[SPARK-21570] File __spark_libs__XXX.zip does not exist on networked file system w/ yarn

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Not A Problem
• Affects Version/s: 2.2.0
• Component/s: Spark Core, YARN
• Assignee: Unassigned
• Reporter: Albert Chu
• Created: 29/Jul/17 00:13
• Updated: 17/May/20 18:14
• Resolved: 04/Aug/17 09:00

Description

I have a set of scripts that run Spark with data in a networked file system. One of my unit tests to make sure things don't break between Spark releases is to simply run a word count (via org.apache.spark.examples.JavaWordCount) on a file in the networked file system. This test broke with Spark 2.2.0 when I use yarn to launch the job (using the spark standalone scheduler things still work). I'm currently using Hadoop 2.7.0. I get the following error:

Diagnostics: File file:/p/lcratery/achu/testing/rawnetworkfs/test/1181015/node-0/spark/node-0/spark-292938be-7ae3-460f-aca7-294083ebb790/__spark_libs__695301535722158702.zip does not exist
java.io.FileNotFoundException: File file:/p/lcratery/achu/testing/rawnetworkfs/test/1181015/node-0/spark/node-0/spark-292938be-7ae3-460f-aca7-294083ebb790/__spark_libs__695301535722158702.zip does not exist
	at org.apache.hadoop.fs.RawLocalFileSystem.deprecatedGetFileStatus(RawLocalFileSystem.java:606)
	at org.apache.hadoop.fs.RawLocalFileSystem.getFileLinkStatusInternal(RawLocalFileSystem.java:819)
	at org.apache.hadoop.fs.RawLocalFileSystem.getFileStatus(RawLocalFileSystem.java:596)
	at org.apache.hadoop.fs.FilterFileSystem.getFileStatus(FilterFileSystem.java:421)
	at org.apache.hadoop.yarn.util.FSDownload.copy(FSDownload.java:253)
	at org.apache.hadoop.yarn.util.FSDownload.access$000(FSDownload.java:63)
	at org.apache.hadoop.yarn.util.FSDownload$2.run(FSDownload.java:361)
	at org.apache.hadoop.yarn.util.FSDownload$2.run(FSDownload.java:359)
	at java.security.AccessController.doPrivileged(Native Method)
	at javax.security.auth.Subject.doAs(Subject.java:422)
	at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1657)
	at org.apache.hadoop.yarn.util.FSDownload.call(FSDownload.java:358)
	at org.apache.hadoop.yarn.util.FSDownload.call(FSDownload.java:62)
	at java.util.concurrent.FutureTask.run(FutureTask.java:266)
	at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
	at java.util.concurrent.FutureTask.run(FutureTask.java:266)
	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
	at java.lang.Thread.run(Thread.java:748)

While debugging, I sat and watched the directory and did see that /p/lcratery/achu/testing/rawnetworkfs/test/1181015/node-0/spark/node-0/spark-292938be-7ae3-460f-aca7-294083ebb790/_spark_libs_695301535722158702.zip does show up at some point.

Wondering if it's possible something racy was introduced. Nothing in the Spark 2.2.0 release notes suggests any type of configuration change that needs to be done.

Thanks