Tags:
create new tag
view all tags
-- ZhuWei - 2011-03-08

gos安装操作步骤

1、 OS:SLI45,用quattor装的,跟计算节点一样。

2、 安装环境准备,需要mysql(4.1.12及以上),ant(1.6.5及以上),ftp,pbs(以下安装的是pbs客户端),C库(gcc,yacc,autoconf,make,dos2unix),证书。除此之外,由于我们需要用自己的环境提交作业,故需要afs和/home/cc目录。

3、 用afs帐号登录,获取安装包和证书,目录为/home/cc/zwei/GOS_4_0_linux_20100713.tar.gz。

4、 解压安装包并重命名为GOS

[ihepgos02] /home/cc/zwei> tar zxvf gos.tar.gz

5、 [ihepgos02] /home/cc/zwei> cd GOS/installBase/

[ihepgos02] /home/cc/zwei> vi gosconf.properties

6、 修改配置文件如下,我们安装为前后段一体即all模式,其中注意db_host应为127.0.0.1,certsBaseDir为证书存放的位置,默认安装8080是关闭的,需要在/etc/sysconfig/iptables里面打开8080端口。

# these lines begin with '#' are comments.

# these configs are 'key = value' pair. commented pair will prompt to user to input value when install.

# You can change the value as your wish.

# the name of the site, it will be used to identify the site when multiple sites form a grid.

# It can be not equal to the hostname.

gos_site_name=ihepgos02

# host information

# The IP address of you machine

gos_host=202.122.33.112

# the port that GOS Core will listen.

gos_port=8080

# the port that GOS system will listen.

gos_system_port=8080

# You machine's hostname. It should never contains '-' and other special characters.

gos_hostname=ihepgos02.ihep.ac.cn

# front machine host information

# The IP address of front machine

gos_front_host=202.122.33.112

# the GOS Core port of front machine that GOS will listen.

gos_front_port=8080

# the GOS System port of front machine that GOS will listen.

gos_front_system_port=8080

# front machine's hostname. It should never contains '-' and other special characters.

gos_front_hostname=ihepgos02.ihep.ac.cn

# database infomation

# the database user for GOS

db_user=gos

# set a new password for database user.

db_pass=gos

# database root, no need to change

db_super=root

# please input the database root's current password.

#db_superpass=

# database host IP address

db_host=127.0.0.1

# the port that the database is listening.

db_port=3306

# whether you want to backup database.

opt_backupdb=false

# whether you want to create new database data for GOS.

opt_createdb=true

# set grid admin infomation

# please input an email as grid admin's user name.

gridAdminUsername=zhuw@ihep.ac.cn

# set a new password for Grid Admin.

#gridAdminPassword=GridAdminPassword

# set default user infomation, Default user is an open user for public use.

#please input an email as default user name.

defaultUsername=andyzhui@sina.com

# set a new password for default user.

defaultUserPassword=

# please set the default agora's name. It's the first part of the agora name, It must not contain ':'.

defaultAgoraName=Agora_ihepgos

#set the certs base direcotry path, including certs directory name

certsBaseDir=/home/cc/zwei/certfromgos02

7、 如果需要多节点互联,需要保证各个gos的管理员用户名和默认用户名不能雷同。还需要保证certsBaseDir所包含的证书不能雷同。gos_hostname的配置项的值必须是运行hostname后得到的。

8、 安装。目前HPCG仅支持命令行安装模式。

[ihepgos02] /home/cc/zwei/GOS/installBase > pwd

/home/cc/zwei/GOS/installBase

[ihepgos02] /home/cc/zwei/GOS/installBase > ./install.sh all

9、 在运行该命令之后,后续将执行验证/文件拷贝/数据库初始化/设置相关密码等操作。

启动gos,特别注意,不能用root来启动gos服务。

[ihepgos02] /home/cc/zwei/GOS > pwd

/home/cc/zwei/GOS

[ihepgos02] /home/cc/zwei/GOS > ./gos.sh start

10、检查

a、通过ps -Af | grep java检查进程状态;

b、查看portal看服务是否启动,

http://202.122.33.162:8080/axis/services

http://202.122.33.162:8080/batchService/services

http://202.122.33.162:8080/hpcgapp/

http://202.122.33.162:8080/hpcgmgmt/

c、testInstall 检查gos的核心能不能正常工作

[ihepgos02] /home/cc/zwei/GOS/all/gosShell/binary > pwd

/home/cc/zwei/GOS/all/gosShell/binary

[ihepgos02] /home/cc/zwei/GOS/all/gosShell/binary > mv gsh gsh_1

[ihepgos02] /home/cc/zwei/GOS/all/gosShell/binary > ./grun.sh -u zhuw@ihep.ac.cn -p 密码

log in successfully

<< Bash with Grid Extension >>

(using C grip APIs)

Hi zhuw@ihep.ac.cn! Welcome to the agora Agora_ihep01.

Use 'grun' tag before a command to run it as a grip.

Enter 'ghelp' to show all GOS commands.

[gos: binary]testInsatll

........................

........................

OK (38 tests)

[gos: binary]

注意:核心测试是否成功关键看最后是否出现OK(38 tests)字样,测试中屏幕输出会有很多异常信息,一般后面都有紧跟着exception occurred as needed,这个是核心测试故意引入的错误操作,用以检验核心对各种异常情况的处理是否正确。

11、作业配置

12、配置sudo

用root用户登录。运行[root@gos01 ~]# visudo,添加如下行:

zwei ALL=(ALL) NOPASSWD: /bin/cat

zwei ALL=(ALL) NOPASSWD: /bin/chmod

zwei ALL=(ALL) NOPASSWD: /bin/sed

zwei ALL= (ALL) NOPASSWD: /home/cc/zwei/GOS/all/work/batch/bin/pbs_driver_function_withoutTracejob.sh *

zwei ALL= (ALL) NOPASSWD: /home/cc/zwei/GOS/all/work/batch/bin/file_driver_function.sh *

zwei ALL= (ALL) NOPASSWD: /home/cc/zwei/GOS/all/work/batch/bin/fms_driver_function.sh *

切换到zwei用户下,用命令

[ihepgos02] /home/cc/zwei> sudo -l

若出现刚加入的信息,则说明配置正确。

13、作业配置文件的检查

在[ihepgos02] /home/cc/zwei/GOS/all/work/batch/bin > pwd此路径下,

[ihepgos02] /home/cc/zwei/GOS/all/work/batch/bin > vi pbs_driver_function.sh 检查qsub路径是否与当前系统的qsub路径一致。

当PBS server不在本节点时(也即如果gos这个机器只有pbs客户端时),需要改的配置 :在/home/cc/zwei/GOS/all/work/batch/bin路径下

a、pbs_driver_function.sh中改一个配置项

把PARA_DRIVER_NAME="pbs_driver_function.sh"中的sh文件改成:pbs_driver_function_withoutTracejob.sh

b、修改sudo中,将pbs_driver_function.sh改成pbs_driver_function_withoutTracejob.sh

c、修改pbs_driver_function_withoutTracejob.sh中qsub,qstat,qdel目录

14、重启gos [ihepgos02] /home/cc/zwei/GOS > ./gos.sh restart

Edit | Attach | Watch | Print version | History: r3 < r2 < r1 | Backlinks | Raw View | Raw edit | More topic actions
Topic revision: r3 - 2011-03-08 - ZhuWei
 
This site is powered by the TWiki collaboration platform Powered by PerlCopyright © 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback