[fix] retry producer creation upon error after succssful topic lookup

pulsar/producer_partition.go

@@ -197,7 +197,49 @@ func newPartitionProducer(client *client, topic string, options *ProducerOptions
 	} else {
 		p.userProvidedProducerName = false
 	}
-	err := p.grabCnx()
+	// retry to create producer when failed with maxRetry
+	var maxRetry int
+	if p.options.MaxReconnectToBroker == nil {
+		maxRetry = -1
+	} else {
+		maxRetry = int(*p.options.MaxReconnectToBroker)
+	}
+
+	var delayReconnectTime time.Duration
+	defaultBackoff := internal.DefaultBackoff{}
+
+	var err error
+	for maxRetry != 0 {
+		if p.options.BackoffPolicy == nil {
+			delayReconnectTime = defaultBackoff.Next()
+		} else {
+			delayReconnectTime = p.options.BackoffPolicy.Next()
+		}
+
+		atomic.AddUint64(&p.epoch, 1)
+		err = p.grabCnx()
+		if err == nil {
+			break
+		}
+		p.log.WithError(err).Error("Failed to create producer at newPartitionProducer")
+		if errors.Is(err, ErrTopicNotfound) {
+			// when topic is not found, do not attempt to reconnect
+			p.log.Warn("Failed to create producer due to Topic Not Found")
+			break
+		}
+
+		if errors.Is(err, ErrTopicTerminated) {
+			p.log.Info("Topic was terminated, failing pending messages, will not create producer")
+			break
+		}
+
+		if maxRetry > 0 {
+			maxRetry--
+		}
+		logger.WithError(err).
+			Error("Failed to create producer at newPartitionProducer retry to create producer")
+		time.Sleep(delayReconnectTime)
+	}
 	if err != nil {
 		p.batchFlushTicker.Stop()
 		logger.WithError(err).Error("Failed to create producer at newPartitionProducer")