Broken Pipe Error causes streaming Elastic MapReduce job on AWS to fail

Question

Everything works fine locally when I do as follows:

cat input | python mapper.py | sort | python reducer.py

However, when I run the streaming MapReduce job on AWS Elastic Mapreduce, the job does not complete successfully. The mapper.py runs part way through (I know this because of writing to stderr along the way). The mapper is interrupted by a "Broken Pipe" error, which I'm able to retrieve from the syslog of the task attempt after it fails:

java.io.IOException: Broken pipe
    at java.io.FileOutputStream.writeBytes(Native Method)
    at java.io.FileOutputStream.write(FileOutputStream.java:282)
    at java.io.BufferedOutputStream.write(BufferedOutputStream.java:105)
    at java.io.BufferedOutputStream.flushBuffer(BufferedOutputStream.java:65)
    at java.io.BufferedOutputStream.write(BufferedOutputStream.java:109)
    at java.io.DataOutputStream.write(DataOutputStream.java:90)
    at org.apache.hadoop.streaming.io.TextInputWriter.writeUTF8(TextInputWriter.java:72)
    at org.apache.hadoop.streaming.io.TextInputWriter.writeValue(TextInputWriter.java:51)
    at org.apache.hadoop.streaming.PipeMapper.map(PipeMapper.java:109)
    at org.apache.hadoop.mapred.MapRunner.run(MapRunner.java:50)
    at org.apache.hadoop.streaming.PipeMapRunner.run(PipeMapRunner.java:36)
    at org.apache.hadoop.mapred.MapTask.runOldMapper(MapTask.java:441)
    at org.apache.hadoop.mapred.MapTask.run(MapTask.java:377)
    at org.apache.hadoop.mapred.Child$4.run(Child.java:255)
    at java.security.AccessController.doPrivileged(Native Method)
    at javax.security.auth.Subject.doAs(Subject.java:396)
    at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1059)
    at org.apache.hadoop.mapred.Child.main(Child.java:249)


2012-03-26 07:19:05,400 WARN org.apache.hadoop.streaming.PipeMapRed (main): java.io.IOException: Broken pipe
    at java.io.FileOutputStream.writeBytes(Native Method)
    at java.io.FileOutputStream.write(FileOutputStream.java:282)
    at java.io.BufferedOutputStream.write(BufferedOutputStream.java:105)
    at java.io.BufferedOutputStream.flushBuffer(BufferedOutputStream.java:65)
    at java.io.BufferedOutputStream.flush(BufferedOutputStream.java:123)
    at java.io.DataOutputStream.flush(DataOutputStream.java:106)
    at org.apache.hadoop.streaming.PipeMapRed.mapRedFinished(PipeMapRed.java:579)
    at org.apache.hadoop.streaming.PipeMapper.map(PipeMapper.java:124)
    at org.apache.hadoop.mapred.MapRunner.run(MapRunner.java:50)
    at org.apache.hadoop.streaming.PipeMapRunner.run(PipeMapRunner.java:36)
    at org.apache.hadoop.mapred.MapTask.runOldMapper(MapTask.java:441)
    at org.apache.hadoop.mapred.MapTask.run(MapTask.java:377)
    at org.apache.hadoop.mapred.Child$4.run(Child.java:255)
    at java.security.AccessController.doPrivileged(Native Method)
    at javax.security.auth.Subject.doAs(Subject.java:396)
    at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1059)
    at org.apache.hadoop.mapred.Child.main(Child.java:249)

2012-03-26 07:19:05,400 INFO org.apache.hadoop.streaming.PipeMapRed (main): mapRedFinished
2012-03-26 07:19:05,400 WARN org.apache.hadoop.streaming.PipeMapRed (main): java.io.IOException: Bad file descriptor
    at java.io.FileOutputStream.writeBytes(Native Method)
    at java.io.FileOutputStream.write(FileOutputStream.java:282)
    at java.io.BufferedOutputStream.write(BufferedOutputStream.java:105)
    at java.io.BufferedOutputStream.flushBuffer(BufferedOutputStream.java:65)
    at java.io.BufferedOutputStream.flush(BufferedOutputStream.java:123)
    at java.io.DataOutputStream.flush(DataOutputStream.java:106)
    at org.apache.hadoop.streaming.PipeMapRed.mapRedFinished(PipeMapRed.java:579)
    at org.apache.hadoop.streaming.PipeMapper.close(PipeMapper.java:135)
    at org.apache.hadoop.mapred.MapRunner.run(MapRunner.java:57)
    at org.apache.hadoop.streaming.PipeMapRunner.run(PipeMapRunner.java:36)
    at org.apache.hadoop.mapred.MapTask.runOldMapper(MapTask.java:441)
    at org.apache.hadoop.mapred.MapTask.run(MapTask.java:377)
    at org.apache.hadoop.mapred.Child$4.run(Child.java:255)
    at java.security.AccessController.doPrivileged(Native Method)
    at javax.security.auth.Subject.doAs(Subject.java:396)
    at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1059)
    at org.apache.hadoop.mapred.Child.main(Child.java:249)

