python - 如何增加bluemix上spark-submit作业的日志输出？

Question

我已将 python 作业作为服务提交给 bluemix Spark，但它失败了。不幸的是，日志记录不足，并且没有给我任何失败原因的线索。

如何提高日志级别输出？

spark 作为服务的输出:

==== Failed Status output =====================================================

Getting status
HTTP/1.1 200 OK
Server: nginx/1.8.0
Date: Thu, 12 May 2016 19:09:30 GMT
Content-Type: application/json;charset=utf-8
Content-Length: 850
Connection: keep-alive

{
  "action" : "SubmissionStatusResponse",
  "driverState" : "ERROR",
  "message" : "Exception from the cluster:
org.apache.spark.SparkUserAppException: User application exited with 255
    org.apache.spark.deploy.PythonRunner$.main(PythonRunner.scala:88)
    org.apache.spark.deploy.PythonRunner.main(PythonRunner.scala)
    sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:95)
    sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:55)
    java.lang.reflect.Method.invoke(Method.java:507)
    org.apache.spark.deploy.ego.EGOClusterDriverWrapper$$anon$3.run(EGOClusterDriverWrapper.scala:430)",
  "serverSparkVersion" : "1.6.0",
  "submissionId" : "xxxxxx",
  "success" : true
}
===============================================================================

我已针对 BigInsights 集群成功运行相同的作业。在 biginsights 集群上运行时，我还会得到更详细的输出。

Answer 1

有 stderr-%timestamp% 和 stdout-%timestamp% 文件从集群下载到您运行 spark-submit.sh< 的本地目录。 通常您会在这两个文件中找到工作问题。

引用:http://spark.apache.org/docs/latest/spark-standalone.html#monitoring-and-logging