1 Commits
Author SHA1 Message Date
sneak 515bc1e5a2 Exit with status 1 when the web server cannot listen (closes #73)
check / check (push) Waiting to run
Server.Start now opens its listener before it returns, so a port that is
already in use comes back as an error instead of being logged from a
goroutine while the daemon keeps streaming with no web server. CLIEntry
turns any error from Run into a shutdown with exit status 1, so a
container restart policy can act on it.

Model: opus-5-5
2026-10-14 14:43:58 +00:00
12 changed files with 88 additions and 608 deletions
+3 -7
View File
@@ -141,15 +141,11 @@ 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 33 lookups for IPv4, 129 for IPv6); prefixes are stored in the
text form Go's `net/netip` prints. Routes shorter than /8 for IPv4 or /16 text form Go's `net/netip` prints
for IPv6, such as a default route, are not used.
### Database Schema ### Database Schema
+13 -31
View File
@@ -37,14 +37,6 @@ const (
ipv6Bits = 128 ipv6Bits = 128
) )
// The IP lookup does not use routes shorter than these mask lengths. A default
// route or a route this short only says where a peer sends traffic it has no
// better route for; it does not say who holds the address.
const (
minLookupMaskLengthV4 = 8
minLookupMaskLengthV6 = 16
)
// SQLite memory tuning. cache_size and busy_timeout go in the DSN so every // SQLite memory tuning. cache_size and busy_timeout go in the DSN so every
// pooled connection gets them; the heap limits are process-wide and set once. // pooled connection gets them; the heap limits are process-wide and set once.
const ( const (
@@ -1632,17 +1624,16 @@ 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"
@@ -1664,12 +1655,11 @@ 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
@@ -1692,18 +1682,13 @@ 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
@@ -1891,12 +1876,11 @@ func (d *Database) GetIPInfo(ip string) (*IPInfo, error) {
// GetIPInfoContext returns comprehensive IP information with context support. // GetIPInfoContext returns comprehensive IP information with context support.
// //
// It finds the most specific live route for the address with point lookups on // It finds the most specific live route for the address with point lookups on
// the prefix index: for each mask length from the longest (32 or 128) down to // the prefix index: for each mask length from the longest (32 or 128) down to 0
// minLookupMaskLengthV4 or minLookupMaskLengthV6 it looks up the address's // it looks up the address's network prefix at that length, and the first length
// network prefix at that length, and the first length with a live route wins. // with a live route wins. That is at most 33 or 129 indexed lookups, however
// That is at most 25 or 113 indexed lookups, however many routes are live. The // many routes are live. The prefix is looked up in the text form net/netip
// prefix is looked up in the text form net/netip prints, which is the form the // prints, which is the form the prefix handler stores.
// prefix handler stores.
func (d *Database) GetIPInfoContext(ctx context.Context, ip string) (*IPInfo, error) { func (d *Database) GetIPInfoContext(ctx context.Context, ip string) (*IPInfo, error) {
addr, err := netip.ParseAddr(ip) addr, err := netip.ParseAddr(ip)
if err != nil { if err != nil {
@@ -1907,11 +1891,9 @@ func (d *Database) GetIPInfoContext(ctx context.Context, ip string) (*IPInfo, er
info := &IPInfo{IP: ip, IPVersion: ipVersionV6} info := &IPInfo{IP: ip, IPVersion: ipVersionV6}
routesTable, prefixesTable := "live_routes_v6", "prefixes_v6" routesTable, prefixesTable := "live_routes_v6", "prefixes_v6"
minMaskLength := minLookupMaskLengthV6
if addr.Is4() { if addr.Is4() {
info.IPVersion = ipVersionV4 info.IPVersion = ipVersionV4
routesTable, prefixesTable = "live_routes_v4", "prefixes_v4" routesTable, prefixesTable = "live_routes_v4", "prefixes_v4"
minMaskLength = minLookupMaskLengthV4
} }
// Route info with peer count and prefix first_seen // Route info with peer count and prefix first_seen
@@ -1952,7 +1934,7 @@ func (d *Database) GetIPInfoContext(ctx context.Context, ip string) (*IPInfo, er
var whoisUpdatedAt sql.NullTime var whoisUpdatedAt sql.NullTime
found := false found := false
for bits := addr.BitLen(); bits >= minMaskLength; bits-- { for bits := addr.BitLen(); bits >= 0; bits-- {
prefix := netip.PrefixFrom(addr, bits).Masked().String() prefix := netip.PrefixFrom(addr, bits).Masked().String()
err := stmt.QueryRowContext(ctx, prefix).Scan( err := stmt.QueryRowContext(ctx, prefix).Scan(
&info.Netblock, &info.Netblock,
-72
View File
@@ -133,78 +133,6 @@ func TestGetIPInfoFindsMostSpecificLiveRoute(t *testing.T) {
lookup("2001:db8:1:2::3", "2001:db8:1::/48") lookup("2001:db8:1:2::3", "2001:db8:1::/48")
} }
// TestGetIPInfoIgnoresShortRoutes checks that the IP lookup does not report a
// default route, or a route shorter than /8 for IPv4 or /16 for IPv6, as the
// route for an address, and that a route of exactly /8 or /16, or longer, still
// is. The prefixes are written out so that a wrong limit fails the test.
