无法使用Sarama Golang软件包创建Kafka生产者客户端-“客户端/元数据在获取元数据时从代理处出错：EOF”

Question

Versions: GoLang 1.10.2 Kafka 4.4.1 Docker 18.03.1

I'm trying to use Shopify's Sarama package to test out my Kafka instance. I used Docker compose to stand up Kafka/Zookeeper and it is all successfully running.

When I try to create a Producer client with Sarama, an error is thrown.

When I run the following

    package main

import (
"fmt"
"log"
"os"
"os/signal"
"time"

"strconv"

"github.com/Shopify/sarama"

)

func main() {


// Setup configuration
config := sarama.NewConfig()
config.Producer.Return.Successes = true
config.Producer.Partitioner = sarama.NewRandomPartitioner
config.Producer.RequiredAcks = sarama.WaitForAll
brokers := []string{"localhost:29092"}
producer, err := sarama.NewAsyncProducer(brokers, config)
if err != nil {
    // Should not reach here
    panic(err)
}

defer func() {
    if err := producer.Close(); err != nil {
        // Should not reach here
        panic(err)
    }
}()

I get this

[sarama] 2018/06/12 17:22:05 Initializing new client

[sarama] 2018/06/12 17:22:05 client/metadata fetching metadata for all topics from broker localhost:29092

[sarama] 2018/06/12 17:22:05 Connected to broker at localhost:29092 (unregistered)

[sarama] 2018/06/12 17:22:05 client/metadata got error from broker while fetching metadata: EOF

[sarama] 2018/06/12 17:22:05 Closed connection to broker localhost:29092

{sarama] 2018/06/12 17:22:05 client/metadata no available broker to send metadata request to

[sarama] 2018/06/12 17:22:06 Closing Client panic: kafka: client has run out of available brokers to talk to (Is your cluster reachable?)

goroutine 1 [running]: main.main() /Users/benwornom/go/src/github.com/acstech/doppler-events/testprod/main.go:29 +0x3ec exit status 2

Sarama did try several times in a row to create a producer client, but failed each time.

My understanding of Sarama's "NewAsyncProducer" method is that it calls "NewClient", which is invoked regardless of whether you are creating a Producer or Consumer. NewClient attempts to gather metadata from the Kafka broker, which is failing in my situation. I know it is connecting to the Kafka broker, but once it connects it seems to break. Any advice would be helpful. My network connection is strong, I can't think of anything interfering with the server. As far as I know, I only have one broker and one partition for the existing topic. I don't think I have to manually assign a topic to a broker. If my client is connecting with the broker, why can't I establish a lasting connection for my producer?

This is from the kafka log file right before it dies.

__consumer_offsets-5 -> Vector(1), connect-offsets-23 -> Vector(1), __consumer_offsets-43 -> Vector(1), __consumer_offsets-32 -> Vector(1), __consumer_offsets-21 -> Vector(1), __consumer_offsets-10 -> Vector(1), connect-offsets-20 -> Vector(1), __consumer_offsets-37 -> Vector(1), connect-offsets-9 -> Vector(1), connect-status-4 -> Vector(1), __consumer_offsets-48 -> Vector(1), __consumer_offsets-40 -> Vector(1), __consumer_offsets-29 -> Vector(1), __consumer_offsets-18 -> Vector(1), connect-offsets-14 -> Vector(1), __consumer_offsets-7 -> Vector(1), __consumer_offsets-34 -> Vector(1), __consumer_offsets-45 -> Vector(1), __consumer_offsets-23 -> Vector(1), connect-offsets-6 -> Vector(1), connect-status-1 -> Vector(1), connect-offsets-17 -> Vector(1), connect-offsets-0 -> Vector(1), connect-offsets-22 -> Vector(1), __consumer_offsets-26 -> Vector(1), connect-offsets-11 -> Vector(1), __consumer_offsets-15 -> Vector(1), __consumer_offsets-4 -> Vector(1), __consumer_offsets-42 -> Vector(1), __consumer_offsets-9 -> Vector(1), __consumer_offsets-31 -> Vector(1), __consumer_offsets-20 -> Vector(1), connect-offsets-3 -> Vector(1), __consumer_offsets-1 -> Vector(1), __consumer_offsets-12 -> Vector(1), connect-offsets-8 -> Vector(1), connect-offsets-19 -> Vector(1), connect-status-3 -> Vector(1), __confluent.support.metrics-0 -> Vector(1), __consumer_offsets-17 -> Vector(1), __consumer_offsets-28 -> Vector(1), __consumer_offsets-6 -> Vector(1), __consumer_offsets-39 -> Vector(1), __consumer_offsets-44 -> Vector(1), connect-offsets-16 -> Vector(1), connect-status-0 -> Vector(1), connect-offsets-5 -> Vector(1), connect-offsets-21 -> Vector(1), __consumer_offsets-47 -> Vector(1), __consumer_offsets-36 -> Vector(1), __consumer_offsets-14 -> Vector(1), __consumer_offsets-25 -> Vector(1), __consumer_offsets-3 -> Vector(1), __consumer_offsets-30 -> Vector(1), __consumer_offsets-41 -> Vector(1), connect-offsets-13 -> Vector(1), connect-offsets-24 -> Vector(1), connect-offsets-2 -> Vector(1), connect-configs-0 -> Vector(1), __consumer_offsets-11 -> Vector(1), __consumer_offsets-22 -> Vector(1), __consumer_offsets-33 -> Vector(1), __consumer_offsets-0 -> Vector(1), connect-offsets-7 -> Vector(1), connect-offsets-18 -> Vector(1))) (kafka.controller.KafkaController) [36mkafka_1 |[0m [2018-06-12 20:24:47,461] DEBUG [Controller id=1] Topics not in preferred replica for broker 1 Map() (kafka.controller.KafkaController) [36mkafka_1 |[0m [2018-06-12 20:24:47,462] TRACE [Controller id=1] Leader imbalance ratio for broker 1 is 0.0 (kafka.controller.KafkaController)

Answer 1

I ran into the same issue and I did the following which solve my problem:

1. Check what version of Kafka you're using. The default Kafka version specified in the config when passed to NewAsyncProducer is V0_8_2_0. Make sure you specify the correct Kafka version in the config:

   config := sarama.NewConfig()
   config.Version = sarama.V1_1_0_0
   

2. Make sure the broker URL you are passing is correct. You should be passing the broker URL, not the zookeeper URL or another URL. The default kafka port is 9092 so the URL should be similar to BROKER_URL:9092 if you used the default port.