Question

我编写了一个小型Scala程序，该程序使用Apache Flink Streaming API来阅读Twitter推文。

object TwitterWordCount {
  private val properties = "/home/twitter-login.properties"
  def main(args: Array[String]) {
    val env = StreamExecutionEnvironment.getExecutionEnvironment
    val twitterStream = env.addSource(new TwitterSource(properties))
    val tweets = twitterStream
      .flatMap(new JSONParseFlatMap[String, String] {
        override def flatMap(in: String, out: Collector[String]): Unit = {
          if (getString(in, "user.lang") == "en") {
            out.collect(getString(in, "text"))
          }
        }
      })
    tweets.print
    env.execute("tweets")
  }
}

执行时遇到以下问题：

14:35:48,353 INFO  com.twitter.hbc.httpclient.ClientBase - twitterSourceClient Establishing a connection
14:35:48,354 DEBUG org.apache.http.impl.conn.PoolingClientConnectionManager - Connection request: [route: {}->http://stream.twitter.com][total kept alive: 0; route allocated: 0 of 2; total allocated: 0 of 20]
14:35:48,354 DEBUG org.apache.http.impl.conn.PoolingClientConnectionManager - Connection leased: [id: 4][route: {}->http://stream.twitter.com][total kept alive: 0; route allocated: 1 of 2; total allocated: 1 of 20]
14:35:48,354 DEBUG org.apache.http.impl.conn.DefaultClientConnectionOperator - Connecting to stream.twitter.com:80
14:35:49,486 DEBUG org.apache.flink.runtime.taskmanager.TaskManager - Received message SendHeartbeat at akka://flink/user/taskmanager_1 from Actor[akka://flink/deadLetters].
14:35:49,486 DEBUG org.apache.flink.runtime.taskmanager.TaskManager - Sending heartbeat to JobManager
14:35:49,487 DEBUG org.apache.flink.runtime.taskmanager.TaskManager - Handled message SendHeartbeat in 1 ms from Actor[akka://flink/deadLetters].
14:35:49,487 DEBUG org.apache.flink.runtime.jobmanager.JobManager - Received message Heartbeat(cb51cdb1bd08879df10bd2198b8e043a,[B@4daaaf5f) at akka://flink/user/jobmanager from Actor[akka://flink/user/taskmanager_1#-64418449].
14:35:49,488 DEBUG org.apache.flink.runtime.jobmanager.JobManager - Received hearbeat message from cb51cdb1bd08879df10bd2198b8e043a.
14:35:49,488 DEBUG org.apache.flink.runtime.instance.InstanceManager - Received heartbeat from TaskManager cb51cdb1bd08879df10bd2198b8e043a @ localhost - 8 slots - URL: akka://flink/user/taskmanager_1
14:35:49,488 DEBUG org.apache.flink.runtime.jobmanager.JobManager - Handled message Heartbeat(cb51cdb1bd08879df10bd2198b8e043a,[B@4daaaf5f) in 0 ms from Actor[akka://flink/user/taskmanager_1#-64418449].
14:35:52,358 DEBUG org.apache.http.impl.conn.DefaultClientConnection - Connection org.apache.http.impl.conn.DefaultClientConnection@64c88f2d closed
14:35:52,358 DEBUG org.apache.http.impl.conn.DefaultClientConnection - Connection org.apache.http.impl.conn.DefaultClientConnection@64c88f2d shut down
14:35:52,358 DEBUG org.apache.http.impl.conn.PoolingClientConnectionManager - Connection [id: 4][route: {}->http://stream.twitter.com] can be kept alive for 9223372036854775807 MILLISECONDS
14:35:52,358 DEBUG org.apache.http.impl.conn.DefaultClientConnection - Connection org.apache.http.impl.conn.DefaultClientConnection@64c88f2d closed
14:35:52,358 DEBUG org.apache.http.impl.conn.PoolingClientConnectionManager - Connection released: [id: 4][route: {}->http://stream.twitter.com][total kept alive: 0; route allocated: 0 of 2; total allocated: 0 of 20]
14:35:52,359 WARN  com.twitter.hbc.httpclient.ClientBase - twitterSourceClient IOException caught when establishing connection to https://stream.twitter.com/1.1/statuses/filter.json?delimited=length
14:35:53,613 WARN  com.twitter.hbc.httpclient.ClientBase - twitterSourceClient failed to establish connection properly
14:35:53,613 INFO  com.twitter.hbc.httpclient.ClientBase - twitterSourceClient Done processing, preparing to close connection
14:35:53,613 DEBUG org.apache.http.impl.conn.PoolingClientConnectionManager - Connection manager is shutting down
14:35:53,613 DEBUG org.apache.http.impl.conn.PoolingClientConnectionManager - Connection manager shut down

程序尝试重新建立连接。因此，继续发出这4行日志消息。

关于这个的奇怪之处在于，当我运行Apache Flink项目中提供的example时，一切正常（我从GitHub中提取了最新版本的master）。我甚至使用相同的属性文件。如果我将该示例类复制到我自己的项目中，则上述问题状态也会发生。

我使用Flink原型来创建自己的项目。我试过版本0.9.1以及0.10-SNAPSHOT。依赖项flink-scala，flink-streaming-scala，flink-clients和flink-connector-twitter用于相应的版本。

是否有人遇到过类似的问题，可以让我走上正轨？

Answer 1

调试com.twitter.hbc.httpclient.ClientBase让我遇到以下异常：org.apache.http.conn.ConnectTimeoutException: Connect to stream.twitter.com:80 timed out

根据Twitter Developer论坛上的post，这是因为Apaches HttpClient 4.2中存在错误。事实上，在我的项目中解析依赖树表明flink-runtime依赖于com.amazonaws：aws-java-sdk：1.81它再次依赖于org.apache.httpcomponents：httpclient：4.2。

将HttpClient 4.2.6添加到我的项目的依赖项中暂时解决了这个问题。

Answer 2

谢谢@ peedeeX21，您的解决方案帮助了我！添加显式依赖关系到pom.xml将解决从eclipse运行时的问题，但是当使用flink集群并使用flink运行提交程序时 - 使用flink发行版打包的版本仍然获胜。

我已经通过将httpclient-4.2.6.jar jar下载到flink / lib并将其重命名为＆＃34; a＆＃34;来解决它。（ahttpclient-4.2.6.jar）所以它将首先添加到flink运行时的类路径中（由bin / config.sh完成）希望它对某人有所帮助。

使用Apache Flink连接到Twitter Streaming API时的IOExcpetion

2 个答案: