本文共 11137 字，大约阅读时间需要 37 分钟。
1. 概述
大数据tensorflowonspark 进行安装和测试。
2 .环境
所选操作系统 | 地址和软件版本 | 节点类型 |
Centos7.3 64位 | 192.168.2.31（master） Java：jdk 1.8 Scala:2.10.4 Hadoop:2.7.3 Spark:2.12.3 TensorFlowOnSpark:0.8.0 Python2.7 | Master |
Centos7.3 64位 | 192.168.2.32（spark worker） Java：jdk 1.8 Hadoop:2.7.3 Spark:2.12.3 | slave001 |
Centos7.3 64位 | 192.168.2.33（spark worker） Java：jdk 1.8 Hadoop:2.7.3 Spark:2.12.3 | slave002 |
3 .安装
1.1 删除系统自带jdk：
# rpm -e --nodeps java-1.7.0-openjdk-1.7.0.99-2.6.5.1.el6.x86_64rpm -e --nodeps java-1.6.0-openjdk-1.6.0.38-1.13.10.4.el6.x86_64rpm -e --nodeps tzdata-java-2016c-1.el6.noarch
1.2 安装jdk
rpm -ivh jdk-8u144-linux-x64.rpm
1.3添加java路径
export JAVA_HOME=/usr/java/jdk1.8.0_144
1.4 验证java
[root@master opt]# java -versionjava version "1.8.0_144"Java(TM) SE Runtime Environment (build 1.8.0_144-b01)Java HotSpot(TM) 64-Bit Server VM (build 25.144-b01, mixed mode)
1.5 Ssh免登陆设置
cd /root/.ssh/ssh-keygen -t rsacat id_rsa.pub >> authorized_keys scp id_rsa.pub authorized_keys root@192.168.2.32:/root/.ssh/scp id_rsa.pub authorized_keys root@192.168.2.31:/root/.ssh/
1.6安装python2.7和pip
yum install -y gcc wget https://www.python.org/ftp/python/2.7.13/Python-2.7.13.tgztar vxf Python-2.7.13.tgzcd Python-2.7.13.tgz./configure --prefix=/usr/localmake && make install [root@master opt]# pythonPython 2.7.13 (default, Aug 24 2017, 16:10:35) [GCC 4.4.7 20120313 (Red Hat 4.4.7-18)] on linux2Type "help", "copyright", "credits" or "license" for more information.
1.7 安装pip和setuptools
tar zxvf pip-1.5.4.tar.gztar zxvf setuptools-2.0.tar.gzcd setuptools-2.0python setup.py installcd pip-1.5.4python setup.py install
1.8 Hadoop安装和配置
1.8.1 三台机器都要安装Hadoop
tar zxvf hadoop-2.7.3.tar.gz -C /usr/local/cd /usr/local/hadoop-2.7.3/bin[root@master bin]# ./hadoop versionHadoop 2.7.3Subversion https://git-wip-us.apache.org/repos/asf/hadoop.git -r baa91f7c6bc9cb92be5982de4719c1c8af91ccffCompiled by root on 2016-08-18T01:41ZCompiled with protoc 2.5.0From source with checksum 2e4ce5f957ea4db193bce3734ff29ff4This command was run using /usr/local/hadoop-2.7.3/share/hadoop/common/hadoop-common-2.7.3.jar
1.8.2 配置hadoop
配置mastervi /usr/local/hadoop-2.7.3/etc/hadoop/core-site.xmlhadoop.tmp.dir file:/usr/local/hadoop/tmp Abase for other temporary directories. fs.defaultFS hdfs://master:9001
配置slave
[root@slave001 hadoop-2.7.3]# vi ./etc/hadoop/core-site.xmlhadoop.tmp.dir file:/usr/local/hadoop/tmp Abase for other temporary directories. fs.defaultFS hdfs://slave001:9001
[root@slave002 hadoop-2.7.3]# vi ./etc/hadoop/core-site.xmlhadoop.tmp.dir file:/usr/local/hadoop/tmp Abase for other temporary directories. fs.defaultFS hdfs://slave002:9001
1.8.3 配置hdfs
vi /usr/local/hadoop-2.7.3/etc/hadoop/hdfs-site.xmldfs.replication 1 dfs.namenode.name.dir file:/usr/local/hadoop/tmp/dfs/name dfs.datanode.data.dir file:/usr/local/hadoop/tmp/dfs/data dfs.namenode.rpc-address master:9001
1.9 安装scala
tar -zxvf scala-2.12.3.tgz -C /usr/local/ #修改变量添加scalavi /etc/profileexport SCALA_HOME=/usr/local/scala-2.12.3/export PATH=$PATH:/usr/local/scala-2.12.3/binsource /etc/profile
2.0三台机器都要安装spark
tar -zxvf spark-2.1.1-bin-hadoop2.7.tgz -C /usr/local/ vi /etc/profileexport JAVA_HOME=/usr/java/jdk1.8.0_144/export SCALA_HOME=/usr/local/scala-2.12.3/export PATH=$PATH:/usr/local/scala-2.12.3/binexport SPARK_HOME=/usr/local/spark-2.1.1-bin-hadoop2.7/export PATH=$PATH:$SPARK_HOME/bin:$SPARK_HOME/sbinsource /etc/profile
