deletions []LiveRouteDeletion) error DeleteLiveRouteBatch(deletions []LiveRouteDeletion) error
GetPrefixDistribution() (ipv4 []PrefixDistribution, ipv6 []PrefixDistribution, err error) GetPrefixDistribution() (ipv4 []PrefixDistribution, ipv6 []PrefixDistribution, err error)
GetPrefixDistributionContext(ctx context.Context) (ipv4 []PrefixDistribution, ipv6 []PrefixDistribution, err error) GetPrefixDistributionContext(ctx context.Context) (ipv4 []PrefixDistribution, ipv6 []PrefixDistribution, err error)
@@ -76,8 +76,8 @@ type Store interface {
GetASDetailsContext(ctx context.Context, asn int) (*ASN, []LiveRoute, error) GetASDetailsContext(ctx context.Context, asn int) (*ASN, []LiveRoute, error)
GetASPeers(asn int) ([]ASPeer, error) GetASPeers(asn int) ([]ASPeer, error)
GetASPeersContext(ctx context.Context, asn int) ([]ASPeer, error) GetASPeersContext(ctx context.Context, asn int) ([]ASPeer, error)
GetPrefixDetails(prefix string) ([]LiveRoute, map[int]ASN, error) GetPrefixDetails(prefix string) ([]LiveRoute, error)
GetPrefixDetailsContext(ctx context.Context, prefix string) ([]LiveRoute, map[int]ASN, error) GetPrefixDetailsContext(ctx context.Context, prefix string) ([]LiveRoute, error)
GetRandomPrefixesByLength(maskLength, ipVersion, limit int) ([]LiveRoute, error) GetRandomPrefixesByLength(maskLength, ipVersion, limit int) ([]LiveRoute, error)
GetRandomPrefixesByLengthContext(ctx context.Context, maskLength, ipVersion, limit int) ([]LiveRoute, error) GetRandomPrefixesByLengthContext(ctx context.Context, maskLength, ipVersion, limit int) ([]LiveRoute, error)
-1
View File
@@ -114,7 +114,6 @@ type IPInfo struct {
// LiveRouteDeletion represents parameters for deleting a live route // LiveRouteDeletion represents parameters for deleting a live route
type LiveRouteDeletion struct { type LiveRouteDeletion struct {
Prefix string Prefix string
OriginASN int
PeerIP string PeerIP string
IPVersion int IPVersion int
} }
+2 -2
View File
@@ -107,7 +107,7 @@ CREATE TABLE IF NOT EXISTS live_routes_v4 (
as_path TEXT NOT NULL, -- JSON array as_path TEXT NOT NULL, -- JSON array
next_hop TEXT NOT NULL, next_hop TEXT NOT NULL,
last_updated DATETIME NOT NULL, last_updated DATETIME NOT NULL,
UNIQUE(prefix, origin_asn, peer_ip) UNIQUE(prefix, peer_ip) -- a peer has at most one route for a prefix
); );
-- IPv6 routing table maintained by PrefixHandler -- 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 as_path TEXT NOT NULL, -- JSON array
next_hop TEXT NOT NULL, next_hop TEXT NOT NULL,
last_updated DATETIME NOT NULL, last_updated DATETIME NOT NULL,
UNIQUE(prefix, origin_asn, peer_ip) UNIQUE(prefix, peer_ip) -- a peer has at most one route for a prefix
); );
-- Indexes for live_routes_v4 table -- Indexes for live_routes_v4 table
+4 -6
View File
@@ -205,7 +205,7 @@ func (m *mockStore) UpsertLiveRoute(route *database.LiveRoute) error {
} }
// DeleteLiveRoute mock implementation // DeleteLiveRoute mock implementation
func (m *mockStore) DeleteLiveRoute(prefix string, originASN int, peerIP string) error { func (m *mockStore) DeleteLiveRoute(prefix, peerIP string) error {
// Simple mock - just return nil // Simple mock - just return nil
return nil return nil
} }
@@ -252,15 +252,13 @@ func (m *mockStore) GetASDetailsContext(ctx context.Context, asn int) (*database
} }
// GetPrefixDetails mock implementation // GetPrefixDetails mock implementation
func (m *mockStore) GetPrefixDetails(prefix string) ([]database.LiveRoute, map[int]database.ASN, error) { func (m *mockStore) GetPrefixDetails(prefix string) ([]database.LiveRoute, error) {
// Return empty routes for now // Return empty routes for now
return []database.LiveRoute{}, map[int]database.ASN{}, nil return []database.LiveRoute{}, nil
} }
