查看 PySpark 脚本的 Glue 作业输出的最佳方式

Posted 2023-04-15

【中文标题】查看 PySpark 脚本的 Glue 作业输出的最佳方式

【英文标题】：Best Way to View Output of Glue Jobs for PySpark Scripts

【发布时间】：2020-09-23 05:52:00

【问题描述】：

所以我创建了一个调用 Python 脚本并进行 PySpark 转换的作业。但是，当我从 AWS Cloudwatch 查看Output 时，输出中有很多对我来说并不重要的信息。例如：

at org.apache.spark.rdd.NewHadoopRDD$$anon$1.liftedTree1$1(NewHadoopRDD.scala:199)
at org.apache.spark.rdd.NewHadoopRDD$$anon$1.<init>(NewHadoopRDD.scala:196)
at org.apache.spark.rdd.NewHadoopRDD.compute(NewHadoopRDD.scala:151)
at org.apache.spark.rdd.NewHadoopRDD.compute(NewHadoopRDD.scala:70)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
at org.apache.spark.rdd.UnionRDD.compute(UnionRDD.scala:105)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)

有没有办法将 Glue 作业配置为仅输出到日志语句，例如来自我设置的 print() 或仅由于错误或异常？

【问题讨论】：

您是否尝试过启用连续记录 docs.aws.amazon.com/glue/latest/dg/… 并使用自定义脚本记录器记录特定于应用程序的消息？

这是启用Continuous Logging 的结果。现在我想过滤这些日志结果以便于调试。

【参考方案1】：

AWS 最近发布了 Glue for V2.0 的更新。这个版本改变了日志的排序/分组方式，现在我只想查看的日志（例如来自 print()）现在放在一个地方，我可以在 AWS Cloudwatch 中轻松查看。

所以解决方案是更新您的 Glue 作业以使用 Glue 2.0。