Installing Applications on Sapelo2
Introduction
We introduce here some guidance on how to install applications and libraries on Sapelo2.
In general, users can build applications under their home directory or other space owned by the users (/home/MyID/).
If an application will be used by several members of a group, the application can be installed in the user's group work space (/work/abclab/).
GACRC team takes requests to install applications at a central place (/apps) if the application satisfies the following conditions:
- The application is Linux compatible.
- The application has general interest among users.
- The application needs to be built with special settings, such as root privilege, shared common data set, complex structures, or other requirements.
- The application needs to be configured to integrate into the GACRC environment, such as queue settings or database connections.
Please use the GACRC Software Installation/Update online form to submit a support ticket to GACRC team if you need any help from us.
General Guidlines
IMPORTANT: Please DO NOT install applications on the login node. Instead, please install applications from an interactive session started with the interact command.
Sapelo2 login node (sapelo2.gacrc.uga.edu) has limited memory. Most of the time software installation fails due to insufficient memory. More importantly, the process of installing applications on the login node can degrade the performance of the cluster for everyone. So please do not install any applications on the cluster while you are on the login node. We strongly advise that you build or install applications from an interactive session that you can start with interact command from the login node:
interact
All of the following installation examples were executed in a interact environment.
How to check if an application is installed
To find if an application, e.g. Trinity, is already installed on Sapelo2, use the following command (the application name is NOT case sensitive):
module spider trinity
or
module avail trinity
To see a description of the module (the application name is case sensitive)
module whatis Trinity
To see configuration details, e.g., environment variables, of the module (the application name is case sensitive)
module show Trinity
How to check if a library is installed
Perl
To check if a perl library is installed, e.g. DBI, first load the Perl package of interest, for example Perl/5.34.1-GCCcore-11.3.0:
module load Perl/5.34.1-GCCcore-11.3.0 perl -MDBI -e 'print "OK\n"'
If the library is not installed, it will warn that the library is not in the path. Otherwise, it will print OK.
Python
To check if a Python library is installed, e.g. numpy, first load the Python package of interest, then run "pip show" or "pip list" commands. The Linux command "which" is to identify if pip is in place. In the following example, we use Python/2.7.18-GCCcore-11.3.0:
module load Python/2.7.18-GCCcore-11.3.0 which pip pip show numpy
To list all installed libraries with their versions:
module load Python/2.7.18-GCCcore-11.3.0 pip list
To generate an alphabetical list of the libraries installed under a Python version:
module load Python/2.7.18-GCCcore-11.3.0
pip freeze| sort | awk -F'@' '{print $1}'
For Python3, some versions have both pip3 and pip. pip3 is the Python3 version of pip. If you load Python3, run pip3, otherwise run pip. You can use "which" command to identify if pip3 is in place. For example:
module load Python/3.10.4-GCCcore-11.3.0 which pip3
Then check if a Python3 library is installed, e.g. using pip3:
module load Python/3.10.4-GCCcore-11.3.0 pip3 show numpy
To list all installed libraries with their versions:
module load Python/3.10.4-GCCcore-11.3.0 pip3 list
To generate an alphabetical list of the libraries installed under a Python version:
module load Python/3.8.2-GCCcore-8.3.0
pip3 freeze| sort | awk -F'@' '{print $1}'
If the library is not installed, it will not be shown by pip or pip3 commands. Otherwise, pip or pip3 will give you information about libraries' version and installation location.
Note: Some Python libraries are installed outside of the Python default location and are provided as a separate module file that needs to be loaded separately. To check if a python library is installed as a separate module, please use the module spider command.
For example, to check if matplotlib is installed as a separate module:
module spider matplotlib
This command will return all the matplotlib versions currently installed on the cluster. For example, the module named matplotlib/3.5.2-foss-2022a provides matplotlib version 3.5.2 for Python 3.10.4 and it uses the foss-2022a toolchain.
