stance(x._2, y._2)
> (toId,score)
> }.toList.sortWith((x,y)=>x._2>y._2).take(20)
> (fromId,docSims)
> }
> res.writeAsText(..)
>
> - 原始邮件 -
> 发件人：Stephan Ewen
> 收件人：user@flink.apache.org
> 抄送人：亘谷
> 主题：Re: flink Broadcast
&
= y._1 val score = 1-cosDistince.distance(x._2,
y._2)(toId,score)
}.toList.sortWith((x,y)=>x._2>y._2).take(20) (fromId,docSims)
}res.writeAsText(..)
- 原始邮件 -
发件人：Stephan Ewen
收件人：user@flink.apache.org
抄送人：亘谷
主题：Re: flink Broadcast
日期：20
The program consists of two executions - one that only collects() back to
the client, one that executes the map function.
Are you running this as a "YARN single job" execution? IN that case, there
may be an issue that this incorrectly tries to submit to a stopping YARN
cluster.
On Fri, Mar 24,
Hi,
Can you provide more logs to help us understand whats going on?
One note regarding your application: You are calling .collect() and send
the collection with the map() call to the cluster again.
This is pretty inefficient and can potentially break your application (in
particular the RPC system
