analysis/
contains the R markdown files used to generated the reports and the website.code/
contains the R code.container/
contains the singularity container.docs/
contains the generated HTML files.guix/
contains the manifest and channel description to create the container.scripts/
contains the container start scripts.
GUIX and Singularity have to be installed on the system you want to use to create the container.
# on a Debian GNU/Linux
sudo apt install singularity-container make rsync
# newer Debian versions may support installation of guix via apt
cd /tmp && \
wget https://git.savannah.gnu.org/cgit/guix.git/plain/etc/guix-install.sh && \
chmod +x guix-install.sh && \
sudo ./guix-install.sh && guix pull
This is a workflowr project. After creating the singularity container it was created by:
(You can skip this if you downloaded/clone this repository.)
singularity exec container/ampel-leipzig-meld.sif \
R -e 'library("workflowr"); wflow_start(".", existing = TRUE)'
- Linux cluster
slurm
(version 21.08.x)singularity
(tested with version 3.4.2)ssh
access.
The login node in _targets.R
(here brain
) has to match an
entry in ~/.ssh_config
or be configured manually.
login <- future::tweak(
future::remote,
workers = "brain",
The partitions/ncpu/mcpu settings have to be changed in the _target.R
.
slurm <- future::tweak(
future.batchtools::batchtools_slurm,
template = "scripts/slurm_batchtools.tmpl",
workers = 16L,
resources = list(
partition = "batch,compute,snowball",
ncpu = 64, # 72, 64
mcpu = 2L * 1024L, # 2L
walltime = 24L * 60L * 60L # seconds
)
)
The scripts/Rscript.sh
has to be adopted to the cluster's module system and
singularity version. Maybe the scripts/slurm_batchtools.tmpl
has to be
modified as well.