To use this version of matplotlib, please load the module matplotlib/3.5.2-foss-2022a. Similarly, you can load other module files that provide other Python libraries, for example, SciPy-bundle module provides numpy, spicy, and pandas, etc..
R
To check if a R library is installed, e.g. xtable, first load the R version of interest, for example R/4.3.1-foss-2022a:
interact
module load R/4.3.1-foss-2022a
R
require("xtable")
Ctrl+d
If the library is not installed, it will warn that the library is not in the path. Otherwise, it will print "Loading required package: xtable".
How to structure directories
Common practice is to set up the following three directories using "mkdir" command in the user's home directory /home/MyID/:
- apps: directory where the applications will be installed
- src: directory under which you can store the source files and build the applications
- modulefiles: directory under which you can put your own module files
These directories can be created with
mkdir ~/apps mkdir ~/src mkdir ~/modulefiles
How to use local Lmod modules
To use your own modules, e.g. trinity/1.0:
module use ~/modulefiles module load trinity/1.0
If there are module files with the same name as in our central place, your local ones will take precedence over the central ones.
How to Install Software Packages (general)
If you are interested in using an application, check the website of this application to see if (a) it is compatible with Linux, (b) it is distributed as binaries or source code, (c) and the kind of package that it is (Python library and scripts, Perl scripts, Java, C, C++, Fortran code, etc).
Perl scripts and Java jar files can be downloaded into the users' home directory and run from there.
If the application is distributed as pre-compiled binaries, check if binaries are available for the Linux OS that our cluster run (Sapelo2 runs Rocky Linux 8). Binaries compiled for Window and for Mac OSX cannot be run on Sapelo2. Also, binaries generated in other Linux distributions (e.g. Ubuntu, Debian, etc) will in general not work on Sapelo2.
If pre-compiled binaries are not available for the OS on Sapelo2 (Rocky Linux 8), then you can compile the code yourself, if the source code is available.
Another option for running binaries compatible with other Linux distributions is to create a Singularity container and run it on Sapelo2 (more information below).
Installing Software Packages
How to build C, C++ applications
Here is an example on how to install an application called GERUDsim3, which comprise a single C++ program.
1. Login to sapelo2 login node. If you do not have a directory called apps, create one with
mkdir ~/apps
2. Download the application into the ~/apps directory. For the example used here, you can download it with
cd ~/apps git clone https://github.com/JonesLabIdaho/GERUDsim3.git
3. Start an interactive session with
interact -c 4 --mem 8gb
4. compile the code on the interactive node with:
cd ~/apps/GERUDsim3/GERUDsim3/Source_files g++ GerudSim3.cpp -o GerudSim3
The g++ command that is on user's default path is version 8.5.0:
which g++ /usr/bin/g++ which gcc /usr/bin/gcc g++ --version g++ (GCC) 8.5.0 20210514 (Red Hat 8.5.0-18) gcc --version gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-18)
If you wish/need to use a different GNU compiler version, or a different compiler (e.g. Intel), then first load the corresponding module and then compile the code. For example, to use g++ 11.3.0:
cd ~/apps/GERUDsim3/GERUDsim3/Source_files module load foss/2022a g++ GerudSim3.cpp -o GerudSim3
5. The name of your executable is GerudSim3. You can run it with the full path
~/apps/GERUDsim3/GERUDsim3/Source_files/GerudSim3
or copy this binary to ~/bin (create this directory first, if it does not exist yet) and run it with
~/bin/GerudSim3
You can also add ~/bin to your default PATH. To do that, add the following in your .bashrc file:
export PATH=/home/MyID/bin/:$PATH
where MyID needs to be replaced by your UGA MyID. With this, any executable you put in ~/bin will be in your PATH and can be invoked without its full path.
For more information on compilers available on Sapelo2, please see Code Compilation on Sapelo2.
Cross build for mixed processor architecture
At Sapelo2, we have two types of processors: Intel and AMD.