// GetPrefixDetailsContext mock implementation with context support // GetPrefixDetailsContext mock implementation with context support
func (m *mockStore) GetPrefixDetailsContext( func (m *mockStore) GetPrefixDetailsContext(ctx context.Context, prefix string) ([]database.LiveRoute, error) {
ctx context.Context, prefix string,
) ([]database.LiveRoute, map[int]database.ASN, error) {
return m.GetPrefixDetails(prefix) return m.GetPrefixDetails(prefix)
} }
+14 -42
View File
@@ -108,14 +108,14 @@ func (h *PrefixHandler) HandleMessage(msg *ristypes.RISMessage) {
// Withdrawals are queued before announcements, because the batch keeps // Withdrawals are queued before announcements, because the batch keeps
// each peer's last update for a prefix: a message that lists a prefix as // 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). // 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 { for _, prefix := range msg.Withdrawals {
h.batch = append(h.batch, prefixUpdate{ h.batch = append(h.batch, prefixUpdate{
prefix: canonicalPrefix(prefix), prefix: canonicalPrefix(prefix),
originASN: originASN, // Use the originASN from path if available
peer: msg.Peer, peer: msg.Peer,
messageType: "withdrawal", messageType: "withdrawal",
timestamp: timestamp, timestamp: timestamp,
path: msg.Path,
}) })
// Record withdrawal in metrics // Record withdrawal in metrics
if h.metrics != nil { if h.metrics != nil {
@@ -221,7 +221,6 @@ func (h *PrefixHandler) flushBatchLocked() {
// Create deletion record for batch delete // Create deletion record for batch delete
routesToDelete = append(routesToDelete, database.LiveRouteDeletion{ routesToDelete = append(routesToDelete, database.LiveRouteDeletion{
Prefix: update.prefix, Prefix: update.prefix,
OriginASN: update.originASN,
PeerIP: update.peer, PeerIP: update.peer,
IPVersion: ipVersion, IPVersion: ipVersion,
}) })
@@ -425,51 +424,24 @@ func (h *PrefixHandler) processAnnouncementDirect(update prefixUpdate) {
// processWithdrawalDirect handles removing a route directly without prefix table // processWithdrawalDirect handles removing a route directly without prefix table
// nolint:unused // kept for potential future use // nolint:unused // kept for potential future use
func (h *PrefixHandler) processWithdrawalDirect(update prefixUpdate) { func (h *PrefixHandler) processWithdrawalDirect(update prefixUpdate) {
// For withdrawals, we need to delete the route from live_routes if err := h.db.DeleteLiveRoute(update.prefix, update.peer); err != nil {
if update.originASN > 0 { h.logger.Error("Failed to delete live route",
if err := h.db.DeleteLiveRoute(update.prefix, update.originASN, update.peer); err != nil { "prefix", update.prefix,
h.logger.Error("Failed to delete live route", "peer", update.peer,
"prefix", update.prefix, "error", err,
"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 // processWithdrawal handles removing a route from the live routing table
// nolint:unused // kept for potential future use // nolint:unused // kept for potential future use
func (h *PrefixHandler) processWithdrawal(_ *database.Prefix, update prefixUpdate) { func (h *PrefixHandler) processWithdrawal(_ *database.Prefix, update prefixUpdate) {
// For withdrawals, we need to delete the route from live_routes if err := h.db.DeleteLiveRoute(update.prefix, update.peer); err != nil {
// Since we have the origin ASN from the update, we can delete the specific route h.logger.Error("Failed to delete live route",
if update.originASN > 0 { "prefix", update.prefix,
if err := h.db.DeleteLiveRoute(update.prefix, update.originASN, update.peer); err != nil { "peer", update.peer,
h.logger.Error("Failed to delete live route", "error", err,
"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,
)
}
} }
} }
+147 -23
View File
@@ -3,6 +3,7 @@ package routewatch
import ( import (
"context" "context"
"errors" "errors"
"fmt"
"slices" "slices"
"testing" "testing"
"time" "time"
@@ -170,7 +171,7 @@ func TestPrefixHandlerAppliesEachPeersLastUpdateInABatch(t *testing.T) {
} }
ctx := context.Background() ctx := context.Background()
routes, _, err := db.GetPrefixDetailsContext(ctx, prefix) routes, err := db.GetPrefixDetailsContext(ctx, prefix)
if err != nil && !errors.Is(err, database.ErrNoRoute) { if err != nil && !errors.Is(err, database.ErrNoRoute) {
t.Fatalf("GetPrefixDetailsContext: %v", err) t.Fatalf("GetPrefixDetailsContext: %v", err)
} }
@@ -182,32 +183,155 @@ func TestPrefixHandlerAppliesEachPeersLastUpdateInABatch(t *testing.T) {
t.Errorf("peers with a live route for %s = %v, want %v", prefix, gotPeers, tt.wantPeers) 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 var wantDistribution []database.PrefixDistribution
if len(tt.wantPeers) > 0 { if len(tt.wantPeers) > 0 {
wantDistribution = []database.PrefixDistribution{{MaskLength: prefixMaskLength, Count: 1}} wantDistribution = []database.PrefixDistribution{{MaskLength: prefixMaskLength, Count: 1}}
} }
queryDistributionV4, _, err := db.GetPrefixDistributionContext(ctx) assertCounts(t, db, ipv4Version, len(tt.wantPeers), wantDistribution)
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)
}
}) })
} }
} }
// TestPrefixHandlerKeepsOneRoutePerPrefixAndPeer runs updates from one peer
// through the prefix handler into a real database, flushing each message as its
// own batch, once with IPv4 prefixes and once with IPv6 prefixes. 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) {
families := []struct {
ipVersion int
prefix, otherPrefix string
// The mask length of both prefixes.
maskLength int
}{
{ipVersion: ipv4Version, prefix: "198.51.100.0/24", otherPrefix: "203.0.113.0/24", maskLength: 24},
{ipVersion: ipv6Version, prefix: "2001:db8:1::/48", otherPrefix: "2001:db8:2::/48", maskLength: 48},
}
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}}},
}
}
for _, family := range families {
prefix, otherPrefix := family.prefix, family.otherPrefix
// 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(fmt.Sprintf("IPv%d %s", family.ipVersion, tt.name), func(t *testing.T) {
db, err := database.New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("failed to create database: %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()
}
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: family.maskLength, Count: len(tt.wantOrigins)},
}
assertCounts(t, db, family.ipVersion, len(tt.wantOrigins), wantDistribution)
})
}
}
}
// assertCounts checks that the live route count and prefix distribution of one
// IP version are wantRoutes and wantDistribution, both as held in memory and as
// read from the route table.
func assertCounts(t *testing.T, db *database.Database, ipVersion, wantRoutes int,
wantDistribution []database.PrefixDistribution) {
t.Helper()
ctx := context.Background()
stats, err := db.GetStatsContext(ctx)
if err != nil {
t.Fatalf("GetStatsContext: %v", err)
}
tableRoutesV4, tableRoutesV6, err := db.GetLiveRouteCountsContext(ctx)
if err != nil {
t.Fatalf("GetLiveRouteCountsContext: %v", err)
}
queryDistributionV4, queryDistributionV6, err := db.GetPrefixDistributionContext(ctx)
if err != nil {
t.Fatalf("GetPrefixDistributionContext: %v", err)
}
memoryRoutes, tableRoutes := stats.IPv4Routes, tableRoutesV4
memoryDistribution, queryDistribution := stats.IPv4PrefixDistribution, queryDistributionV4
if ipVersion == ipv6Version {
memoryRoutes, tableRoutes = stats.IPv6Routes, tableRoutesV6
memoryDistribution, queryDistribution = stats.IPv6PrefixDistribution, queryDistributionV6
}
if memoryRoutes != wantRoutes || tableRoutes != wantRoutes {