func TestGetIPInfoIgnoresShortRoutes(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)
announce := func(prefix string, asn int) {
t.Helper()
err := db.UpsertLiveRouteBatch([]*LiveRoute{{
ID: uuid.New(),
Prefix: prefix,
MaskLength: netip.MustParsePrefix(prefix).Bits(),
IPVersion: detectIPVersion(prefix),
OriginASN: asn,
PeerIP: testPeerIP,
ASPath: []int{asn},
NextHop: testPeerIP,
LastUpdated: ts,
}})
if err != nil {
t.Fatalf("UpsertLiveRouteBatch(%s): %v", prefix, err)
}
}
// lookup checks that ip resolves to the live prefix netblock from asn.
lookup := func(ip, netblock string, asn int) {
t.Helper()
info, err := db.GetIPInfo(ip)
if err != nil {
t.Errorf("GetIPInfo(%s): %v", ip, err)
return
}
if info.Netblock != netblock || info.ASN != asn {
t.Errorf("GetIPInfo(%s) = %s AS%d, want %s AS%d", ip, info.Netblock, info.ASN, netblock, asn)
}
}
// Default and very short routes, and routes one bit shorter than each limit.
announce("2000::/3", 64500)
announce("0.0.0.0/0", 64500)
announce("2000::/15", 64500)
announce("192.0.0.0/7", 64500)
for _, ip := range []string{"2001:db8::1", "192.0.2.1"} {
if info, err := db.GetIPInfo(ip); !errors.Is(err, ErrNoRoute) {
t.Errorf("GetIPInfo(%s) with only short routes = %+v, %v; want ErrNoRoute", ip, info, err)
}
}
// Routes of exactly the minimum length are used.
announce("2001::/16", 64501)
announce("192.0.0.0/8", 64501)
lookup("2001:db8::1", "2001::/16", 64501)
lookup("192.0.2.1", "192.0.0.0/8", 64501)
announce("2001:db8::/32", 64502)
lookup("2001:db8::1", "2001:db8::/32", 64502)
}
// 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
+2 -2
View File
@@ -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)
+3 -5
View File
@@ -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)
} }
+26 -27
View File
@@ -105,24 +105,7 @@ func (h *PrefixHandler) HandleMessage(msg *ristypes.RISMessage) {
h.mu.Lock() h.mu.Lock()
defer h.mu.Unlock() defer h.mu.Unlock()
// Withdrawals are queued before announcements, because the batch keeps // Process announcements
// 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).
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 {
h.metrics.RecordWithdrawal()
}
}
for _, announcement := range msg.Announcements { for _, announcement := range msg.Announcements {
for _, prefix := range announcement.Prefixes { for _, prefix := range announcement.Prefixes {
h.batch = append(h.batch, prefixUpdate{ h.batch = append(h.batch, prefixUpdate{
@@ -140,6 +123,22 @@ func (h *PrefixHandler) HandleMessage(msg *ristypes.RISMessage) {
} }
} }
// Process withdrawals
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 {
h.metrics.RecordWithdrawal()
}
}
// Check if we need to flush // Check if we need to flush
if len(h.batch) >= prefixBatchSize { if len(h.batch) >= prefixBatchSize {
h.flushBatchLocked() h.flushBatchLocked()
@@ -180,14 +179,14 @@ func (h *PrefixHandler) flushBatchLocked() {
startTime := time.Now() startTime := time.Now()
batchSize := len(h.batch) batchSize := len(h.batch)
// For each prefix and peer, keep only the last update the feed delivered. // Group updates by prefix to deduplicate
// Updates from different peers for the same prefix are all kept. Arrival // For each prefix, keep the latest update
// order decides, not the timestamp: timestamps are whole seconds, so prefixMap := make(map[string]prefixUpdate)
// updates within one second carry the same one.
type prefixAndPeer struct{ prefix, peer string }
lastUpdates := make(map[prefixAndPeer]prefixUpdate)
for _, update := range h.batch { for _, update := range h.batch {
lastUpdates[prefixAndPeer{update.prefix, update.peer}] = update key := update.prefix
if existing, ok := prefixMap[key]; !ok || update.timestamp.After(existing.timestamp) {
prefixMap[key] = update
}
} }
// Collect routes to upsert and delete // Collect routes to upsert and delete
@@ -197,7 +196,7 @@ func (h *PrefixHandler) flushBatchLocked() {
// Collect unique prefixes to update // Collect unique prefixes to update
prefixesToUpdate := make(map[string]time.Time) prefixesToUpdate := make(map[string]time.Time)
for _, update := range lastUpdates { for _, update := range prefixMap {
// Track prefix for both announcements and withdrawals // Track prefix for both announcements and withdrawals
if _, exists := prefixesToUpdate[update.prefix]; !exists || update.timestamp.After(prefixesToUpdate[update.prefix]) { if _, exists := prefixesToUpdate[update.prefix]; !exists || update.timestamp.After(prefixesToUpdate[update.prefix]) {
prefixesToUpdate[update.prefix] = update.timestamp prefixesToUpdate[update.prefix] = update.timestamp
@@ -256,7 +255,7 @@ func (h *PrefixHandler) flushBatchLocked() {
elapsed := time.Since(startTime) elapsed := time.Since(startTime)
h.logger.Debug("Flushed prefix batch", h.logger.Debug("Flushed prefix batch",
"batch_size", batchSize, "batch_size", batchSize,
"unique_prefixes", len(prefixesToUpdate), "unique_prefixes", len(prefixMap),
"success", successCount, "success", successCount,
"duration_ms", elapsed.Milliseconds(), "duration_ms", elapsed.Milliseconds(),
) )
-139
View File
@@ -1,9 +1,7 @@
package routewatch package routewatch
import ( import (
"context"
"errors" "errors"
"slices"
"testing" "testing"
"time" "time"
@@ -74,140 +72,3 @@ func TestPrefixHandlerStoresPrefixesTheIPLookupFinds(t *testing.T) {
} }
} }
} }
// TestPrefixHandlerAppliesEachPeersLastUpdateInABatch runs updates for one
// prefix through the prefix handler into a real database. All updates carry the
// same timestamp, as updates within one second of the feed do. Within a batch,
// the last update each peer sent for the prefix decides whether that peer has a
// live route for it, and updates from different peers are all applied. The
// in-memory route count and prefix distribution must match the route table.
func TestPrefixHandlerAppliesEachPeersLastUpdateInABatch(t *testing.T) {
const (
prefix = "198.51.100.0/24"
prefixMaskLength = 24
otherPeerIP = "2001:db8:ffff::2"
)
ts := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
announce := func(peer string) *ristypes.RISMessage {
return &ristypes.RISMessage{
Peer: peer,
Path: ristypes.ASPath{testASNA, testASNB},
ParsedTimestamp: ts,
Announcements: []ristypes.RISAnnouncement{{NextHop: peer, Prefixes: []string{prefix}}},
}
}
withdraw := func(peer string) *ristypes.RISMessage {
return &ristypes.RISMessage{Peer: peer, ParsedTimestamp: ts, Withdrawals: []string{prefix}}
}
announceAndWithdraw := func(peer string) *ristypes.RISMessage {
msg := announce(peer)
msg.Withdrawals = []string{prefix}
return msg
}
tests := []struct {
name string
// Each inner list is handled and then flushed as one batch.
batches [][]*ristypes.RISMessage
// The peers left with a live route for the prefix, by peer IP.
wantPeers []string
}{
{
name: "two peers announce the prefix",
batches: [][]*ristypes.RISMessage{{announce(testPeerIP), announce(otherPeerIP)}},
wantPeers: []string{testPeerIP, otherPeerIP},
},
{
name: "two peers withdraw the prefix",
batches: [][]*ristypes.RISMessage{
{announce(testPeerIP)},
{announce(otherPeerIP)},
{withdraw(testPeerIP), withdraw(otherPeerIP)},
},
},
{
name: "a peer announces and then withdraws the prefix",
batches: [][]*ristypes.RISMessage{{announce(testPeerIP), withdraw(testPeerIP)}},
},
{
name: "a peer withdraws and then re-announces the prefix",
batches: [][]*ristypes.RISMessage{
{announce(testPeerIP)},
{withdraw(testPeerIP), announce(testPeerIP)},
},
wantPeers: []string{testPeerIP},
},
{
// RFC 4271 section 4.3: the announcement counts, not the withdrawal.
