Installation

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX''">XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX''">

1. The master host is the machine which continually runs the Analysis Manager daemon (called QueMgr). This is also referred to as the master node.

2. The submit host is the machine from which the analysis is submitted, sometimes referred to as the client also.

3. The analysis host is the machine which actually executes the analysis.

1. One master node must be chosen for each organizational group (for each Queue Manager that will be running - typical installation only have one!).

2. Queue Manager (QueMgr) should run as root on the master node. This is not a strict requirement but recommended on Unix. On Windows it can run as a user or as administrator.

3. Each node (submit and analysis hosts) in the Analysis Manager configuration must be reachable to and from the master node via a TCP/IP network.

4. Each analysis host must have a Remote Manager (RmtMgr) running with the same port number (for each QueMgr). It is recommended that each Submit machine also (especially on Windows) however this is not a strict requirement. (This takes the place of rsh (remsh) remote access capabilities that used to be used in older versions of the Analysis Manager.)

5. The Analysis Manager software will come off of the installation media onto any machine (master, submit, or analysis host) under the $P3_HOME/p3manager_files directory. The $P3_HOME variable is the installation directory and is typically set up as /msc/patran200x and is usually defined as an environment variable. This p3manager_files directory and tree must exist as-is and not be renamed.

6. Each analysis host machine in the Analysis Manager configuration must be able to identically see the installation tree. If a RmtMgr is running this is not an issue because the RmtMgr knows where the Analysis Manager executables are.

7. The root user should run the administration program (p3am_admin (AdmMgr)) on the master node to test and ensure that new users can correctly access the Analysis Manager. See Configuration Management Interface.

1. Users who are using the Analysis Manager should have the same login name, user and group ids on all hosts / nodes in the Analysis Manager configuration. This will prevent file access problems. In specific cases, users may run jobs on different accounts other than their own, but this must be set up by the system administrator. This is described in Examples of Configuration Files.

2. Users must have uname in their default search path (path or PATH environment variable in the user's .cshrc or .profile file).

1. Unload the p3manager_files directory from the installation media. (Consult the Installation guide for more information on how this is done.)

2. Decide on a master node (typically the node the Patran software is located on), and login to that node as root.

3. Decide which machine(s), that have MSC Nastran, MSC.Marc, ABAQUS, or other analysis modules to be used, will be included in the Analysis Manager’s configuration. Find out where each runtime script and/or configuration files are located (i.e. /msc/bin/nast200x, /msc/conf/nast200xrc for MSC Nastran) for each machine. Only these machines will be enabled for later job submission, monitoring, and management.

4. Each analysis host machine that will be configured to run an analysis code must be able to see the p3manager_files directory structure as outlined in Directory Structure. This directory structure must also exist on the master node as well as client (submit) nodes. This can be done in one of two ways. Either the directory structure can be copied directly to each machine so that it can be accessed in the same manner as on the master node, or symbolic links and NFS mounts can be created. In any case, if on one machine you type

As an example of setting up a link, suppose that the machine venus is the master host and has the installation directory structure in /venus/users/msc/patran200x. A link can be established on venus by typing:

This will ensure that analysis host jupiter can see the installation directory structure. Repeat this for all analysis hosts. NFS mounts are not necessary if you wish to copy the installation directory structures to each host separately instead of creating links.

Each submit host (hosts that submit jobs) does not necessarily need to see the directory structure in exactly the same way as the master and analysis hosts do. They only need to be able to see an installation directory structure to find the user interface executable (P3Mgr).

Note:	The above description sounds a bit more restrictive than it really is. In actuality, if a RmtMgr is started on each analysis host, the directory structure can be seen because RmtMgr knows from where it was launched and thus knows where all the Analysis Manager executable are. However, it is still recommended to follow the above procedure if at all possible.

5. Start up the RmtMgr daemon or service on each and every analysis node. It is recommended to start RmtMgr on submit machines also. Starting the Queue/Remote Managers explains this procedure. This must be done before configuration testing can be done.

6. Use the p3am_admin program to set up the configuration files. This program is located in

Modify Configuration Files explains the use of this program and the format of the generated configuration files as a result of running this program. The configuration file will be placed in the correct locations automatically. The following configuration files will be generated:

host.cfg	Host file configuration file
disk.cfg	Disk space configuration file
lsf.cfg	LSF configuration file (if you plan to use LSF as your scheduler, and not the Analysis Manager own built-in scheduler.
nqs.cfg	NQS configuration file (if you plan to use NQS as your scheduler, and not the Analysis Manager own built-in scheduler.

Note:	For a minimal configuration with a single Queue Manager, you should remove or rename the file $P3_HOME/p3manager_files/org.cfg. See step 12. for more information.

7. Test the configuration setup using p3am_admin’s testing features. Specifically, do basic tests and network tests for each user that wishes to access the Analysis Manager. Test Configuration explains this procedure in detail.

8. Start up the QueMgr daemon on the master node. Starting the Queue/Remote Managers explains this procedure.

9. Add commands to the appropriate rc files for automatic start-up of the QueMgr and RmtMgr daemons when the master, submit or analysis nodes have to be rebooted. Starting the Queue/Remote Managers also explains this procedure.

10. Invoke the Analysis Manager user interface as a normal user and check that the installation was performed properly. Invoking the Analysis Manager is explained in Invoking the Analysis Manager Manually.

11. Repeat the procedure from step 2. for each organizational group (Queue Manager) you wish to
set up.

12. When more than one organizational group (Queue Manager) is to be accessed, either modify the org.cfg file and add the port numbers and group names, or have users set the appropriate environment variables to access them. See Organization Environment Variables for an explanation of these variables and see Examples of Configuration Files for setting up the org.cfg file.

13. Make sure users have $P3_HOME/bin in their path. Most Analysis Manager executables can be invoked from$P3_HOME/bin or are links from $P3_HOME/bin for setting all environment variables. These include: