Question

我遇到了在hadoop流中运行简单python代码的问题。我已尝试过以前帖子中的所有建议，但出现类似错误并仍然存在问题。

添加了usr / bin / env python
chmod a + x mapper和reducer python code
put＆＃34;＆＃34; -mapper＆＃34; python mapper.py -n 1 -r 0.4＆＃34;

我已在外面运行代码并且运行良好。

更新：我使用以下代码运行hadoop流之外的代码：

cat file |python mapper.py -n 5 -r 0.4 |sort|python reducer.py -f 3618

这很好..但现在我需要将它运行到HADOOP STREAMING

hadoop jar /usr/lib/hadoop-mapreduce/hadoop-streaming.jar \
-D mapreduce.job.reduces=5  \
-files lr \
-mapper "python lr/mapper.py -n 5 -r 0.4"  \
-reducer "python lr/reducer.py -f 3618"  \
-input training \
-output models

hadoop流媒体是失败的。我查看了日志，但是我没有看到任何告诉我它为什么会发生的事情？

我有以下 mapper.py ：

#!/usr/bin/env python

import sys
import random

from optparse import OptionParser

parser = OptionParser()
parser.add_option("-n", "--model-num", action="store", dest="n_model",
                  help="number of models to train", type="int")
parser.add_option("-r", "--sample-ratio", action="store", dest="ratio",
                  help="ratio to sample for each ensemble", type="float")

options, args = parser.parse_args(sys.argv)

random.seed(8803)
r = options.ratio
for line in sys.stdin:
    # TODO
    # Note: The following lines are only there to help 
    #       you get started (and to have a 'runnable' program). 
    #       You may need to change some or all of the lines below.
    #       Follow the pseudocode given in the PDF.
    key = random.randint(0, options.n_model-1)
    value = line.strip()
    for i in range(1, options.n_model+1):
        m = random.random()
        if m < r:
            print "%d\t%s" % (i, value)

和我的 reducer.py ：

#!/usr/bin/env python
import sys
import pickle
from optparse import OptionParser
from lrsgd import LogisticRegressionSGD
from utils import parse_svm_light_line

parser = OptionParser()
parser.add_option("-e", "--eta", action="store", dest="eta",
                  default=0.01, help="step size", type="float")
parser.add_option("-c", "--Regularization-Constant", action="store", dest="C",
                  default=0.0, help="regularization strength", type="float")
parser.add_option("-f", "--feature-num", action="store", dest="n_feature",
                  help="number of features", type="int")
options, args = parser.parse_args(sys.argv)

classifier = LogisticRegressionSGD(options.eta, options.C, options.n_feature)

for line in sys.stdin:
    key, value = line.split("\t", 1)
    value = value.strip()
    X, y = parse_svm_light_line(value)
    classifier.fit(X, y)

pickle.dump(classifier, sys.stdout)

当我在代码之外运行它时，它运行正常，但是当我在hadoop-streaming中运行它时，它给了我错误：

17/02/07 07:44:34 INFO mapreduce.Job: Task Id : attempt_1486438814591_0038_m_000001_2, Status : FAILED
Error: java.lang.RuntimeException: PipeMapRed.waitOutputThreads(): subprocess failed with code 2
    at org.apache.hadoop.streaming.PipeMapRed.waitOutputThreads(PipeMapRed.java:322)
    at org.apache.hadoop.streaming.PipeMapRed.mapRedFinished(PipeMapRed.java:535)
    at org.apache.hadoop.streaming.PipeMapper.close(PipeMapper.java:130)

Answer 1

在帖子-How to resolve java.lang.RuntimeException: PipeMapRed.waitOutputThreads(): subprocess failed with code 2?

中使用Harishanker的答案

确保使用chmod可执行映射器文件和化简器文件。（例如：“ chmod 744 mapper.py”）

然后像这样执行流式传输命令：

hadoop jar /usr/lib/hadoop-mapreduce/hadoop-streaming.jar \
-D mapreduce.job.reduces=5  \
-files lr \
-mapper lr/mapper.py -n 5 -r 0.4  \
-reducer lr/reducer.py -f 3618  \
-input training \
-output models

现在它应该可以工作了。如果没有，请发表评论。

Python如何解决错误：java.lang.RuntimeException：PipeMapRed.waitOutputThreads（）：子进程失败，代码为2

1 个答案: