在 OS/X 上找不到 Hadoop 原生库

Question

我从 github 下载了hadoop 源代码并使用native 选项编译：

mvn package -Pdist,native -DskipTests -Dtar -Dmaven.javadoc.skip=true

然后我将 .dylib 文件复制到 $HADOOP_HOME/lib

cp -p hadoop-common-project/hadoop-common/target/hadoop-common-2.7.1/lib/native/*.dylib /usr/local/Cellar/hadoop/2.7.2/libexec/share/hadoop/lib

LD_LIBRARY_PATH 已更新，hdfs 已重新启动：

 echo $LD_LIBRARY_PATH
 /usr/local/Cellar/hadoop/2.7.2/libexec/lib:
 /usr/local/Cellar/hadoop/2.7.2/libexec/share/hadoop/common/lib:/Library/Java/JavaVirtualMachines/jdk1.8.0_92.jdk/Contents/Home//jre/lib

（注意：这也意味着Hadoop “Unable to load native-hadoop library for your platform” error on docker-spark?的答案对我不起作用..）

但是checknative还是统一返回false：

$stop-dfs.sh && start-dfs.sh && hadoop checknative
16/06/13 16:12:32 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Stopping namenodes on [sparkbook]
sparkbook: stopping namenode
localhost: stopping datanode
Stopping secondary namenodes [0.0.0.0]
0.0.0.0: stopping secondarynamenode
16/06/13 16:12:50 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
16/06/13 16:12:50 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Starting namenodes on [sparkbook]
sparkbook: starting namenode, logging to /usr/local/Cellar/hadoop/2.7.2/libexec/logs/hadoop-macuser-namenode-sparkbook.out
localhost: starting datanode, logging to /usr/local/Cellar/hadoop/2.7.2/libexec/logs/hadoop-macuser-datanode-sparkbook.out
Starting secondary namenodes [0.0.0.0]
0.0.0.0: starting secondarynamenode, logging to /usr/local/Cellar/hadoop/2.7.2/libexec/logs/hadoop-macuser-secondarynamenode-sparkbook.out
16/06/13 16:13:05 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
16/06/13 16:13:05 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Native library checking:
hadoop:  false
zlib:    false
snappy:  false
lz4:     false
bzip2:   false
openssl: false

Answer 1

要在全新安装的 macOS 10.12 上运行，我必须执行以下操作：

1. 使用homebrew 安装构建依赖项：

   brew install cmake maven openssl protobuf@2.5 snappy
   

2. 查看 hadoop 源代码

   git clone https://github.com/apache/hadoop.git
   cd hadoop
   git checkout rel/release-2.7.3
   

3. 将以下补丁应用于构建：

   diff --git a/hadoop-common-project/hadoop-common/src/CMakeLists.txt b/hadoop-common-project/hadoop-common/src/CMakeLists.txt
   index 942b19c..8b34881 100644
   --- a/hadoop-common-project/hadoop-common/src/CMakeLists.txt
   +++ b/hadoop-common-project/hadoop-common/src/CMakeLists.txt
   @@ -16,6 +16,8 @@
    # limitations under the License.
    #
   
   +SET(CUSTOM_OPENSSL_PREFIX /usr/local/opt/openssl)
   +
    cmake_minimum_required(VERSION 2.6 FATAL_ERROR)
   
    # Default to release builds
   @@ -116,8 +118,8 @@ set(T main/native/src/test/org/apache/hadoop)
    GET_FILENAME_COMPONENT(HADOOP_ZLIB_LIBRARY ${ZLIB_LIBRARIES} NAME)
   
    SET(STORED_CMAKE_FIND_LIBRARY_SUFFIXES ${CMAKE_FIND_LIBRARY_SUFFIXES})
   -set_find_shared_library_version("1")
   -find_package(BZip2 QUIET)
   +set_find_shared_library_version("1.0")
   +find_package(BZip2 REQUIRED)
    if (BZIP2_INCLUDE_DIR AND BZIP2_LIBRARIES)
        GET_FILENAME_COMPONENT(HADOOP_BZIP2_LIBRARY ${BZIP2_LIBRARIES} NAME)
        set(BZIP2_SOURCE_FILES
   diff --git a/hadoop-common-project/hadoop-common/src/main/conf/core-site.xml b/hadoop-common-project/hadoop-common/src/main/conf/core-site.xml
   index d2ddf89..ac8e351 100644
   --- a/hadoop-common-project/hadoop-common/src/main/conf/core-site.xml
   +++ b/hadoop-common-project/hadoop-common/src/main/conf/core-site.xml
   @@ -17,4 +17,8 @@
    <!-- Put site-specific property overrides in this file. -->
   
    <configuration>
   +<property>
   +<name>io.compression.codec.bzip2.library</name>
   +<value>libbz2.dylib</value>
   +</property>
    </configuration>
   diff --git a/hadoop-tools/hadoop-pipes/pom.xml b/hadoop-tools/hadoop-pipes/pom.xml
   index 34c0110..70f23a4 100644
   --- a/hadoop-tools/hadoop-pipes/pom.xml
   +++ b/hadoop-tools/hadoop-pipes/pom.xml
   @@ -52,7 +52,7 @@
                        <mkdir dir="${project.build.directory}/native"/>
                        <exec executable="cmake" dir="${project.build.directory}/native" 
                            failonerror="true">
   -                      <arg line="${basedir}/src/ -DJVM_ARCH_DATA_MODEL=${sun.arch.data.model}"/>
   +                      <arg line="${basedir}/src/ -DJVM_ARCH_DATA_MODEL=${sun.arch.data.model} -DOPENSSL_ROOT_DIR=/usr/local/opt/openssl"/>
                        </exec>
                        <exec executable="make" dir="${project.build.directory}/native" failonerror="true">
                          <arg line="VERBOSE=1"/>
   

4. 从源代码构建 hadoop：

   mvn package -Pdist,native -DskipTests -Dtar -Dmaven.javadoc.skip=true
   

5. 运行hadoop时指定JAVA_LIBRARY_PATH：

   $ JAVA_LIBRARY_PATH=/usr/local/opt/openssl/lib:/opt/local/lib:/usr/lib hadoop-dist/target/hadoop-2.7.3/bin/hadoop checknative -a
   16/10/14 20:16:32 INFO bzip2.Bzip2Factory: Successfully loaded & initialized native-bzip2 library libbz2.dylib
   16/10/14 20:16:32 INFO zlib.ZlibFactory: Successfully loaded & initialized native-zlib library
   Native library checking:
   hadoop:  true /Users/admin/Desktop/hadoop/hadoop-dist/target/hadoop-2.7.3/lib/native/libhadoop.dylib
   zlib:    true /usr/lib/libz.1.dylib
   snappy:  true /usr/local/lib/libsnappy.1.dylib
   lz4:     true revision:99
   bzip2:   true /usr/lib/libbz2.1.0.dylib
   openssl: true /usr/local/opt/openssl/lib/libcrypto.dylib
   

Answer 2

@andrewdotn 在上面的回复中有一些缺失的步骤：

1) 对于步骤 (3)，通过将发布到文本文件的文本添加来创建补丁，例如“patch.txt”，然后执行“git apply patch.txt”

2) 除了按照 javadba 的指示复制文件外，某些应用程序还需要您设置：

export HADOOP_OPTS="-Djava.library.path=${HADOOP_HOME}/lib/native"
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${HADOOP_HOME}/lib/native
export JAVA_LIBRARY_PATH=$JAVA_LIBRARY_PATH:${HADOOP_HOME}/lib/native

Answer 3

所需的步骤是将*.dylib 从git 源构建目录复制到您平台的$HADOOP_HOME/<common dir>lib 目录中。对于通过brew 安装的OS/X，它是：

cp /git/hadoop/hadoop-dist/target/hadoop-2.7.1/lib/native/ /usr/local/Cellar/hadoop/2.7.2/libexec/share/hadoop/common/

