test(storage/remote/queue_manager_test.go): use synctest in TestShutdown for better

control over time

The test becomes flaky after it was asked to run on parallel
and "fight" for resources

let's hide all of that

Signed-off-by: machine424 <ayoubmrini424@gmail.com>

storage/remote/queue_manager.go

@@ -962,10 +962,10 @@ func (t *QueueManager) Stop() {
 	defer t.logger.Info("Remote storage stopped.")
 
 	close(t.quit)
-	t.wg.Wait()
 	// Wait for all QueueManager routines to end before stopping shards, metadata watcher, and WAL watcher. This
 	// is to ensure we don't end up executing a reshard and shards.stop() at the same time, which
 	// causes a closed channel panic.
+	t.wg.Wait()
 	t.shards.stop()
 	t.watcher.Stop()
 	if t.mcfg.Send {
@@ -1458,9 +1458,18 @@ func (q *queue) ReturnForReuse(batch []timeSeries) {
 // FlushAndShutdown stops the queue and flushes any samples. No appends can be
 // made after this is called.
 func (q *queue) FlushAndShutdown(done <-chan struct{}) {
+	ticker := time.NewTicker(time.Second)
+	defer ticker.Stop()
+
+loop:
 	for q.tryEnqueueingBatch(done) {
-		time.Sleep(time.Second)
+		select {
+		case <-done:
+			break loop
+		case <-ticker.C:
 		}
+	}
+
 	q.batchMtx.Lock()
 	defer q.batchMtx.Unlock()
 	q.batch = nil

storage/remote/queue_manager_test.go

@@ -24,6 +24,7 @@ import (
 	"strings"
 	"sync"
 	"testing"
+	"testing/synctest"
 	"time"
 
 	"github.com/gogo/protobuf/proto"
@@ -34,6 +35,7 @@ import (
 	"github.com/prometheus/common/promslog"
 	"github.com/stretchr/testify/require"
 	"go.uber.org/atomic"
+	"go.uber.org/goleak"
 
 	"github.com/prometheus/prometheus/config"
 	"github.com/prometheus/prometheus/model/histogram"
@@ -53,6 +55,10 @@ import (
 
 const defaultFlushDeadline = 1 * time.Minute
 
+func TestMain(m *testing.M) {
+	goleak.VerifyTestMain(m)
+}
+
 func TestBasicContentNegotiation(t *testing.T) {
 	t.Parallel()
 	queueConfig := config.DefaultQueueConfig
@@ -456,8 +462,9 @@ func TestSampleDeliveryOrder(t *testing.T) {
 }
 
 func TestShutdown(t *testing.T) {
-	// Not t.Parallel() because the test became flaky; see https://github.com/prometheus/prometheus/issues/17045
+	t.Parallel()
-	deadline := 1 * time.Second
+	synctest.Run(func() {
+		deadline := 15 * time.Second
 		c := NewTestBlockedWriteClient()
 
 		cfg := config.DefaultQueueConfig
@@ -473,7 +480,7 @@ func TestShutdown(t *testing.T) {
 		go func() {
 			m.Append(samples)
 		}()
-	time.Sleep(100 * time.Millisecond)
+		synctest.Wait()
 
 		// Test to ensure that Stop doesn't block.
 		start := time.Now()
@@ -481,13 +488,8 @@ func TestShutdown(t *testing.T) {
 		// The samples will never be delivered, so duration should
 		// be at least equal to deadline, otherwise the flush deadline
 		// was not respected.
-	duration := time.Since(start)
+		require.Equal(t, time.Since(start), deadline)
-	if duration > deadline+(deadline/10) {
+	})
-		t.Errorf("Took too long to shutdown: %s > %s", duration, deadline)
-	}
-	if duration < deadline {
-		t.Errorf("Shutdown occurred before flush deadline: %s < %s", duration, deadline)
-	}
 }
 
 func TestSeriesReset(t *testing.T) {