修改spark配置
cd /usr/local/spark-2.1.1-bin-hadoop2.7/
vi ./conf/spark-env.sh.template
export JAVA_HOME=/usr/java/jdk1.8.0_144/
export SCALA_HOME=/usr/local/scala-2.12.3/
#export SPARK_HOME=/usr/local/spark-2.1.1-bin-hadoop2.7/
export SPARK_MASTER_IP=192.168.2.31
export SPARK_WORKER_MEMORY=1g
export HADOOP_CONF_DIR=/usr/local/hadoop-2.7.3/etc/hadoop
export HADOOP_HDFS_HOME=/usr/local/hadoop-2.7.3/
export SPARK_DRIVER_MEMORY=1g
保存退出
mv spark-env.sh.template spark-env.sh
#修改slaves
[root@master conf]# vi slaves.template
192.168.2.32
192.168.2.33
[root@master conf]# mv slaves.template slaves
2.1 三台主机上修改hosts
vi /etc/hosts
192.168.2.31 master
192.168.2.32 slave001
192.168.2.33 slave002
4. 启动服务
[root@master local]# cd hadoop-2.7.3/sbin/
修改配置文件vi /usr/local/hadoop-2.7.3/etc/hadoop/hadoop-env.sh
export JAVA_HOME=/usr/java/jdk1.8.0_144/
./start-all.sh
localhost: Warning: Permanently added 'localhost' (RSA) to the list of known hosts.
localhost: Error: JAVA_HOME is not set and could not be found.
修改配置文件
vi /usr/local/hadoop-2.7.3/etc/hadoop/hadoop-env.sh
export JAVA_HOME=/usr/java/jdk1.8.0_144/
重新启动服务
sbin/start-all.sh
#启动spark
cd /usr/local/spark-2.1.1-bin-hadoop2.7/sbin/
./start-all.sh
4. 安装tensorflow
前提下先安装cudavim /etc/yum.repos.d/linuxtech.testing.repo 添加内容：[cpp] view plain copy[linuxtech-testing] name=LinuxTECH Testing baseurl=http://pkgrepo.linuxtech.net/el6/testing/ enabled=0 gpgcheck=1 gpgkey=http://pkgrepo.linuxtech.net/el6/release/RPM-GPG-KEY-LinuxTECH.NET sudo rpm -i cuda-repo-rhel6-8.0.61-1.x86_64.rpmsudo yum clean allsudo yum install cudarpm -ivh --nodeps dkms-2.1.1.2-1.el6.rf.noarch.rpm yum install cudayum install epel-releaseyum install -y zlib* #软连接cudaln -s /usr/local/cuda-8.0 /usr/local/cudaldconfig /usr/local/cuda/lib64Vi /etc/profileexport LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64"export CUDA_HOME=/usr/local/cuda
更新pippip install --upgrade pip下载tensorflowpip install --upgrade https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.8.0-cp27-none-linux_x86_64.whl安装好后#python>>> import tensorflowTraceback (most recent call last): File "", line 1, in File "/usr/local/lib/python2.7/site-packages/tensorflow/__init__.py", line 23, in from tensorflow.python import * File "/usr/local/lib/python2.7/site-packages/tensorflow/python/__init__.py", line 45, in from tensorflow.python import pywrap_tensorflow File "/usr/local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 28, in _pywrap_tensorflow = swig_import_helper() File "/usr/local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 24, in swig_import_helper _mod = imp.load_module('_pywrap_tensorflow', fp, pathname, description)ImportError: libcudart.so.7.5: cannot open shared object file: No such file or directory
#这是因为lib库不完整yum install openssl -yyum install openssl-devel -yyum install gcc gcc-c++ gcc*#更新pip install --upgrade pippip install --upgrade https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.8.0-cp27-none-linux_x86_64.whl
