Private module files allow you to load your own installed software packages into your environment and to handle different versions without getting into conflicts. Private modules can be setup for a single user as well as all users of a project group. The workflow and settings for user private as well as project private module files are described in the following.
0. Build and Install Software¶
Obviously the first step is to build and install the software you'd like to use. Please follow the
instructions and tips provided on the page building_software.
For consistency, we use the placeholder variable
<sw_name> in this documentation. When following this
instructions, please substitute it with the actual software name within the commands.
1. Create Directory¶
Now, create the directory
privatemodules to store all your private module files and the directory
sw_name therein. All module files for different versions or build options of
sw_name should be
located in this directory.
marie@compute$ cd $HOME marie@compute$ mkdir --verbose --parents privatemodules/<sw_name> marie@compute$ cd privatemodules/<sw_name>
Project private module files for software that can be used by all members of your group should be
located in your global projects directory, e.g.,
/projects/p_number_crunch/privatemodules. Thus, create
marie@compute$ mkdir --verbose --parents /projects/p_number_crunch/privatemodules/<sw_name> marie@compute$ cd /projects/p_number_crunch/privatemodules/<sw_name>
Make sure, that the directory is group-readable.
2. Create Modulefile¶
Within the directory
<sw_name> create the module file. The file can either be a TCL or a Lua. We
recommend to use Lua. The module file name should reflect the particular version of the software,
If you create a group private module file, make sure it is group-readable.
A template module file is:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
The most important functions to adjust the environment are listed and described in the following table.
||Message when the help command is called.|
||The current modulefile will only load if all listed modules are NOT loaded.|
||Loads all modules. When unloading only dependent modules are unloaded.|
||Load all modules. Report error if unable to load.|
||Prepend the value to a path-like variable.|
||Assign the value to the environment variable
In order to use private module files and the corresponding software, you need to expand the module search path. This is done by invoking the command
marie@login$ module use $HOME/privatemodules
for your private module files and
marie@login$ module use /projects/p_number_crunch/privatemodules
for group private module files, respectively.
Afterwards, you can use the module commands to, e.g., load and unload your private modules as usual.
An automated backup system provides security for the home directories on the cluster on a daily
basis. This is the reason why we urge users to store (large) temporary data (like checkpoint files)
/scratch filesystem or at local scratch disks.
This is also why we have set
ulimit -c 0 as a default setting to prevent users from filling the
home directories with dumps of crashed programs. In particular,
ulimit -c 0 sets the core file
size (blocks) to 0, which disables creation of core dumps in case an application crashes.
Enable core files for debugging
If you use
bash as shell and you need these core files for analysis, set