#失败的映射任务超过了允许的限制

Question

我正在尝试使用Python开发Hadoop流。我通过从here获得帮助，编写了简单的map和reduce脚本

map脚本如下：

#!/usr/bin/env python

import sys, urllib, re

title_re = re.compile("<title>(.*?)</title>", re.MULTILINE | re.DOTALL | re.IGNORECASE)

for line in sys.stdin:
    url = line.strip()
    match = title_re.search(urllib.urlopen(url).read())
    if match :
        print url, "\t", match.group(1).strip()

而reduce脚本如下：

#!/usr/bin/env python

from operator import itemgetter
import sys

for line in sys.stdin :
    line = line.strip()
    print line

在使用hadoop streaming jar运行这些脚本之后，map任务将完成，我可以看到它们已经100%完成，但是reduce作业将停留在22%的位置，并且经过很长一段时间后，它将出现ERROR streaming.StreamJob: Job not successful. Error: # of failed Map Tasks exceeded allowed limit. FailedCount: 1.错误。

我无法找出这背后的确切原因。

我的终端窗口如下所示：

shekhar@ubuntu:/host/Shekhar/Softwares/hadoop-1.0.0$ hadoop jar contrib/streaming/hadoop-streaming-1.0.0.jar -mapper /host/Shekhar/HadoopWorld/MultiFetch.py -reducer /host/Shekhar/HadoopWorld/reducer.py -input /host/Shekhar/HadoopWorld/urls/* -output /host/Shekhar/HadoopWorld/titles3
Warning: $HADOOP_HOME is deprecated.

packageJobJar: [/tmp/hadoop-shekhar/hadoop-unjar2709939812732871143/] [] /tmp/streamjob1176812134999992997.jar tmpDir=null
12/05/27 11:27:46 INFO util.NativeCodeLoader: Loaded the native-hadoop library
12/05/27 11:27:46 INFO mapred.FileInputFormat: Total input paths to process : 3
12/05/27 11:27:46 INFO streaming.StreamJob: getLocalDirs(): [/tmp/hadoop-shekhar/mapred/local]
12/05/27 11:27:46 INFO streaming.StreamJob: Running job: job_201205271050_0006
12/05/27 11:27:46 INFO streaming.StreamJob: To kill this job, run:
12/05/27 11:27:46 INFO streaming.StreamJob: /host/Shekhar/Softwares/hadoop-1.0.0/libexec/../bin/hadoop job  -Dmapred.job.tracker=localhost:9001 -kill job_201205271050_0006
12/05/27 11:27:46 INFO streaming.StreamJob: Tracking URL: http://localhost:50030/jobdetails.jsp?jobid=job_201205271050_0006
12/05/27 11:27:47 INFO streaming.StreamJob:  map 0%  reduce 0%
12/05/27 11:28:07 INFO streaming.StreamJob:  map 67%  reduce 0%
12/05/27 11:28:37 INFO streaming.StreamJob:  map 100%  reduce 0%
12/05/27 11:28:40 INFO streaming.StreamJob:  map 100%  reduce 11%
12/05/27 11:28:49 INFO streaming.StreamJob:  map 100%  reduce 22%
12/05/27 11:31:35 INFO streaming.StreamJob:  map 67%  reduce 22%
12/05/27 11:31:44 INFO streaming.StreamJob:  map 100%  reduce 22%
12/05/27 11:34:52 INFO streaming.StreamJob:  map 67%  reduce 22%
12/05/27 11:35:01 INFO streaming.StreamJob:  map 100%  reduce 22%
12/05/27 11:38:11 INFO streaming.StreamJob:  map 67%  reduce 22%
12/05/27 11:38:20 INFO streaming.StreamJob:  map 100%  reduce 22%
12/05/27 11:41:29 INFO streaming.StreamJob:  map 67%  reduce 22%
12/05/27 11:41:35 INFO streaming.StreamJob:  map 100%  reduce 100%
12/05/27 11:41:35 INFO streaming.StreamJob: To kill this job, run:
12/05/27 11:41:35 INFO streaming.StreamJob: /host/Shekhar/Softwares/hadoop-1.0.0/libexec/../bin/hadoop job  -Dmapred.job.tracker=localhost:9001 -kill job_201205271050_0006
12/05/27 11:41:35 INFO streaming.StreamJob: Tracking URL: http://localhost:50030/jobdetails.jsp?jobid=job_201205271050_0006
12/05/27 11:41:35 ERROR streaming.StreamJob: Job not successful. Error: # of failed Map Tasks exceeded allowed limit. FailedCount: 1. LastFailedTask: task_201205271050_0006_m_000001
12/05/27 11:41:35 INFO streaming.StreamJob: killJob...
Streaming Job Failed!

有人能帮我吗？？

编辑 工作跟踪器详细信息如下：

Hadoop job_201205271050_0006 on localhost

User: shekhar
Job Name: streamjob1176812134999992997.jar
Job File: file:/tmp/hadoop-shekhar/mapred/staging/shekhar/.staging/job_201205271050_0006/job.xml
Submit Host: ubuntu
Submit Host Address: 127.0.1.1
Job-ACLs: All users are allowed
Job Setup: Successful
Status: Failed
Failure Info:# of failed Map Tasks exceeded allowed limit. FailedCount: 1. LastFailedTask: task_201205271050_0006_m_000001
Started at: Sun May 27 11:27:46 IST 2012
Failed at: Sun May 27 11:41:35 IST 2012
Failed in: 13mins, 48sec
Job Cleanup: Successful
Black-listed TaskTrackers: 1
Kind    % Complete  Num Tasks   Pending Running Complete    Killed  Failed/Killed
Task Attempts
map 100.00%
3   0   0   2   1   4 / 0
reduce  100.00%
1   0   0   0   1   0 / 1

Answer 1

先检查一下你的标准。 您的信息不足以决定它是什么错误，stderr通常在： {your hadoop temp dir here}/mapred/local/userlogs/{your job id}/{your attemp id}/stderr

Sean的答案是第一次使用hadoop时的大多数情况，所以我猜您可能会得到一个“env:python\r:No such file or directory”错误。如果是的话，只要把你的CR换成LF就可以解决这个问题。只需运行一个脚本将其替换为

Answer 2

此错误只是一个常规错误，太多映射任务未能完成：

of failed Map Tasks exceeded allowed limit

您可以使用EMR控制台导航到各个Map/Reduce任务的日志。 那么你应该能看到问题所在。

在我的例子中，当我犯了一些小错误时，我就犯了这个错误，比如错误地设置了映射脚本的路径。

查看任务日志的步骤：

http://antipatterns.blogspot.nl/2013/03/amazon-emr-map-reduce-error-of-failed.html

Answer 3

我刚才也出现了同样的错误。在我的例子中，结果是一个解析错误。有一个“意外”的新线在一些地方，stdin分裂线。我建议检查一下你的数据文件。一旦我移除了有这些新行的列，它就工作得很好了。