Closed ChiaraF32 closed 2 years ago
You should be able to use the original Snakefile and set the tmpdir flag on the command line. You have to wrap the tmpdir
variable in double quotes.
Look at snakemake's resource section: https://snakemake.readthedocs.io/en/stable/snakefiles/rules.html#standard-resources
snakemake rnaVariantCalling --cores 10 --default-resources "tmpdir='/data/tmp_drop'" --keep-going
EDIT
In my test run the path/to/tmpdir
must be the full absolute path, otherwise the R
processing sections will fail.
Hopefully this works. If not I would reccommend changing the environment's $TMPDIR
variable to align with this:
https://docs.python.org/3/library/tempfile.html#tempfile.gettempdir
Thanks for your help, it is now working so I will close this thread!
Hi, I am trying to run the DROP variant calling pipeline on a group of around 30 skeletal muscle rnaseq data.
I am running the pipeline on a virtual machine with 16 CPU and 64 GB of RAM, OS Ubuntu 18.04
I keep running into issues with the pipeline writing tmp files to
/tmp
and apparently maxing out the available storage, causing the run to fail. In my set-up, thetmp
directory is in a root volume with limited space (40 GB) so it would be preferable to write to a directory in my/data
volume which has more space.An example error I am getting is below:
I first tried adding a
default-resources
specification to myconfig.yaml
file (after reading this article). I also added the--default-resources
flag into my run command:However, this did not seem to alter what resources were being used in each step, so the pipeline failed again. See below for a copy of my
config.yaml
file.This didn't work though I first had a look in the Snakefile which I found in a sub-directory of my project directory:
/data/drop/Scripts/rnaVariantCalling/pipeline/Snakefile
I tried modifying the Snakefile manually for each step that uses a gatk tool by adding in the
--TMP_DIR
flag and specifying a directory within my/data
partition, see attached Snakefile, and below for an example. This didn't seem to work either; the log files were still saying that the default/tmp
directory was being used. I then tried deleting all occurrences of-java-options "-Djava.io.tmpdir={resources.tmpdir}
on top of using the--TMP_DIR
flag. This didn't work either, with the pipeline still writing files to/tmp
.I feel a bit stuck as to what to do next. Maybe I am just missing something really obvious and simple! Any help would be very much appreciated. Snakefile.txt