Fix two goroutine leaks: stats handlers on timeout, streamer tickers on reconnect (closes #12)
check / check (push) Failing after 1s

The stats handlers ran the database query in a goroutine that sent on an
unbuffered channel. When the 4s request timeout won, nothing received and
the goroutine blocked forever; the status page polls every 2s, so once the
query exceeds the timeout every poll leaked one goroutine. Give both
channels capacity 1 so the send always completes.

The streamer started two ticker goroutines per connection that exited only
with the streamer's lifetime context, leaking two on every reconnect. Scope
them to a per-connection context cancelled when the stream call returns.

Tests force the stats timeout repeatedly and drive many reconnects, then
assert the goroutine count settles back to its starting value. The streamer
gains an internal endpoint field so a test can point it at a local server.

Model: opus-4-8
This commit was merged in pull request #19.
This commit is contained in:
2026-09-21 16:29:32 +02:00
parent cb2374f033
commit 63d62b7bc1
4 changed files with 191 additions and 9 deletions
+10 -6
View File
@@ -179,9 +179,11 @@ func (s *Server) handleStatusJSON() http.HandlerFunc {
metrics := s.streamer.GetMetrics()
// Get database stats with timeout
statsChan := make(chan database.Stats)
errChan := make(chan error)
// Get database stats with timeout. The channels are buffered so the
// goroutine's send never blocks if the timeout wins and nothing here
// receives; otherwise it would block forever and leak.
statsChan := make(chan database.Stats, 1)
errChan := make(chan error, 1)
go func() {
dbStats, err := s.db.GetStatsContext(ctx)
@@ -398,9 +400,11 @@ func (s *Server) handleStats() http.HandlerFunc {
metrics := s.streamer.GetMetrics()
// Get database stats with timeout
statsChan := make(chan database.Stats)
errChan := make(chan error)
// Get database stats with timeout. The channels are buffered so the
// goroutine's send never blocks if the timeout wins and nothing here
// receives; otherwise it would block forever and leak.
statsChan := make(chan database.Stats, 1)
errChan := make(chan error, 1)
go func() {
dbStats, err := s.db.GetStatsContext(ctx)
+99
View File
@@ -0,0 +1,99 @@
package server
import (
"context"
"net/http"
"net/http/httptest"
"runtime"
"testing"
"time"
"git.eeqj.de/sneak/routewatch/internal/database"
"git.eeqj.de/sneak/routewatch/internal/logger"
"git.eeqj.de/sneak/routewatch/internal/metrics"
"git.eeqj.de/sneak/routewatch/internal/streamer"
)
// blockingStatsDB embeds database.Store (left nil) and overrides only
// GetStatsContext, which blocks until release is closed. The stats handlers
// call it in a goroutine; every other Store method is unused on the timeout
// path and would panic if called.
type blockingStatsDB struct {
database.Store
release chan struct{}
}
func (d blockingStatsDB) GetStatsContext(_ context.Context) (database.Stats, error) {
<-d.release
return database.Stats{}, nil
}
// TestStatsHandlersDoNotLeakOnTimeout drives each stats handler repeatedly with
// a request whose context times out before the database responds, then releases
// the blocked queries and asserts the goroutine count returns to its starting
// value. Before the fix the per-request goroutine sent on an unbuffered channel
// that nothing received once the timeout won, so it blocked forever and every
// poll leaked one goroutine.
func TestStatsHandlersDoNotLeakOnTimeout(t *testing.T) {
release := make(chan struct{})
db := blockingStatsDB{release: release}
s := New(db, streamer.New(logger.New(), metrics.New()), logger.New())
handlers := map[string]http.HandlerFunc{
"status.json": s.handleStatusJSON(),
"stats": s.handleStats(),
}
baseline := settledGoroutineCount()
const (
iterations = 20
requestTimeout = 50 * time.Millisecond
)
for _, handler := range handlers {
for range iterations {
ctx, cancel := context.WithTimeout(context.Background(), requestTimeout)
req := httptest.NewRequest(http.MethodGet, "/", nil).WithContext(ctx)
handler(httptest.NewRecorder(), req)
cancel()
}
}
// Let the blocked queries finish; with buffered channels each goroutine's
// send now succeeds and the goroutine exits.
close(release)
if !waitForGoroutines(baseline) {
t.Fatalf("goroutines did not return to baseline %d, got %d",
baseline, runtime.NumGoroutine())
}
}
// settledGoroutineCount lets transient goroutines finish, then reports the
// current count.
func settledGoroutineCount() int {
prev := runtime.NumGoroutine()
for range 20 {
time.Sleep(10 * time.Millisecond)
cur := runtime.NumGoroutine()
if cur == prev {
return cur
}
prev = cur
}
return prev
}
// waitForGoroutines waits until the goroutine count drops to target or below.
func waitForGoroutines(target int) bool {
for range 100 {
if runtime.NumGoroutine() <= target {
return true
}
time.Sleep(10 * time.Millisecond)
}
return false
}