MS6 linux 安装备忘

SLES10 服务器上安装MS6.

安装小木虫上安装

http://emuch.net/bbs/viewthread.php?tid=4702095

接着安装torque（其他都是浮云，官方网站才是正道，以后切记了）

如果系统比较旧的 还是选低点版本如SLES10 里使用torque 3.0.6 就可以了，最新的4.1.2 反而在configure是提示缺少limx2.dev 库，安装起来会比较烦

http://www.clusterresources.com/torquedocs/1.1installation.shtml

其中要注意的问题是里面提到的提交路径：After installation, verify you have the PATH environment variable configured to include/usr/local/bin/ and/usr/local/sbin/.

如果提示Torque /usr/local/sbin/pbs_mom: error while loading shared libraries: libtorque.so.2

参考http://blog.csdn.net/kongxx/article/details/8096032

这个问题原因是没有将/usr/local/lib加入动态连接库，缺少的这个文件/usr/local/lib下，因此此时需要使用root用户运行下面的命令

1. # echo '/usr/local/lib' > /etc/ld.so.conf.d/torque.conf  
2. # ldconfig  

然后再运行命令启动Torque服务就不会出现上面的错误了。

一般linux系统的path里会含有/usr/local/bin 而不会包含 /usr/local/sbin

如果明明设置了nodes 而是要pbsnodes 总提示

pbsnodes: Server has no node list MSG=node list is empty - check 'server_priv/ndes' file
就是因为/usr/local/sbin不在系统路径下。一下午啊就折在这上了
知道就好办，在etc/profile 里添加：export PATH=/usr/local/sbin:$PATH
source /etc/profile

要启动pbs：
管理员权限 使用torque安装目录下的torque.setup <user>  添加队列
模式就是batch

结束qterm 结束pbs_server
接着重启pbs_server

sudo pbs_mom
sudo pbs_sched
sudo pbs_server

在路径MSROOT]/etc/Gateway/root_default/dsd/commands/queues/TORQUE

下 修改torque.cfg里的

# Add any additional parameters to qsub command.
#
 params = -l nodes=node1
#
# For example if you want to get notification in e-mail when your job finishes,
# this line may look like:
# params = -m e -M someuser@somedomain.com
# For the list of qsub parameters please refer to the Torque documentation.
# If Materials Studio is installed on a Linux cluster, the job may run on any node
# belonging to the cluster. For all other Linux configurations, you must insure that the
# job is executed on the same node where Materials Studio is installed.
#
# Hence, if your PBS Linux configuration allows starting jobs on a remote node and if this
# is not a Linux cluster, you must add  -l nodes=<host>  to the params line and substitute
# <host> by the actual name of the local host. This will prevent jobs from being submitted
# to a different node. You can find out the local host name by executing command hostname.
#
#  If the gateway is installed on a Linux cluster.
#  it is worth putting correct number of cpu per node
#  the information can be obtained by running pbsnodes and examining parameters  np to find the most
#  representative number, otherwise Gateway will try to guess it... The guess is correct if cluster is homogeneous
#
#  NumCPUSsperNode = <numcorespernode>

#
# The following line specifies the default queue.
#
 default_queue=batch
# <queue name> should be substituted by the actual queue name.
# This queue will be shown first in the user interface on the client site.
# If there is a default queue in the queuing system you can omit this step.

进入/home/msi/Accelrys/MaterialsStudio6.0/etc/Gateway，执行
    ./msgateway_control_18888 restart

还要在windows上的MS server console 中添加 远程服务器

正常情况就可在windows下提交任务到linux 计算了