我们现在可以在那里看到所需的库：

$ll /usr/local/Cellar/hadoop/2.7.2/libexec/share/hadoop/common/*.dylib
-rwxr-xr-x  1 macuser  staff  149100 Jun 13 13:44 /usr/local/Cellar/hadoop/2.7.2/libexec/share/hadoop/common/libhadoop.dylib
-rwxr-xr-x  1 macuser  staff  149100 Jun 13 13:44 /usr/local/Cellar/hadoop/2.7.2/libexec/share/hadoop/common/libhadoop.1.0.0.dylib

现在hadoop checknative 命令起作用了：

$hadoop checknative

6/06/15 09:10:59 INFO zlib.ZlibFactory: Successfully loaded & initialized native-zlib library
Native library checking:
hadoop:  true /usr/local/Cellar/hadoop/2.7.2/libexec/share/hadoop/common/libhadoop.dylib
zlib:    true /usr/lib/libz.1.dylib
snappy:  false
lz4:     true revision:99
bzip2:   false
openssl: false build does not support openssl.

Answer 4

作为@andrewdotn 答案的更新，这里是用于 Hadoop 2.8.1 的 patch.txt 文件：

diff --git a/hadoop-common-project/hadoop-common/src/CMakeLists.txt b/hadoop-common-project/hadoop-common/src/CMakeLists.txt
index c93bfe78546..e8918f9ca29 100644
--- a/hadoop-common-project/hadoop-common/src/CMakeLists.txt
+++ b/hadoop-common-project/hadoop-common/src/CMakeLists.txt
@@ -20,6 +20,8 @@
 # CMake configuration.
 #

+SET(CUSTOM_OPENSSL_PREFIX /usr/local/opt/openssl)
+
 cmake_minimum_required(VERSION 2.6 FATAL_ERROR)

 list(APPEND CMAKE_MODULE_PATH ${CMAKE_SOURCE_DIR}/..)
@@ -50,8 +52,8 @@ get_filename_component(HADOOP_ZLIB_LIBRARY ${ZLIB_LIBRARIES} NAME)

 # Look for bzip2.
 set(STORED_CMAKE_FIND_LIBRARY_SUFFIXES ${CMAKE_FIND_LIBRARY_SUFFIXES})
-hadoop_set_find_shared_library_version("1")
-find_package(BZip2 QUIET)
+hadoop_set_find_shared_library_version("1.0")
+find_package(BZip2 REQUIRED)
 if(BZIP2_INCLUDE_DIR AND BZIP2_LIBRARIES)
     get_filename_component(HADOOP_BZIP2_LIBRARY ${BZIP2_LIBRARIES} NAME)
     set(BZIP2_SOURCE_FILES
diff --git a/hadoop-common-project/hadoop-common/src/main/conf/core-site.xml b/hadoop-common-project/hadoop-common/src/main/conf/core-site.xml
index d2ddf893e49..ac8e351f1c8 100644
--- a/hadoop-common-project/hadoop-common/src/main/conf/core-site.xml
+++ b/hadoop-common-project/hadoop-common/src/main/conf/core-site.xml
@@ -17,4 +17,8 @@
 <!-- Put site-specific property overrides in this file. -->

 <configuration>
+<property>
+<name>io.compression.codec.bzip2.library</name>
+<value>libbz2.dylib</value>
+</property>
 </configuration>
diff --git a/hadoop-tools/hadoop-pipes/pom.xml b/hadoop-tools/hadoop-pipes/pom.xml
index 8aafad0f7eb..d4832542265 100644
--- a/hadoop-tools/hadoop-pipes/pom.xml
+++ b/hadoop-tools/hadoop-pipes/pom.xml
@@ -55,7 +55,7 @@
                     <mkdir dir="${project.build.directory}/native"/>
                     <exec executable="cmake" dir="${project.build.directory}/native" 
                         failonerror="true">
-                      <arg line="${basedir}/src/ -DJVM_ARCH_DATA_MODEL=${sun.arch.data.model}"/>
+                      <arg line="${basedir}/src/ -DJVM_ARCH_DATA_MODEL=${sun.arch.data.model} -DOPENSSL_ROOT_DIR=/usr/local/opt/openssl"/>
                     </exec>
                     <exec executable="make" dir="${project.build.directory}/native" failonerror="true">
                       <arg line="VERBOSE=1"/>