t.Errorf("IPv%d live routes: %d in memory, %d in the table, want %d",
ipVersion, memoryRoutes, tableRoutes, wantRoutes)
}
if !slices.Equal(memoryDistribution, wantDistribution) ||
!slices.Equal(queryDistribution, wantDistribution) {
t.Errorf("IPv%d prefix distribution: %v in memory, %v from the table, want %v",
ipVersion, memoryDistribution, queryDistribution, wantDistribution)
}
}
+12 -4
View File
@@ -759,7 +759,7 @@ func (s *Server) handlePrefixDetailJSON() http.HandlerFunc {
} }
prefix := parsed.Masked().String() prefix := parsed.Masked().String()
routes, _, err := s.db.GetPrefixDetailsContext(r.Context(), prefix) routes, err := s.db.GetPrefixDetailsContext(r.Context(), prefix)
if err != nil { if err != nil {
if errors.Is(err, database.ErrNoRoute) { if errors.Is(err, database.ErrNoRoute) {
writeJSONError(w, http.StatusNotFound, err.Error()) writeJSONError(w, http.StatusNotFound, err.Error())
@@ -930,7 +930,7 @@ func (s *Server) handlePrefixDetail() http.HandlerFunc {
} }
prefix := parsed.Masked().String() prefix := parsed.Masked().String()
routes, originASNs, err := s.db.GetPrefixDetailsContext(r.Context(), prefix) routes, err := s.db.GetPrefixDetailsContext(r.Context(), prefix)
if err != nil { if err != nil {
if errors.Is(err, database.ErrNoRoute) { if errors.Is(err, database.ErrNoRoute) {
http.Error(w, "Prefix not found", http.StatusNotFound) http.Error(w, "Prefix not found", http.StatusNotFound)
@@ -952,10 +952,18 @@ func (s *Server) handlePrefixDetail() http.HandlerFunc {
originMap := make(map[int]*ASNInfo) originMap := make(map[int]*ASNInfo)
for _, route := range routes { for _, route := range routes {
if _, exists := originMap[route.OriginASN]; !exists { if _, exists := originMap[route.OriginASN]; !exists {
// Get AS info from database
asInfo, _, _ := s.db.GetASDetailsContext(r.Context(), route.OriginASN)
handle := ""
description := ""
if asInfo != nil {
handle = asInfo.Handle
description = asInfo.Description
}
originMap[route.OriginASN] = &ASNInfo{ originMap[route.OriginASN] = &ASNInfo{
ASN: route.OriginASN, ASN: route.OriginASN,
Handle: originASNs[route.OriginASN].Handle, Handle: handle,
Description: originASNs[route.OriginASN].Description, Description: description,
PeerCount: 0, PeerCount: 0,
} }
} }
-70
View File
@@ -3,13 +3,10 @@ package server
import ( import (
"context" "context"
"encoding/json" "encoding/json"
"errors"
"html"
"net/http" "net/http"
"net/http/httptest" "net/http/httptest"
"runtime" "runtime"
"slices" "slices"
"strings"
"testing" "testing"
"time" "time"
@@ -18,7 +15,6 @@ import (
"git.eeqj.de/sneak/routewatch/internal/logger" "git.eeqj.de/sneak/routewatch/internal/logger"
"git.eeqj.de/sneak/routewatch/internal/metrics" "git.eeqj.de/sneak/routewatch/internal/metrics"
"git.eeqj.de/sneak/routewatch/internal/streamer" "git.eeqj.de/sneak/routewatch/internal/streamer"
"git.eeqj.de/sneak/routewatch/pkg/asinfo"
"github.com/google/uuid" "github.com/google/uuid"
) )
@@ -204,72 +200,6 @@ func TestPrefixDetailHandlersParsePrefix(t *testing.T) {
} }
} }
// failingASDetailsDB wraps a real store and fails every AS details lookup.
type failingASDetailsDB struct {
database.Store
}
func (failingASDetailsDB) GetASDetailsContext(_ context.Context, _ int) (*database.ASN, []database.LiveRoute, error) {
return nil, nil, errors.New("AS details lookup failed")
}
// TestPrefixDetailShowsOriginNames checks that the prefix page shows each
// origin AS's handle and description without looking up the AS's details,
// which also read every route the AS originates. Before the fix the page did
// that lookup once per origin, so with it failing the names were missing
// (https://git.eeqj.de/sneak/routewatch/issues/67). The AS path column shows
// handles too, so the test looks for the origin list's "(handle)" form.
