我编写了一个小型scala程序,它使用apache flink流api来读取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,这4行日志消息将继续发出。
奇怪的是,当我运行apache flink项目中提供的示例时,一切正常(我从github获取了master的最新版本)。我甚至使用相同的属性文件。如果我将这个示例类复制到自己的项目中,上面的问题状态也会出现。
我用flink原型来创建我自己的项目。我尝试了版本0.9.1和0.10-snapshot。依赖关系 flink-scala
, flink-streaming-scala
, flink-clients
以及 flink-connector-twitter
在相应的版本中使用。
有没有人经历过类似的问题,能让我走上正轨?
2条答案
按热度按时间4bbkushb1#
调试
com.twitter.hbc.httpclient.ClientBase
给我带来了以下例外:org.apache.http.conn.ConnectTimeoutException: Connect to stream.twitter.com:80 timed out
根据twitter开发者论坛上的一篇文章,发生这种情况是因为apacheshttpclient4.2中的一个bug。事实上,解析我的项目上的依赖树表明flink运行时对com有依赖关系。amazonaws:aws-java-sdk:1.81,它同样依赖于org.apache.httpcomponents:httpclient:4.2.将httpclient4.2.6添加到我的项目的依赖项中暂时解决了这个问题。
ckx4rj1h2#
谢谢@peedeex21你的解决方案帮了我!在pom.xml中添加显式依赖项可以解决从eclipse运行时的问题,但是当使用flink cluster并提交带有flink run的程序时,使用flink发行版打包的版本仍然会获胜。
我将httpclient-4.2.6.jar下载到flink/lib中,并将其重命名为“a”(ahttpclient-4.2.6.jar),因此它将首先添加到flink运行时的类路径中(由bin/config.sh完成),希望对某些人有所帮助。