In some C or C++ applications, compiler performance-optimization flags, such as -march=native, may be introduced in the configuration step or in the Makefile. Executables compiled with such options on one type of processor might not run on a different processor type. To enable the compiled application to run on both types of processors, these processor specific compiler optimization options should be removed. One option is to set the flags as -mtune=generic -march=x86-64.
If you want to compile or test your program on different types of processors to ensure that they work there, you could start an interactive session on each of the processor types. For example, to start an interactive session on an AMD EPYC node, use
interact --constraint EPYC
You could choose a different processor type with the --constraint option. For more information on how to start interactive sessions on Sapelo2, please see Running interactive jobs on Sapelo2
How to install a Perl module
A convenient way to build and install perl modules is to use CPAN, which can be used to install Perl modules in the user's home directory. We suggest that you do this installation in an interactive session.
1. Start an interactive session with
interact
2. Decide where you want the perl modules to be installed, e.g. ~/perlmods, and create this directory if it does not exist yet
mkdir ~/perlmods
3. Load the perl module that you want to use, e.g. Perl/5.34.1-GCCcore-11.3.0
module load Perl/5.34.1-GCCcore-11.3.0
4. Start the CPAN shell with
cpan
When you start cpan for the first time, some configurations will be set (unless you already have configurations set in ~/.cpan/CPAN/MyConfig.pm):
cpan cpan shell -- CPAN exploration and modules installation (v2.28) Enter 'h' for help. cpan[1]>
5. From within the CPAN shell, enter the following two (2) commands to specify the installation directory:
cpan[1]> o conf mbuildpl_arg "--install_base /home/MyID/perlmods"
mbuildpl_arg [--install_base /home/MyID/perlmods]
Please use 'o conf commit' to make the config permanent!
cpan[2]> o conf makepl_arg "PREFIX=/home/MyID/perlmods"
makepl_arg [PREFIX=/home/MyID/perlmods]
Please use 'o conf commit' to make the config permanent!
cpan[3]>
Note that in the two commands above you will need to replace MyID by your own MyID and change perlmods if you are using a different directory name. If you want to set the above installation path as the default one, you can make the settings above permanent by entering "o conf commit". If this is not done, you will need to reset this value every time you restart CPAN. If you do make the settings permanent, you can always change them later and re-commit as shown above.
6. To install a module (for example, if you want to install "IO::CaptureOutput") enter:
cpan[3]> install IO::CaptureOutput
Respond to any prompts for information that might be requested. When you are finished, enter:
cpan[4]> quit
7. After you have successfully installed a local Perl module, set the PERL5LIB environmental variable to tell Perl where to find the module. For example:
export PERL5LIB=/home/MyID/perlmods/lib/perl5/site_perl/5.34.1/:$PERL5LIB perl -MIO::CaptureOutput -e 'print "OK\n"' OK
where /home/MyID/perlmods has to be replaced by the path to your installation directory.
You can add this export command to your .bashrc file if you'd like to ensure that this PERL5LIB environment variable is always set when you login and for your non-interactive scripts. If this export line is not added in your .bashrc file, you can add it to your job submission scripts. You can also define a module file where this variable is defined.
How to install Python packages
When installing Python packages it's important to keep them organized based on the project you're working on (as opposed to installing all Python packages you'll ever use into one location). This is typically done with Python virtual environments (or Conda environments if there's a scientific package you need only available through Conda). For example, if you were working one project that involved machine learning and another project involving data visualization, you may need different versions of some of the same Python packages and should keep them separate. A Python virtual environment is just a directory in which Python packages are installed, separate from other unrelated packages. To create a virtual environment in which to install Python packages in your /home directory, follow these steps below:
Create a virtual environment in your /home directory
1. Start an interactive job:
interact
2. Search for the Python version you would like to use:
gacrc-test@d2-13 ~$ module spider Python
3. Load the software module for the version of Python you would like to use:
gacrc-test@d2-13 ~$ module load Python/3.10.4-GCCcore-11.3.0
4. Create the Python virtual environment in your /home directory. You may find it beneficial to have a directory in your /home directory for all of your Python virtual environments.
gacrc-test@d2-13 ~$ python -m venv ~/env/mypyenv
This creates a base virtual environment to work from, in this case called "mypyenv". You should use more meaningful names for your actual Python environments. The ~/envs directory in which I would install all of my Python virtual environments in this example doesn't have to exist already when I run the above command. If it doesn't, the above command would create it. This gives us a base Python virtual environment to work with:
gacrc-test@d2-13 ~$ ls ~/env/mypyenv/
bin include lib lib64 pyvenv.cfg
gacrc-test@d2-13 ~$ ls ~/env/mypyenv/bin/
Activate.ps1 activate activate.csh activate.fish pip pip3 pip3.10 python python3 python3.10
5. Now that the environment is created, we can activate it and install Python packages. To do so, we need to source the bin/activate file. This file sets the appropriate environment variables so that any Python packages you install will be installed in your virtual environment. To source file and thus activate the virtual environment:
gacrc-test@d2-13 ~$ . ~/env/mypyenv/bin/activate
Note the space in between the period (source command) and the path to the activate file. Upon running this command, you will see that your command prompt has changed, showing you the name of your active environment:
gacrc-test@d2-13 ~$ . ~/env/mypyenv/bin/activate
(mypyenv) gacrc-test@d2-13 ~$
Now any Python packages you install will be installed in the "lib" directory of your Python virtual environment (whether you use pip or download the package and install it with its setup.py script) and are usable when your Python virtual environment is active.
6. To deactivate a Python virtual environment:
deactivate
Installing Python packages in your virtual environment using pip
The first time you ever use pip in your Python virtual environment you should probably upgrade pip.
pip install --upgrade pip
Then all you have to do is pip install the package, and it will be installed in your virtual environment:
pip install <package name>
or using a requirements.txt file:
pip install -r requirements.txt
Installing Python packages in your virtual environment from the source code
If the Python package that you want to install in your virtual environment is not available through pip for some reason, you can download the source code in your home directory and run the package's setup.py script. As long as your Python virtual environment is active when you do this it will be installed in that virtual environment. Suppose you have downloaded a Python package tarball file, e.g., myPackage-1.0.tar.gz, in the source folder (src) in your /home directory with the full path being:
/home/MyID/src/myPackage-1.0.tar.gz
Change directory to /home/MyID/src/:
cd ~/src
Uncompress and untar the tarball file using "tar xzvf" command:
tar xzvf myPackage-1.0.tar.gz
cd into the new directory:
cd myPackage-1.0Run the setup.py script:
python setup.py --install
Using your Python virtual environment in a job
Once you have created your Python virtual environment and installed the packages you need, you can use it in a job on the cluster by loading the Python module you used to create the environment and then activating the environment:
module load Python/3.10.4-GCCcore-11.3.0
. ~/envs/mypyenv/activate
How to install Conda packages
To install a software package as a conda environment in user's Sapelo2 home directory, please use the option "-p" to define your environment installation path, e.g., /home/MyID/busco_conda.
For example, to install busco v3.0.2 conda environment at /home/MyID/busco_conda:
module load Miniconda3/4.12.0 conda create -p /home/MyID/busco_conda -c bioconda busco=3.0.2
where bioconda is the channel where busco v3.0.2 source is downloaded. Other common conda channels are, for example, conda-forge, r, defaults, and qiime2, etc.. Please refer to Conda channels
If the busco version 3.0.2 is not specified, the most recent version of busco will be downloaded and installed, i.e.,
module load Miniconda3/4.12.0 conda create -p /home/MyID/busco_conda -c bioconda busco
To list conda environments you installed:
module load Miniconda3/4.12.0 conda env list
To activate your conda environment, you need to give your environment path to source activate command:
module load Miniconda3/4.12.0 source activate /home/MyID/busco_conda/
Once a conda environment is activated, you can list packages installed inside, with their names, versions, and downloading channels. For example, to list packages installed in busco v3.0.2 conda environment:
module load Miniconda3/4.12.0 source activate /home/MyID/busco_conda/ conda list
You can add or install new packages to an existing environment. For example, to install scipy v1.5.0 package to busco v3.0.2 conda environment:
module load Miniconda3/4.12.0 source activate /home/MyID/busco_conda/ conda install scipy=1.5.0
Likewise, if the scipy version 1.5.0 is not specified, the most recent version of scipy will be downloaded and installed for you.
To remove an installed package, e.g., the scipy package, from your conda environment:
module load Miniconda3/4.12.0 source activate /home/MyID/busco_conda/ conda remove scipy
To deactivate a conda environment:
module load Miniconda3/4.12.0 conda deactivate /home/MyID/busco_conda/
Once a conda environment is deactivated, its conda layer will be removed from your Linux shell environment.
How to install R packages
If you wish to install an R package in your home directory, first decide the directory where you will install it. For example, to install it in ~/Rlibs, first create this dir if it does not exist yet:
mkdir ~/Rlibs
Then, create a file called ~/.Renviron containing the following line:
R_LIBS_USER=/path/to/Rlibs
Please replace /path/to/Rlibs with the path that you want to use. For example, to /home/MyID/Rlibs, where MyID needs to be replaced by your UGA MyID.
Start an interactive session and load the module for the version of R you want to use (e.g. R/4.3.1-foss-2022a)
interact module load R/4.3.1-foss-2022a
Install R package using R command line
If you download the R package tarball (e.g. brocolors_0.1.tar.gz) to your home dir, you can install it with R CMD INSTALL at the command line, but use the flag --library=/path/to/Rlibs, as follows:
R CMD INSTALL --library=/path/to/Rlibs brocolors_0.1.tar.gz
Note that the path for --library flag should be the same path contained in ~/.Renviron as described above.
Install R package in an interactive R session
You can install a package in an interactive R session using install.packages(). To use this method, you need to create the ~/.Renviron file as described above. For example, to open an interactive R session to install ggplot2 package into the path given in ~/.Renviron:
$ R
R version 4.3.1 (2023-06-16) -- "Beagle Scouts"
Copyright (C) 2023 The R Foundation for Statistical Computing
Platform: x86_64-pc-linux-gnu (64-bit)
R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.
R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.
Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.
> install.packages('ggplot2')
You will be prompted to enter your selection of a CRAN mirror site for downloading package source. Usually we select site 72: USA (OH) [https]. If one site does not work well for you, you could try others.
How to install Java applications
Most third-party Java applications are distributed as pre-compiled binaries (jar file), which users can download into their own home directories. For example, to install picard 2.4.1 in your home directory.
1. Create a directory where you want to install the java application (e.g. picard). For example
mkdir -p ~/apps/picard
2. Download the package (e.g. picard-tools-2.4.1.zip) from their website, put into your chosen directory (e.g. ~/apps/picard) and extract the file. For example
cd ~/apps/picard unzip picard-tools-2.4.1.zip
This will create a directory called picard-tools-2.4.1 that contains picard.jar. You can rename this dir e.g.
cd ~/apps/picard mv picard-tools-2.4.1 2.4.1
3. To run this application, load a java module and invoke this application in your job submission script. For example
module load Java/15.0.1 java -Xmx20g -classpath "/home/MyID/apps/picard/2.4.1" -jar /home/MyID/apps/picard/2.4.1/picard.jar [options]
where MyID needs to be replaced by your own MyID.
How to build complex applications
Many applications use a configure step to check on system libraries and path to dependencies in order to create Makefiles. The Makefiles are then used to build and install the application. To illustrate how this process is typically set up, we will describe how to install the GNU Scientific Libraries (GSL) v2.6 using the GNU 11.3.0 compilers.