func TestPrefixDetailShowsOriginNames(t *testing.T) {
db, err := database.New(&config.Config{StateDir: t.TempDir()}, logger.New())
if err != nil {
t.Fatalf("database.New: %v", err)
}
t.Cleanup(func() { _ = db.Close() })
ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
origins := []int{15169, 13335}
for _, asn := range origins {
if _, err := db.GetOrCreateASN(asn, ts); err != nil {
t.Fatalf("GetOrCreateASN(%d): %v", asn, err)
}
}
if err := db.UpsertLiveRouteBatch([]*database.LiveRoute{
{
ID: uuid.New(), Prefix: "192.0.2.0/24", MaskLength: 24, IPVersion: 4,
OriginASN: origins[0], PeerIP: "198.51.100.1", ASPath: []int{origins[0]}, NextHop: "198.51.100.1",
LastUpdated: ts,
},
{
ID: uuid.New(), Prefix: "192.0.2.0/24", MaskLength: 24, IPVersion: 4,
OriginASN: origins[1], PeerIP: "198.51.100.2", ASPath: []int{origins[1]}, NextHop: "198.51.100.2",
LastUpdated: ts,
},
}); err != nil {
t.Fatalf("UpsertLiveRouteBatch: %v", err)
}
s := New(failingASDetailsDB{db}, streamer.New(logger.New(), metrics.New()), logger.New(), &config.Config{})
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, "/prefix/192.0.2.0/24", nil))
if rec.Code != http.StatusOK {
t.Fatalf("status %d, want %d; body %s", rec.Code, http.StatusOK, rec.Body)
}
page := rec.Body.String()
for _, asn := range origins {
info, ok := asinfo.Get(asn)
if !ok {
t.Fatalf("AS%d is not in the AS info data", asn)
}
if want := "(" + html.EscapeString(info.Handle) + ")"; !strings.Contains(page, want) {
t.Errorf("page does not show AS%d's handle %q", asn, want)
}
if want := html.EscapeString(info.Description); !strings.Contains(page, want) {
t.Errorf("page does not show AS%d's description %q", asn, want)
}
}
}
// settledGoroutineCount lets transient goroutines finish, then reports the // settledGoroutineCount lets transient goroutines finish, then reports the
// current count. // current count.
func settledGoroutineCount() int { func settledGoroutineCount() int {
+15 -37
View File
@@ -7,7 +7,6 @@ import (
"compress/gzip" "compress/gzip"
"context" "context"
"encoding/json" "encoding/json"
"errors"
"fmt" "fmt"
"io" "io"
"math" "math"
@@ -49,7 +48,6 @@ const (
metricsUpdateRate = time.Second metricsUpdateRate = time.Second
minBackoffDelay = 5 * time.Second minBackoffDelay = 5 * time.Second
maxBackoffDelay = 320 * time.Second maxBackoffDelay = 320 * time.Second
streamIdleTimeout = 60 * time.Second // Reconnect when no line arrives for this long
metricsLogInterval = 10 * time.Second metricsLogInterval = 10 * time.Second
bytesPerKB = 1024 bytesPerKB = 1024
bytesPerMB = 1024 * 1024 bytesPerMB = 1024 * 1024
@@ -60,9 +58,6 @@ const (
backpressureSlope = 2.0 // Slope for linear drop probability increase backpressureSlope = 2.0 // Slope for linear drop probability increase
) )
// errStreamEnded is returned by stream when RIS Live ends the response.
var errStreamEnded = errors.New("RIS Live ended the stream")
// MessageHandler defines the interface for processing RIS messages. // MessageHandler defines the interface for processing RIS messages.
// Implementations must specify which message types they want to receive, // Implementations must specify which message types they want to receive,
// how to process messages, and their desired queue capacity. // how to process messages, and their desired queue capacity.
@@ -112,8 +107,6 @@ type Streamer struct {
logger *logger.Logger logger *logger.Logger
client *http.Client client *http.Client
url string url string
idleTimeout time.Duration // Drop the connection when no line arrives for this long
minBackoff time.Duration // First wait before reconnecting; doubles up to maxBackoffDelay
handlers []*handlerInfo handlers []*handlerInfo
rawHandler RawMessageHandler rawHandler RawMessageHandler
mu sync.RWMutex mu sync.RWMutex
@@ -131,10 +124,8 @@ type Streamer struct {
// The metrics tracker is used to record message counts, bytes received, and connection status. // The metrics tracker is used to record message counts, bytes received, and connection status.
