在EMR Spark上,通过数据帧向{3}写入RDD[String]
。
rddString
.toDF()
.coalesce(16)
.write
.option("compression", "gzip")
.mode(SaveMode.Overwrite)
.json(s"s3n://my-bucket/some/new/path")
保存模式为Overwrite
且s3n://my-bucket/some/new/path
尚不存在。
我一直得到IOException: File already exists
:
org.apache.spark.SparkException: Job aborted due to stage failure: Task 15 in stage 55.0 failed 4 times, most recent failure: Lost task 15.3 in stage 55.0 (TID 8441, ip-172-31-17-30.us-west-2.compute.internal, executor 3): org.apache.spark.SparkException: Task failed while writing rows
at org.apache.spark.sql.execution.datasources.FileFormatWriter$.org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask(FileFormatWriter.scala:270)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$apply$mcV$sp$1.apply(FileFormatWriter.scala:189)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$apply$mcV$sp$1.apply(FileFormatWriter.scala:188)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
at org.apache.spark.scheduler.Task.run(Task.scala:108)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:338)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)
Caused by: java.io.IOException: File already exists:s3n://my-bucket/some/new/path/part-00015-03a0c001-fc99-4055-9be5-68a1fb0cf6d3-c000.json.gz
at com.amazon.ws.emr.hadoop.fs.s3n.S3NativeFileSystem.create(S3NativeFileSystem.java:625)
at org.apache.hadoop.fs.FileSystem.create(FileSystem.java:932)
at org.apache.hadoop.fs.FileSystem.create(FileSystem.java:913)
at org.apache.hadoop.fs.FileSystem.create(FileSystem.java:810)
at com.amazon.ws.emr.hadoop.fs.EmrFileSystem.create(EmrFileSystem.java:176)
at org.apache.spark.sql.execution.datasources.CodecStreams$.createOutputStream(CodecStreams.scala:81)
at org.apache.spark.sql.execution.datasources.CodecStreams$.createOutputStreamWriter(CodecStreams.scala:92)
at org.apache.spark.sql.execution.datasources.json.JsonOutputWriter.<init>(JsonFileFormat.scala:140)
at org.apache.spark.sql.execution.datasources.json.JsonFileFormat$$anon$1.newInstance(JsonFileFormat.scala:80)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$SingleDirectoryWriteTask.newOutputWriter(FileFormatWriter.scala:303)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$SingleDirectoryWriteTask.execute(FileFormatWriter.scala:312)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:256)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:254)
at org.apache.spark.util.Utils$.tryWithSafeFinallyAndFailureCallbacks(Utils.scala:1371)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$.org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask(FileFormatWriter.scala:259)
... 8 more
Spark v2.2.1,EMR v5.12.0
在抛出异常之前,文件将写入目标。但是,我不知道它们是否完整。
答案 0 :(得分:4)
当我用胶水工作EMR时遇到了类似的问题。简而言之,通常不是导致您的工作失败的真正根源。火花任务可能由于其他原因而失败。重试原始故障后,它最终引发此“ IOException:文件已存在”。
所以找到并解决真正的根本原因,它也将消失。
就我而言,报告的错误在CloudWatch ErrorLogs中如下所示:
: org.apache.spark.SparkException: Job aborted.
at ...
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:338)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)
Caused by: java.io.IOException: File already exists:s3://personal-tests/xdqian/zappos_triplet_loss/output_cache_test/part-00003-8eaa7c78-e227-4476-b96d-4300e7350bc7-c000.csv
我没有任何线索,但是当我检查日志时,发现了如下异常:
18/12/05 06:14:15 ERROR Utils: Aborting task
org.apache.spark.api.python.PythonException: Traceback (most recent call last):
File "/mnt/yarn/usercache/root/appcache/application_1543990079218_0001/container_1543990079218_0001_01_000101/pyspark.zip/pyspark/worker.py", line 177, in main
process()
File "/mnt/yarn/usercache/root/appcache/application_1543990079218_0001/container_1543990079218_0001_01_000101/pyspark.zip/pyspark/worker.py", line 172, in process
serializer.dump_stream(func(split_index, iterator), outfile)
File "/mnt/yarn/usercache/root/appcache/application_1543990079218_0001/container_1543990079218_0001_01_000101/pyspark.zip/pyspark/serializers.py", line 268, in dump_stream
vs = list(itertools.islice(iterator, batch))
File "/mnt/yarn/usercache/root/appcache/application_1543990079218_0001/container_1543990079218_0001_01_000001/GoldenGardensGluePythonScripts.zip/golden_gardens_glue_python_scripts/job.py", line 62, in <lambda>
TypeError: 'NoneType' object has no attribute '__getitem__'
最后,在解决此NoneType错误之后,“文件已存在”异常消失了。我读过其他一些材料(对不起,我无法继续追踪),“文件已存在”错误总是由任务失败引起,并由于其他一些问题(在我的情况下为NoneType)而重试。我期望执行程序任务创建一个文件并逐行输出数据。由于NoneType错误,它可能在第34行失败并被中止,而文件仍然存在前33行。 据说失败的任务将重试4次。重试任务时,它将在开始时通过先前的运行找到存在的文件。 因此,根本原因实际上是记录为Loggs,在ErrorLogs中带有“文件已存在”异常,因为这是作业终止之前的最终异常。 覆盖模式在这里无济于事,因为只会在开始时进行检查,而不会在这种情况下提供控制标志。
答案 1 :(得分:2)
将文件方案从s3n
更改为s3a
后,不再出现错误。