1. Create a directory to use for building the application, e.g. ~/src/gsl
mkdir -p ~/src/gsl
2. Download the source tarball into your chosen directory, e.g.
cd ~/src/gsl wget http://mirror.sbb.rs/gnu/gsl/gsl-2.6.tar.gz
3. Start an interactive session
interact
4. Load the module for the compiler suite that you want to use, e.g. GCC 11.3.0:
module load foss/2022a
5. Change into your working directory and check the tarball with
cd ~/src/gsl tar ztvf gsl-2.6.tar.gz
and extract the files with
tar zxvf gsl-2.6.tar.gz
6. Change into the extracted directory
cd gsl-2.6
7. Create a subdirectory to build the application
mkdir build_gcc1130 cd build_gcc1130
8. Check all the configure options with
../configure --help
9. Configure the application with e.g.
../configure --prefix=/home/MyID/apps/gsl/2.6/gcc1130
where /home/MyID/apps/gsl/2.6/gcc1130 should be replaced by the directory where you want to install GSL.
You can capture the standard error and standard output of the configure step into a file, to help troubleshoot the step if it encounters any issues. This can be done with
../configure --prefix=/home/MyID/apps/gsl/2.6/gcc1130 2>&1 | tee my.config.log
10. If the configure step worked, you can build the application with
make 2>&1 | tee my.make.log
11. Some applications, including GSL, provide some tests that can be run after the build step to ensure the application was built correctly. For example, for GSL you can run
make check 2>&1 | tee my.check.log
12. Install the application with
make install 2>&1 | tee my.install.log
How to link dependencies
Some applications need third party program, library, header files to build. There are various ways to introduce these to the build command.
Usually the manual/README of the application would elaborate more about the needed configuration variables. To explore these configuration variables from command is to issue --help options following the main build command.
Such as cmake build,
cmake --help
Or if it is a configuration build,
./configure -h
To set needed components in build environment, this could simply be done by loading modules prior to build.
module load CMake/3.24.3-GCCcore-11.3.0 module load zlib/1.2.12-GCCcore-11.3.0 cmake -DCMAKE_INSTALL_PREFIX:PATH=/home/MyID/app/diamond/1.0 ...[skipped] make install
Line 1: load CMake module and zlib module.
Line 2: set prefix, there are other variable could be defined here.
Or set the variables explicitly by export command
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/apps/eb/GSL/2.7-GCC-11.3.0/lib export CFLAGS="-I/apps/eb/GSL/2.7-GCC-11.3.0/include" export LDFLAGS="-L/apps/eb/GSL/2.7-GCC-11.3.0/lib" export LIBS="-lgsl" python setup.py build_ext --inplace
Line 1: specify the location of gsl shared dynamic lib, pass this to shared library loader at runtime
Line 2: specify the location of gsl c header file, pass this to GNU C compiler at compilation time
Line 3 and 4: specify the location of gsl shared dynamic lib and the lib, pass this to GNU linker at linkage time
Another way is to define variables at configuration line.
./configure --prefix=/home/MyID/app/diamond/1.0 --with-jemalloc=/apps/eb/jemalloc/5.3.0-GCCcore-11.3.0/lib
How to download Singularity images
Singularity image can be searched and downloaded from Singularity Container Library. For example, to pull Trinity v2.9.1 Singularity image built for amd64 (default) architecture:
singularity pull --arch amd64 library://colinsauze/default/trinity:v2.9.1
Docker container is the most well-known container system. Docker also has a bigger ecosystem than Singularity. However, Docker was initially designed for ephemeral servers; by default Docker tries to isolate the running container as much as possible, which makes it not suitable for running in a HPC environment. Like Docker, Singularity is a container runtime too. But it starts from a very different place. It favors integration rather than isolation. Singularity is also the best friend of Docker and can import images from Docker registries. You can search a docker image at Docker Hub. If the image you need is on Docker Hub, you can pull and build it into a Singularity image.
For example, to pull and build Trinity latest version Singularity image from Docker Hub:
singularity pull docker://trinityrnaseq/trinityrnaseq:latest
Detailed instructions on how to build Singularity container can be found at GACRC Singularity wiki page or at Singularity Build a Container