func New(logger *logger.Logger, metrics *metrics.Tracker) *Streamer { func New(logger *logger.Logger, metrics *metrics.Tracker) *Streamer {
return &Streamer{ return &Streamer{
logger: logger, logger: logger,
url: risLiveURL, url: risLiveURL,
idleTimeout: streamIdleTimeout,
minBackoff: minBackoffDelay,
client: &http.Client{ client: &http.Client{
Timeout: 0, // No timeout for streaming Timeout: 0, // No timeout for streaming
Transport: &http.Transport{ Transport: &http.Transport{
@@ -414,7 +405,7 @@ func (s *Streamer) updateMetrics(messageBytes int) {
// streamWithReconnect handles streaming with automatic reconnection and exponential backoff // streamWithReconnect handles streaming with automatic reconnection and exponential backoff
func (s *Streamer) streamWithReconnect(ctx context.Context) { func (s *Streamer) streamWithReconnect(ctx context.Context) {
backoffDelay := s.minBackoff backoffDelay := minBackoffDelay
consecutiveFailures := 0 consecutiveFailures := 0
for { for {
@@ -431,9 +422,8 @@ func (s *Streamer) streamWithReconnect(ctx context.Context) {
err := s.stream(ctx) err := s.stream(ctx)
streamDuration := time.Since(startTime) streamDuration := time.Since(startTime)
// Only Stop ends the streamer. A stream that ends any other way, if err == nil {
// with or without an error, is a disconnect and is reconnected. // Clean exit (context cancelled)
if ctx.Err() != nil {
return return
} }
@@ -444,13 +434,17 @@ func (s *Streamer) streamWithReconnect(ctx context.Context) {
"stream_duration", streamDuration) "stream_duration", streamDuration)
s.metrics.SetConnected(false) s.metrics.SetConnected(false)
// A stream that lasted more than 30 seconds resets the backoff. That // Check if context is cancelled
// includes a stream dropped after streamIdleTimeout with no line, so if ctx.Err() != nil {
// a silent feed is retried after the first backoff delay every time. return
}
// If we streamed for more than 30 seconds, reset the backoff
// This indicates we had a successful connection that received data
if streamDuration > 30*time.Second { if streamDuration > 30*time.Second {
s.logger.Info("Resetting backoff delay due to successful connection", s.logger.Info("Resetting backoff delay due to successful connection",
"stream_duration", streamDuration) "stream_duration", streamDuration)
backoffDelay = s.minBackoff backoffDelay = minBackoffDelay
consecutiveFailures = 0 consecutiveFailures = 0
} else { } else {
// Increment consecutive failures // Increment consecutive failures
@@ -483,13 +477,7 @@ func (s *Streamer) stream(ctx context.Context) error {
connCtx, connCancel := context.WithCancel(ctx) connCtx, connCancel := context.WithCancel(ctx)
defer connCancel() defer connCancel()
// The idle timer drops the connection, by cancelling connCtx, when no req, err := http.NewRequestWithContext(ctx, "GET", s.url, nil)
// line arrives for s.idleTimeout. Each line read starts it over. It
// also covers the wait for the server to answer.
idleTimer := time.AfterFunc(s.idleTimeout, connCancel)
defer idleTimer.Stop()
req, err := http.NewRequestWithContext(connCtx, "GET", s.url, nil)
if err != nil { if err != nil {
return fmt.Errorf("failed to create request: %w", err) return fmt.Errorf("failed to create request: %w", err)
} }
@@ -571,8 +559,6 @@ func (s *Streamer) stream(ctx context.Context) error {
scanner := bufio.NewScanner(reader) scanner := bufio.NewScanner(reader)
for scanner.Scan() { for scanner.Scan() {
idleTimer.Reset(s.idleTimeout)
select { select {
case <-ctx.Done(): case <-ctx.Done():
s.logger.Info("Stream stopped by context") s.logger.Info("Stream stopped by context")
@@ -721,19 +707,11 @@ func (s *Streamer) stream(ctx context.Context) error {
s.mu.RUnlock() s.mu.RUnlock()
} }
if ctx.Err() != nil {
return ctx.Err()
}
if connCtx.Err() != nil {
return fmt.Errorf("no line received for %s", s.idleTimeout)
}
if err := scanner.Err(); err != nil { if err := scanner.Err(); err != nil {
return fmt.Errorf("scanner error: %w", err) return fmt.Errorf("scanner error: %w", err)
} }
return errStreamEnded return nil
} }
// shouldDropForBackpressure determines if a message should be dropped based on queue utilization // shouldDropForBackpressure determines if a message should be dropped based on queue utilization
+10 -142
View File
@@ -7,7 +7,6 @@ import (
"net/http" "net/http"
"net/http/httptest" "net/http/httptest"
"runtime" "runtime"
"sync/atomic"
"testing" "testing"
"time" "time"
@@ -40,14 +39,6 @@ func TestNewStreamer(t *testing.T) {
if s.metrics != metricsTracker { if s.metrics != metricsTracker {
t.Error("metrics tracker not set correctly") t.Error("metrics tracker not set correctly")
} }
if s.idleTimeout != streamIdleTimeout {
t.Errorf("idleTimeout = %s, want %s", s.idleTimeout, streamIdleTimeout)
}
if s.minBackoff != minBackoffDelay {
t.Errorf("minBackoff = %s, want %s", s.minBackoff, minBackoffDelay)
}
} }
// TestStreamDoesNotLeakTickersAcrossReconnects drives many short-lived // TestStreamDoesNotLeakTickersAcrossReconnects drives many short-lived
@@ -66,8 +57,8 @@ func TestStreamDoesNotLeakTickersAcrossReconnects(t *testing.T) {
// One warm-up connection so any persistent HTTP transport goroutine exists // One warm-up connection so any persistent HTTP transport goroutine exists
// before we take the baseline. // before we take the baseline.
