e. Spack Intro
1. Check Spack Installation
Spack is already installed on the cluster, to check the version run:
$ spack --version
0.15.3
2. Use Binary Mirror
By default Spack installs everything from source, to install things faster by using a binary mirror, run:
export AWS_REGION=$(curl -s http://169.254.169.254/latest/dynamic/instance-identity/document | grep -oP '\"region\"[[:space:]]*:[[:space:]]*\"\K[^\"]+')
. /shared/spack-1share/spack/setup-tutorial-env.sh # this errors, just hit Ctrl-c
spack mirror add tutorial s3://spack-tutorial-container/mirror/
spack gpg trust /shared/spack-0.15/share/spack/keys/tutorial.pub
3. Install Software
On the cluster, the ubuntu user has full permissions to install software via the sudo command. Note, however, in order to use software in a batch scheduled job, you must either:
- Install the software on a shared filesystem common to Master and Compute nodes, or
- Install the software individually on every compute node.
We recommend option 1, to this end, we’ve mounted /shared as a shared file system common to Master and Compute nodes. Furthermore, we have pre-configured Spack to install packages to /shared/spack and make them accessible available as a Modulefile.
For example, to install SRA-Toolkit, you might use the following workflow to search for, and then install the package:
spack list sra
spack install sra-toolkit
Once it’s been installed you can load and unload SRA-Toolkit with
module avail
module load intelmpi sra-toolkit/2.9.6-gcc-7.5.0