2012-03-26 07:19:05,400 INFO org.apache.hadoop.streaming.PipeMapRed (main): mapRedFinished
2012-03-26 07:19:05,405 INFO org.apache.hadoop.streaming.PipeMapRed (Thread-13): MRErrorThread done
2012-03-26 07:19:05,408 INFO org.apache.hadoop.mapred.TaskLogsTruncater (main): Initializing logs' truncater with mapRetainSize=-1 and reduceRetainSize=-1
2012-03-26 07:19:05,519 INFO org.apache.hadoop.io.nativeio.NativeIO (main): Initialized cache for UID to User mapping with a cache timeout of 14400 seconds.
2012-03-26 07:19:05,520 INFO org.apache.hadoop.io.nativeio.NativeIO (main): Got UserName hadoop for UID 106 from the native implementation
2012-03-26 07:19:05,522 WARN org.apache.hadoop.mapred.Child (main): Error running child
java.io.IOException: log:null
R/W/S=7018/3/0 in:NA [rec/s] out:NA [rec/s]
minRecWrittenToEnableSkip_=9223372036854775807 LOGNAME=null
HOST=null
USER=hadoop
HADOOP_USER=null
last Hadoop input: |null|
last tool output: |text/html    1|
Date: Mon Mar 26 07:19:05 UTC 2012
java.io.IOException: Broken pipe
    at java.io.FileOutputStream.writeBytes(Native Method)
    at java.io.FileOutputStream.write(FileOutputStream.java:282)
    at java.io.BufferedOutputStream.write(BufferedOutputStream.java:105)
    at java.io.BufferedOutputStream.flushBuffer(BufferedOutputStream.java:65)
    at java.io.BufferedOutputStream.write(BufferedOutputStream.java:109)
    at java.io.DataOutputStream.write(DataOutputStream.java:90)
    at org.apache.hadoop.streaming.io.TextInputWriter.writeUTF8(TextInputWriter.java:72)
    at org.apache.hadoop.streaming.io.TextInputWriter.writeValue(TextInputWriter.java:51)
    at org.apache.hadoop.streaming.PipeMapper.map(PipeMapper.java:109)
    at org.apache.hadoop.mapred.MapRunner.run(MapRunner.java:50)
    at org.apache.hadoop.streaming.PipeMapRunner.run(PipeMapRunner.java:36)
    at org.apache.hadoop.mapred.MapTask.runOldMapper(MapTask.java:441)
    at org.apache.hadoop.mapred.MapTask.run(MapTask.java:377)
    at org.apache.hadoop.mapred.Child$4.run(Child.java:255)
    at java.security.AccessController.doPrivileged(Native Method)
    at javax.security.auth.Subject.doAs(Subject.java:396)
    at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1059)
    at org.apache.hadoop.mapred.Child.main(Child.java:249)


    at org.apache.hadoop.streaming.PipeMapper.map(PipeMapper.java:125)
    at org.apache.hadoop.mapred.MapRunner.run(MapRunner.java:50)
    at org.apache.hadoop.streaming.PipeMapRunner.run(PipeMapRunner.java:36)
    at org.apache.hadoop.mapred.MapTask.runOldMapper(MapTask.java:441)
    at org.apache.hadoop.mapred.MapTask.run(MapTask.java:377)
    at org.apache.hadoop.mapred.Child$4.run(Child.java:255)
    at java.security.AccessController.doPrivileged(Native Method)
    at javax.security.auth.Subject.doAs(Subject.java:396)
    at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1059)
    at org.apache.hadoop.mapred.Child.main(Child.java:249)
2012-03-26 07:19:05,525 INFO org.apache.hadoop.mapred.Task (main): Runnning cleanup for the task
2012-03-26 07:19:05,526 INFO org.apache.hadoop.mapred.DirectFileOutputCommitter (main): Nothing to clean up on abort since there are no temporary files written

Here is mapper.py. Note that I write to stderr to provide myself with debugging info:

#!/usr/bin/env python

import sys
from warc import ARCFile

def main():
    warc_file = ARCFile(fileobj=sys.stdin)
    for web_page in warc_file:
        print >> sys.stderr, '%s\t%s' % (web_page.header.content_type, 1) #For debugging
        print '%s\t%s' % (web_page.header.content_type, 1)
    print >> sys.stderr, 'done' #For debugging
if __name__ == "__main__":
    main()

Here is what I get in the stderr for the task attempt when the mapper.py is run:

text/html   1
text/html   1
text/html   1

Basically, the loop runs through 3 times and then stops abruptly without python throwing any error. (Note: it should be outputting thousands of lines). Even an uncaught exception should appear in stderr.

Because the MapReduce runs completely fine on my local computer, my guess is that this is a problem with how Hadoop is dealing with the output I'm printing from mapper.py. But I'm clueless as to what the problem could be.

Answer 1

This is said in the accepted error, but let me attempt to clarify--you must block on stdin, even if you don't need it! This is not the same as Linux pipes, so don't let that fool you. What happens, intuitively, is, Streaming stands up your executable, then says, "wait here while I go get input for you". If your executable stops for any reason before Streaming sends you 100% of the input, Streaming says, "Hey, where did that executable go that I stood up?...Hmmmm...the pipe is broken, let me raise that exception!" So, here is some python code, all it does is what cat does, but you'll note, this code won't exit until all input is processed, and that is the key point:

#!/usr/bin/python
import sys

while True:
    s = sys.stdin.readline()
    if not s:
        break
    sys.stdout.write(s)