name: "one message announces and withdraws the prefix",
batches: [][]*ristypes.RISMessage{
{announce(testPeerIP)},
{announceAndWithdraw(testPeerIP)},
},
wantPeers: []string{testPeerIP},
},
}
for _, tt := range tests {
t.Run(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 _, batch := range tt.batches {
for _, msg := range batch {
h.HandleMessage(msg)
}
h.mu.Lock()
h.flushBatchLocked()
h.mu.Unlock()
}
ctx := context.Background()
routes, _, err := db.GetPrefixDetailsContext(ctx, prefix)
if err != nil && !errors.Is(err, database.ErrNoRoute) {
t.Fatalf("GetPrefixDetailsContext: %v", err)
}
var gotPeers []string
for _, route := range routes {
gotPeers = append(gotPeers, route.PeerIP)
}
if !slices.Equal(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
if len(tt.wantPeers) > 0 {
wantDistribution = []database.PrefixDistribution{{MaskLength: prefixMaskLength, Count: 1}}
}
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)
}
})
}
}
+16 -23
View File
@@ -7,7 +7,6 @@ import (
"errors" "errors"
"net" "net"
"net/http" "net/http"
"net/netip"
"runtime" "runtime"
"sort" "sort"
"strconv" "strconv"
@@ -749,17 +748,10 @@ func (s *Server) handlePrefixDetailJSON() http.HandlerFunc {
return return
} }
// Look the prefix up in the text form net/netip prints, the form the // Combine prefix and length into CIDR notation
// tables store, so that any valid way of writing it finds the route. prefix := prefixParam + "/" + lenParam
parsed, err := netip.ParsePrefix(prefixParam + "/" + lenParam)
if err != nil {
writeJSONError(w, http.StatusBadRequest, err.Error())
return routes, err := s.db.GetPrefixDetailsContext(r.Context(), prefix)
}
prefix := parsed.Masked().String()
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())
@@ -920,17 +912,10 @@ func (s *Server) handlePrefixDetail() http.HandlerFunc {
return return
} }
// Look the prefix up in the text form net/netip prints, the form the // Combine prefix and length into CIDR notation
// tables store, so that any valid way of writing it finds the route. prefix := prefixParam + "/" + lenParam
parsed, err := netip.ParsePrefix(prefixParam + "/" + lenParam)
if err != nil {
http.Error(w, "Invalid prefix", http.StatusBadRequest)
return routes, err := s.db.GetPrefixDetailsContext(r.Context(), prefix)
}
prefix := parsed.Masked().String()
routes, originASNs, 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 +937,18 @@ 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,
} }
} }
-122
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"
) )
@@ -152,124 +148,6 @@ func TestStatsHandlersAnswerFromMemory(t *testing.T) {
} }
} }
// TestPrefixDetailHandlersParsePrefix checks that both prefix pages find a
// stored route whatever valid text form the URL writes its prefix in, and
// answer 400 for a prefix that does not parse. The tables store the form
// net/netip prints. Before the fix the URL text was looked up as typed, so
// "2001:DB8:0:0::/32" was not found although "2001:db8::/32" was stored, and
// "foo/24" answered 500 (https://git.eeqj.de/sneak/routewatch/issues/71).
func TestPrefixDetailHandlersParsePrefix(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)
if err := db.UpsertLiveRouteBatch([]*database.LiveRoute{
{
ID: uuid.New(), Prefix: "192.0.2.0/24", MaskLength: 24, IPVersion: 4,
OriginASN: 64500, PeerIP: "198.51.100.1", ASPath: []int{64500}, NextHop: "198.51.100.1",
LastUpdated: ts,
},
{
ID: uuid.New(), Prefix: "2001:db8::/32", MaskLength: 32, IPVersion: 6,
OriginASN: 64501, PeerIP: "2001:db8::1", ASPath: []int{64501}, NextHop: "2001:db8::1",
LastUpdated: ts,
},
}); err != nil {
t.Fatalf("UpsertLiveRouteBatch: %v", err)
}
s := New(db, streamer.New(logger.New(), metrics.New()), logger.New(), &config.Config{})
tests := []struct {
path string
want int
}{
{"/api/v1/prefix/foo/24", http.StatusBadRequest},
{"/prefix/foo/24", http.StatusBadRequest},
{"/api/v1/prefix/2001:DB8:0:0::/32", http.StatusOK},
{"/prefix/2001:DB8:0:0::/32", http.StatusOK},
{"/api/v1/prefix/192.0.2.1/24", http.StatusOK},
{"/prefix/192.0.2.1/24", http.StatusOK},
}
for _, tt := range tests {
rec := httptest.NewRecorder()
s.router.ServeHTTP(rec, httptest.NewRequest(http.MethodGet, tt.path, nil))
if rec.Code != tt.want {
t.Errorf("%s: status %d, want %d; body %s", tt.path, rec.Code, tt.want, rec.Body)
}
}
}
// 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}}