KFS构建云存储之：KFS编译安装1

环境：Centos 6.2 x64 bit，最小安装。

基础工具安装：
yum install make
yum install wget


1、安装KFS相关依赖包


yum -y install gcc gcc-c++


yum -y install log4cpp


yum -y install cmake


yum -y install boost boost-devel


yum -y install ant


yum -y install openssh*
yum -y install openssl*


yum -y install fuse


yum -y python-devel


yum -y install gettext
yum -y install uuid-devel


源码安装e2xfsprogs,xfsprogs
--yum -y install xfsprogs xfsprogs-devel
--yum -y install e2fsprogs e2fsprogs-devel


源码安装e2fsprogs-1.41.12.tar.gz  目的将uuid库放到/usr/include/文件夹下
tar xzvf e2fsprogs-1.41.12.tar.gz
cd e2fsprogs-1.41.12
./configure
make
make install 


cd /lib/uuid
make install




源码安装xfsprogs-3.1.2.tar.gz
tar xzvf xfsprogs-3.1.2.tar.gz
cd xfsxfsprogs-3.1.2
./configure
make
make install
makde install-dev






2、安装kfs-0.5.tar.gz
将kfs-0.5.tar.gz解压到根目录/


cd /kfs-0.5


vi CMakeLists.txt


:set nu  显示行号  在第110行：set (CMAKE_BUILD_TYPE "Debug")后面插入面内容：


SET(JAVA_INCLUDE_PATH "/usr/lib/jvm/java-1.6.0-openjdk-1.6.0.0.x86_64/include")
SET(JAVA_INCLUDE_PATH2 "/usr/lib/jvm/java-1.6.0-openjdk-1.6.0.0.x86_64/include/linux")


保存编辑CMakeLists.txt


mkdir build
cd build
cmake -D CMAKE_BUILD_TYPE=RelWithDebInfo /kfs-0.5/
make
make install




编译Java部分


返回一上层目录，编译生成jar包


cd ..
ant jar


如果build目录下生成了kfs-0.5.jar和classes文件夹说明成功。




安装Pathon扩展
cd /kfs-0.5/src/cc/access
vi kfs_setup.py


将：


kfsext = Extension('kfs',
                include_dirs = ['/home/srao/p4trees/kfssort/kosmosfs/src/cc/', 


'/usr/local/include/boost-1_37/'],
                libraries = ['kfsClient'],
                library_dirs = [kfs_lib_dir],
        runtime_library_dirs = ['/home/qmr_ksort/sortmaster/lib'],
                sources = ['KfsModulePy.cc'])


修改为：


kfsext = Extension('kfs',
                include_dirs ['/kfs-0.5/src/cc/', '/usr/local/include/boost/'],
                libraries = ['kfsClient'],
                library_dirs = [kfs_lib_dir],
        runtime_library_dirs = ['/kfs-0.5/sortmaster/lib'],
                sources = ['KfsModulePy.cc'])




其中：/usr/local/include/boost/为boost的实际安装目录


保存kfs_setup.py


python kfs_setup.py /kfs-0.5/build/lib/ build


在目录/kfs-0.5/src/cc/access/build/lib.linux-x86_64-2.6/下生成了共享链接：kfs.so，表示成功






配置kfs


cd /kfs-0.5/scripts
vi machines.cfg


先拿本机做试验，本机即做metaServer也是chunkServer，整个集群只有一个node
# KFS Machine configuration file
[metaserver]
node: localhost
clusterkey:kfs-test-cluster
rundir: /kfsRun/meta
baseport: 20000
[chunkserver1]
node: localhost
rundir: /kfsRun/chunk1
baseport: 30000
space: 500 M
[chunkserver2]
node: localhost
rundir: /kfsRun/chunk2
baseport: 40000
space: 500 M
[chunkserver3]
node: localhost
rundir: /kfsRun/chunk3
baseport: 50000
space: 500 M


保存退出




修改ChunkServer.prp配置


cd /kfs-0.5/conf


vi ChunkServer.prp


# Configuration for chunk server:
# meta server location
# hostname改主机名，域名解析
chunkServer.metaServer.hostname = localhost
chunkServer.metaServer.port = 30000
# port to open for client connections
chunkServer.clientPort = 22000
# Directory for storing the chunks
chunkServer.chunkDir = ./chunks
chunkServer.logDir = ./logs
# provide 300G of storage space
# 改为3G
chunkServer.totalSpace = 3000






保存退出,进入/kfs-0.5/scripts目录，安装 metaserver 和 chunkserver：
python kfssetup.py -f machines.cfg  -b ../build -w ../webui -s