如何使用 Kafka .NET 客户端 API 列出每个分区和消费者组的 Kafka 消费者滞后和最新偏移量?

How can I list the Kafka consumer lag and latest offset per partition and consumer group using the Kafka .NET client API?

我正在尝试使用 .NET Confluent.Kafka 1.4.0-RC1(对于 Net472)来解决消费者延迟问题。 我可以使用此脚本获得所需的结果:

$ bin/kafka-run-class.sh kafka.admin.ConsumerGroupCommand --bootstrap-server localhost:9092 --group Grp1 --describe

结果:

GROUP           TOPIC           PARTITION  CURRENT-OFFSET  LOG-END-OFFSET  LAG             CONSUMER-ID                                  HOST            CLIENT-ID
Grp1            test3           1          15              15              0               rdkafka-ca76855f-7b66-4bf1-82bc-73e9a1c1cf71 /10.186.129.93  rdkafka
Grp1            test3           2          13              13              0               rdkafka-d64379dc-881a-4f6f-a793-51e832cc2f5a /10.186.129.93  rdkafka
Grp1            test3           0          9               9               0               rdkafka-a25bdb80-3b70-4e42-963e-d41ad9e2a99a /10.186.129.93  rdkafka
Grp1            test            0          68              68              0               -                                            -               -

我无法使用 .NET 客户端代码获得类似的报告。这是我一直在尝试的代码 - 但由于 consumer.Assignment 属性 有一个空集合,所以一无所获。

private string WriteConsumerGroupLags(string bootstrapServers, string consumerGroupName) {
    // kafka-console-consumer.bat --zookeeper MW45670117:2380 --topic powertelemetry --consumer-property group.id=test123 --consumer-property enable.auto.commit=true

    StringBuilder sb = new StringBuilder();
    sb.AppendLine("\n");
    sb.AppendLine("Consumer Group Lag Report");
    sb.AppendLine("-------------------------");
    ConsumerConfig config = new ConsumerConfig {
        BootstrapServers = bootstrapServers,
        GroupId = consumerGroupName,
        AutoOffsetReset = AutoOffsetReset.Earliest
    };

    using (var consumer = new ConsumerBuilder<Ignore, string>(config).Build()) {
        foreach (TopicPartition tp in consumer.Assignment) {
            string topic = tp.Topic;
            int partitionID = tp.Partition.Value;
            // gets the current position (offset) for the specific topic/partition
            Offset offset = consumer.Position(new TopicPartition(topic, new Partition(partitionID)));
            sb.AppendLine($"Offset value is: {offset.Value}");

            // returns current commited offset for the current assignment
            List<TopicPartitionOffset> tpos = consumer.Committed(TimeSpan.FromSeconds(4));
            foreach (TopicPartitionOffset tpo in tpos) {
                sb.AppendLine($"Commited offset for partition {tpo.TopicPartition} is {tpo.Offset}");
            }
        }
    }

    return sb.ToString();
}

正在寻找每个 partition/consumer-group 的消费者滞后和最新偏移量。

好的,我终于找到了怎么做。看起来您需要将主题分区分配给消费者组。在我的例子中,主题是 test3,我有 3 个分区。这必须在代码中设置。我想一个更通用的实现是将与该组关联的列表传递给该方法。 以下是我如何获得消费者延迟:

private string WriteConsumerGroupLags(string bootstrapServers, string consumerGroupName) {
    StringBuilder sb = new StringBuilder();
    sb.AppendLine("\n");
    sb.AppendLine("Consumer Group Lag Report");
    sb.AppendLine("-------------------------");
    ConsumerConfig config = new ConsumerConfig {
        BootstrapServers = bootstrapServers,
        GroupId = consumerGroupName,
        AutoOffsetReset = AutoOffsetReset.Earliest
    };

    using (var consumer = new ConsumerBuilder<Ignore, string>(config).Build()) {
        List<TopicPartition> topic_partitions = new List<TopicPartition>() {
            { new TopicPartition("test3", new Partition(0)) },
            { new TopicPartition("test3", new Partition(1) ) },
            { new TopicPartition("test3", new Partition(2) ) } };

        consumer.Assign(topic_partitions);

        List<TopicPartitionOffset> tpos = consumer.Committed(TimeSpan.FromSeconds(40));
        foreach (TopicPartitionOffset tpo in tpos) {
            WatermarkOffsets w = consumer.QueryWatermarkOffsets(tpo.TopicPartition, TimeSpan.FromSeconds(40));
            long commited = tpo.Offset.Value;
            long log_end_offset = w.High.Value;
            long lag = log_end_offset - commited;
            sb.AppendLine($"Commited offset for Topic {tpo.TopicPartition.Topic} Partition {tpo.TopicPartition.Partition} is {commited} out of Wantermark end offset {log_end_offset} Lag is: {lag}");
        }

        sb.AppendLine();
    }

    return sb.ToString();
}

我对您的回答进行了扩展,包括为主题分区生成模型。

public class Offsets
{
   public string WriteConsumerGroupLags(string bootstrapServers, string consumerGroupName)
   {
      var sb = new StringBuilder();
      sb.AppendLine("\n");
      sb.AppendLine("Consumer Group Lag Report");
      sb.AppendLine("-------------------------");

      var config = new ConsumerConfig
      {
         BootstrapServers = bootstrapServers,
         GroupId = consumerGroupName,
         SecurityProtocol = SecurityProtocol.Ssl
      };

      using var adminClient = new AdminClientBuilder(config).Build();
      using var consumer = new ConsumerBuilder<Ignore, string>(config).Build();


      var meta = adminClient.GetMetadata(TimeSpan.FromSeconds(20));

      var topicPartitions = new List<TopicPartition>();

      foreach (var topicMeta in meta.Topics.Where(x => !x.Topic.Contains("__")))
      {
         topicPartitions.AddRange(topicMeta.Partitions.Select(topicMetaPartition => new TopicPartition(topicMeta.Topic, new Partition(topicMetaPartition.PartitionId))));
      }

      consumer.Assign(topicPartitions);

      List<TopicPartitionOffset> tpos = consumer.Committed(TimeSpan.FromSeconds(40));
      foreach (TopicPartitionOffset tpo in tpos)
      {
         WatermarkOffsets w = consumer.QueryWatermarkOffsets(tpo.TopicPartition, TimeSpan.FromSeconds(40));
         long commited = tpo.Offset.Value;
         long log_end_offset = w.High.Value;
         long lag = log_end_offset - commited;
         sb.AppendLine(
            $"Commited offset for Topic {tpo.TopicPartition.Topic} Partition {tpo.TopicPartition.Partition} is {commited} out of Wantermark end offset {log_end_offset} Lag is: {lag}");
      }

      sb.AppendLine();

      return sb.ToString();
   }
}