if err := s.stream(context.Background()); !errors.Is(err, errStreamEnded) { if err := s.stream(context.Background()); err != nil {
t.Fatalf("warm-up stream returned %v, want %v", err, errStreamEnded) t.Fatalf("warm-up stream returned error: %v", err)
} }
s.client.CloseIdleConnections() s.client.CloseIdleConnections()
@@ -75,8 +66,8 @@ func TestStreamDoesNotLeakTickersAcrossReconnects(t *testing.T) {
const reconnects = 20 const reconnects = 20
for range reconnects { for range reconnects {
if err := s.stream(context.Background()); !errors.Is(err, errStreamEnded) { if err := s.stream(context.Background()); err != nil {
t.Fatalf("stream returned %v, want %v", err, errStreamEnded) t.Fatalf("stream returned error: %v", err)
} }
} }
s.client.CloseIdleConnections() s.client.CloseIdleConnections()
@@ -87,35 +78,23 @@ func TestStreamDoesNotLeakTickersAcrossReconnects(t *testing.T) {
} }
} }
// updateLine is one UPDATE message as RIS Live sends it. // updateHandler wants UPDATE messages and does nothing with them.
const updateLine = `{"type":"ris_message","data":{"type":"UPDATE","peer":"192.0.2.1",` + type updateHandler struct{}
`"peer_asn":"64496","timestamp":1700000000}}` + "\n"
// updateHandler wants UPDATE messages. For each one it handles it sends on
// handled if there is room, and otherwise does nothing; a nil handled is
// never sent on.
type updateHandler struct {
handled chan struct{}
}
func (updateHandler) WantsMessage(messageType string) bool { return messageType == "UPDATE" } func (updateHandler) WantsMessage(messageType string) bool { return messageType == "UPDATE" }
func (updateHandler) HandleMessage(*ristypes.RISMessage) {}
func (updateHandler) QueueCapacity() int { return 10 } func (updateHandler) QueueCapacity() int { return 10 }
func (h updateHandler) HandleMessage(*ristypes.RISMessage) {
select {
case h.handled <- struct{}{}:
default:
}
}
// TestStopBeforeMessageReachesQueues stops the streamer after the read loop // TestStopBeforeMessageReachesQueues stops the streamer after the read loop
// has checked for cancellation but before it hands the message to the handler // has checked for cancellation but before it hands the message to the handler
// queues. That is the gap a stop from another goroutine can land in, and it // queues. That is the gap a stop from another goroutine can land in, and it
// used to end in "send on closed channel". The raw handler runs in that gap on // used to end in "send on closed channel". The raw handler runs in that gap on
// the read loop itself, so calling Stop from it hits the gap every time. // the read loop itself, so calling Stop from it hits the gap every time.
func TestStopBeforeMessageReachesQueues(t *testing.T) { func TestStopBeforeMessageReachesQueues(t *testing.T) {
const line = `{"type":"ris_message","data":{"type":"UPDATE","peer":"192.0.2.1",` +
`"peer_asn":"64496","timestamp":1700000000}}` + "\n"
srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) { srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
_, _ = io.WriteString(w, updateLine) _, _ = io.WriteString(w, line)
})) }))
defer srv.Close() defer srv.Close()
@@ -138,117 +117,6 @@ func TestStopBeforeMessageReachesQueues(t *testing.T) {
s.Stop() s.Stop()
} }
// TestReconnectWhenStreamEnds has the server end the first response
// normally, as when RIS Live restarts. The streamer must treat that as a
// disconnect and connect again.