>>> import tensorflowTraceback (most recent call last): File "", line 1, in File "/usr/local/lib/python2.7/site-packages/tensorflow/__init__.py", line 23, in from tensorflow.python import * File "/usr/local/lib/python2.7/site-packages/tensorflow/python/__init__.py", line 45, in from tensorflow.python import pywrap_tensorflow File "/usr/local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 28, in _pywrap_tensorflow = swig_import_helper() File "/usr/local/lib/python2.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 24, in swig_import_helper _mod = imp.load_module('_pywrap_tensorflow', fp, pathname, description)ImportError: /lib64/libc.so.6: version `GLIBC_2.15' not found (required by /usr/local/lib/python2.7/site-packages/tensorflow/python/_pywrap_tensorflow.so)#这是因为tensorflow 使用的glibc版本库太高，系统自带太低了。可以使用。
# strings /usr/lib64/libstdc++.so.6 | grep GLIBCXX
GLIBCXX_3.4
GLIBCXX_3.4.1
GLIBCXX_3.4.2
GLIBCXX_3.4.3
GLIBCXX_3.4.4
GLIBCXX_3.4.5
GLIBCXX_3.4.6
GLIBCXX_3.4.7
GLIBCXX_3.4.8
GLIBCXX_3.4.9
GLIBCXX_3.4.10
GLIBCXX_3.4.11
GLIBCXX_3.4.12
GLIBCXX_3.4.13
GLIBCXX_FORCE_NEW
GLIBCXX_DEBUG_MESSAGE_LENGTH
放入最新的glibc库，解压出6.0.20
libstdc++.so.6.0.20 覆盖原来的libstdc++.so.6
[root@master 4.4.7]# ln -s /opt/libstdc++.so.6/libstdc++.so.6.0.20 /usr/lib64/libstdc++.so.6
ln: creating symbolic link `/usr/lib64/libstdc++.so.6': File exists
[root@master 4.4.7]# mv /usr/lib64/libstdc++.so.6 /root/
[root@master 4.4.7]# ln -s /opt/libstdc++.so.6/libstdc++.so.6.0.20 /usr/lib64/libstdc++.so.6
[root@master 4.4.7]# strings /usr/lib64/libstdc++.so.6 | grep GLIBCXX
[root@master ~]# strings /usr/lib64/libstdc++.so.6 | grep GLIBCXX
GLIBCXX_3.4
GLIBCXX_3.4.1
GLIBCXX_3.4.2
GLIBCXX_3.4.3
GLIBCXX_3.4.4
GLIBCXX_3.4.5
GLIBCXX_3.4.6
GLIBCXX_3.4.7
GLIBCXX_3.4.8
GLIBCXX_3.4.9
GLIBCXX_3.4.10
GLIBCXX_3.4.11
GLIBCXX_3.4.12
GLIBCXX_3.4.13
GLIBCXX_3.4.14
GLIBCXX_3.4.15
GLIBCXX_3.4.16
GLIBCXX_3.4.17
GLIBCXX_3.4.18
GLIBCXX_3.4.19
GLIBCXX_3.4.20
GLIBCXX_DEBUG_MESSAGE_LENGTH
这个地方特别要注意坑特别多，一定要覆盖原来的。
pip install tensorflowonspark
这样就可以使用了
报错信息：
报错：ImportError: /lib64/libc.so.6: version `GLIBC_2.17' not found (required by /usr/local/lib/python2.7/site-packages/tensorflow/python/_pywrap_tensorflow.so)
tar zxvf glibc-2.17.tar.gz
mkdir build
cd build
../glibc-2.17/configure --prefix=/usr --disable-profile --enable-add-ons --with-headers=/usr/include --with-binutils=/usr/bin
make -j4
make install
测试验证tensorflow
import tensorflow as tfimport numpy as npx_data = np.float32(np.random.rand(2, 100)) y_data = np.dot([0.100, 0.200], x_data) + 0.300 b = tf.Variable(tf.zeros([1]))W = tf.Variable(tf.random_uniform([1, 2], -1.0, 1.0))y = tf.matmul(W, x_data) + b loss = tf.reduce_mean(tf.square(y - y_data))optimizer = tf.train.GradientDescentOptimizer(0.5)train = optimizer.minimize(loss) init = tf.initialize_all_variables() sess = tf.Session()sess.run(init) for step in xrange(0, 201): sess.run(train) if step % 20 == 0: print step, sess.run(W), sess.run(b) # 得到最佳拟合结果 W: [[0.100 0.200]], b: [0.300]
确保etc/profileexport JAVA_HOME=/usr/java/jdk1.8.0_144/export SCALA_HOME=/usr/local/scala-2.12.3/export PATH=$PATH:/usr/local/scala-2.12.3/binexport SPARK_HOME=/usr/local/spark-2.1.1-bin-hadoop2.7/export PATH=$PATH:$SPARK_HOME/bin:$SPARK_HOME/sbinexport LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64"export CUDA_HOME=/usr/local/cudaexport PYTHONPATH=$SPARK_HOME/python:$SPARK_HOME/python/lib/py4j-0.10.4-src.zip:$PYTHONPATH
完成实验。
下载地址：
转载地址：http://lfbdo.baihongyu.com/