fix: avoid infinite retry in blue-green migration test

pulsar/blue_green_migration_test.go

@@ -21,6 +21,7 @@ package pulsar
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"net/http"
 	"runtime"
@@ -148,6 +149,7 @@ func testTopicMigrate(
 	// Signals both producer and consumer have processed `messageCountBeforeUnload` messages
 	wgSendAndReceiveMessages := sync.WaitGroup{}
 	wgSendAndReceiveMessages.Add(2)
+	errCh := make(chan error, 1)
 
 	// Producer
 	go func() {
@@ -159,14 +161,29 @@ func testTopicMigrate(
 			}
 
 			pm := ProducerMessage{Payload: []byte(fmt.Sprintf("hello-%d", i))}
+			retryStarted := time.Now()
 
 			for true {
 				ctx, cancel := context.WithTimeout(context.Background(), 1*time.Second)
-				defer cancel()
 				_, err := producer.Send(ctx, &pm)
+				cancel()
 				if err == nil {
 					break
 				}
+				if errors.Is(err, ErrTopicTerminated) || errors.Is(err, ErrProducerClosed) {
+					select {
+					case errCh <- fmt.Errorf("producer became terminal during migration at message %d: %w", i, err):
+					default:
+					}
+					return
+				}
+				if time.Since(retryStarted) > 30*time.Second {
+					select {
+					case errCh <- fmt.Errorf("producer send retry exceeded 30s at message %d: %w", i, err):
+					default:
+					}
+					return
+				}
 				time.Sleep(1000 * time.Millisecond)
 			}
 
@@ -185,16 +202,24 @@ func testTopicMigrate(
 				wgUnload.Wait()
 			}
 
+			retryStarted := time.Now()
 			for true {
 				ctx, cancel := context.WithTimeout(context.Background(), 1*time.Second)
-				defer cancel()
 				m, err := consumer.Receive(ctx)
+				cancel()
 				if err == nil {
 					err = consumer.Ack(m)
 					if err == nil {
 						break
 					}
 				}
+				if time.Since(retryStarted) > 30*time.Second {
+					select {
+					case errCh <- fmt.Errorf("consumer receive/ack retry exceeded 30s at message %d: %w", i, err):
+					default:
+					}
+					return
+				}
 				time.Sleep(100 * time.Millisecond)
 			}
 
@@ -205,7 +230,27 @@ func testTopicMigrate(
 	}()
 
 	// Unload the bundle, triggering the producers and consumers to reconnect to the specified broker.
-	wgSendAndReceiveMessages.Wait()
+	waitWithError := func(wg *sync.WaitGroup, stage string) bool {
+		doneCh := make(chan struct{})
+		go func() {
+			wg.Wait()
+			close(doneCh)
+		}()
+
+		select {
+		case <-doneCh:
+			return true
+		case err := <-errCh:
+			req.NoError(err, stage)
+			return false
+		case <-time.After(90 * time.Second):
+			req.FailNow(stage + " timeout")
+			return false
+		}
+	}
+	if !waitWithError(&wgSendAndReceiveMessages, "waiting for pre-unload send/receive") {
+		return
+	}
 
 	topicMigrationURL = fmt.Sprintf(
 		"/admin/v2/clusters/%s/migrate?migrated=true", cluster)
@@ -248,5 +293,7 @@ func testTopicMigrate(
 		return req.Equal(dstTopicBrokerURL, topicBrokerURL)
 	})
 
-	wgRoutines.Wait()
+	if !waitWithError(&wgRoutines, "waiting for producer/consumer routines") {
+		return
+	}
 }