func TestReconnectWhenStreamEnds(t *testing.T) {
endResponse := func(http.ResponseWriter, *http.Request) {}
checkReconnect(t, endResponse, streamIdleTimeout)
}
// TestReconnectWhenStreamGoesIdle has the server hold the first response
// open without sending anything. The streamer must drop it once the idle
// timeout passes and connect again.
func TestReconnectWhenStreamGoesIdle(t *testing.T) {
const idleTimeout = 50 * time.Millisecond
holdResponse := func(w http.ResponseWriter, r *http.Request) {
w.(http.Flusher).Flush()
<-r.Context().Done()
}
checkReconnect(t, holdResponse, idleTimeout)
}
// TestReconnectWhenServerNeverAnswers has the server accept the first
// request and never answer it, not even with headers. The streamer must give
// up on it once the idle timeout passes and connect again.
func TestReconnectWhenServerNeverAnswers(t *testing.T) {
const idleTimeout = 50 * time.Millisecond
neverAnswer := func(_ http.ResponseWriter, r *http.Request) {
<-r.Context().Done()
}
checkReconnect(t, neverAnswer, idleTimeout)
}
// TestStreamNotDroppedWhileLinesArrive has the server send lines well within
// the idle timeout for several times the timeout, then end the response. Each
// line must start the idle timeout over, so the stream lasts until the server
// ends it instead of being dropped as idle.
func TestStreamNotDroppedWhileLinesArrive(t *testing.T) {
const (
idleTimeout = 100 * time.Millisecond
lineInterval = idleTimeout / 10
sendFor = 5 * idleTimeout
)
srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
deadline := time.Now().Add(sendFor)
for time.Now().Before(deadline) {
_, _ = io.WriteString(w, updateLine)
w.(http.Flusher).Flush()
time.Sleep(lineInterval)
}
}))
defer srv.Close()
s := New(logger.New(), metrics.New())
s.url = srv.URL
s.idleTimeout = idleTimeout
if err := s.stream(context.Background()); !errors.Is(err, errStreamEnded) {
t.Fatalf("stream returned %v, want %v", err, errStreamEnded)
}
}
// checkReconnect runs a streamer against a fake RIS Live server. The first
// response is served by firstResponse; every later one carries one UPDATE
// line and then stays open. The streamer must come back for a second
// response, must not be marked connected when it does, and must handle the
// UPDATE in it.
func checkReconnect(t *testing.T, firstResponse http.HandlerFunc, idleTimeout time.Duration) {
t.Helper()
const maxWait = 5 * time.Second
tracker := metrics.New()
connectedAtReconnect := make(chan bool, 1)
var requests atomic.Int32
srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if requests.Add(1) == 1 {
firstResponse(w, r)
return
}
select {
case connectedAtReconnect <- tracker.IsConnected():
default:
}
_, _ = io.WriteString(w, updateLine)
w.(http.Flusher).Flush()
<-r.Context().Done()
}))
defer srv.Close()
s := New(logger.New(), tracker)
s.url = srv.URL
s.idleTimeout = idleTimeout
s.minBackoff = time.Millisecond
handler := updateHandler{handled: make(chan struct{}, 1)}
s.RegisterHandler(handler)
if err := s.Start(); err != nil {
t.Fatalf("Start returned error: %v", err)
}
defer s.Stop()
select {
case <-handler.handled:
case <-time.After(maxWait):
t.Fatalf("no message from a second response handled within %s", maxWait)
}
if <-connectedAtReconnect {
t.Error("streamer was still marked connected when it came back for a second response")
}
}
// settledGoroutineCount lets transient goroutines finish, then reports the // settledGoroutineCount lets transient goroutines finish, then reports the
// current count. // current count.
func settledGoroutineCount() int { func settledGoroutineCount() int {
-1
View File
@@ -283,7 +283,6 @@
<div class="origin-item"> <div class="origin-item">
<a href="/as/{{.ASN}}" class="as-link">AS{{.ASN}}</a> <a href="/as/{{.ASN}}" class="as-link">AS{{.ASN}}</a>
{{if .Handle}} ({{.Handle}}){{end}} {{if .Handle}} ({{.Handle}}){{end}}
{{if .Description}} {{.Description}}{{end}}
<span style="color: #7f8c8d; margin-left: 10px;">{{.PeerCount}} peer{{if ne .PeerCount 1}}s{{end}}</span> <span style="color: #7f8c8d; margin-left: 10px;">{{.PeerCount}} peer{{if ne .PeerCount 1}}s{{end}}</span>
</div> </div>
{{end}} {{end}}