Using queues on clusters
To enable queuing on a Linux cluster with PBS Pro, OGE, or LSF installed
1、Follow the installation instructions provided with your queuing system to ensure that it is started automatically after a system restart.
2、Log in under the account where you are going to perform the installation of the Materials Studio servers and verify that the path to the queuing commands is listed in the PATH environment variable .
3、Confirm that the queuing system is properly configured and active; for example, verify that queuing system commands such as qstat are functioning correctly and that the appropriate daemons are running.
4、Ensure that the queues are configured to allow job submission from the user under whose account the Materials Studio servers will be installed, as well as from the accounts of all end users.
5、Install Materials Studio on your Linux cluster, carefully following the instructions provided.
6、To specify additional parameters for the qsub command for jobs submitted from the gateway, you can edit the respective configuration files (pbs.cfg, sge.cfg, or lsf.cfg) in the <MS_INSTALL>/etc/Gateway/root_default/dsd/commands/queues/[queuing system] directory, where [queuing system] is either PBS Pro, OGE, or LSF. Please take special note of the restrictions described in the PBS configuration file (pbs.cfg) regarding PBS Pro 10.4, especially for Windows.
7、You must now configure the gateway for use with the queuing system. This is best done through the Remote View dialog of the Server Console, which can be accessed by selecting Tools | Server Console from the Windows Start menu on the Materials Studio client machine. Follow the instructions provided in the "Queuing system support" topic in the help system of the Server Console application. Note that the Server Console has its own help system that is separate from the Materials Studio help.
8、When a queuing system is installed on a Linux cluster, the Materials Studio gateway automatically sets the number of cores per CPU parameter, corespercpu, to AUTO. The license pack calculates this value for Intel processors; however, the number of cores for AMD processors must be set manually. If the gateway is subsequently reconfigured from the shell via the configure script, you will need to reset this value.
9、The total number of cores available (across all processors and nodes) parameter, cpucorestotal, is calculated by the Materials Studio gateway based on the head node's resources only. You may need to adjust this to match the number of clustered CPU cores you wish to use. These nodes should also be explicitly named in the machines.LINUX file.
To enable queuing on a Linux cluster with PBS Pro, OGE, or LSF installed
1、Follow the installation instructions provided with your queuing system to ensure that it is started automatically after a system restart.
2、Log in under the account where you are going to perform the installation of the Materials Studio servers and verify that the path to the queuing commands is listed in the PATH environment variable .
3、Confirm that the queuing system is properly configured and active; for example, verify that queuing system commands such as qstat are functioning correctly and that the appropriate daemons are running.
4、Ensure that the queues are configured to allow job submission from the user under whose account the Materials Studio servers will be installed, as well as from the accounts of all end users.
5、Install Materials Studio on your Linux cluster, carefully following the instructions provided.
6、To specify additional parameters for the qsub command for jobs submitted from the gateway, you can edit the respective configuration files (pbs.cfg, sge.cfg, or lsf.cfg) in the <MS_INSTALL>/etc/Gateway/root_default/dsd/commands/queues/[queuing system] directory, where [queuing system] is either PBS Pro, OGE, or LSF. Please take special note of the restrictions described in the PBS configuration file (pbs.cfg) regarding PBS Pro 10.4, especially for Windows.
7、You must now configure the gateway for use with the queuing system. This is best done through the Remote View dialog of the Server Console, which can be accessed by selecting Tools | Server Console from the Windows Start menu on the Materials Studio client machine. Follow the instructions provided in the "Queuing system support" topic in the help system of the Server Console application. Note that the Server Console has its own help system that is separate from the Materials Studio help.
8、When a queuing system is installed on a Linux cluster, the Materials Studio gateway automatically sets the number of cores per CPU parameter, corespercpu, to AUTO. The license pack calculates this value for Intel processors; however, the number of cores for AMD processors must be set manually. If the gateway is subsequently reconfigured from the shell via the configure script, you will need to reset this value.
9、The total number of cores available (across all processors and nodes) parameter, cpucorestotal, is calculated by the Materials Studio gateway based on the head node's resources only. You may need to adjust this to match the number of clustered CPU cores you wish to use. These nodes should also be explicitly named in the machines.LINUX file.