Altering resource usage

This workflow sets a default number of threads and runtime for each rule. Memory is not set per rule as development took place on clusters that did not allow direct requests of memory when submitting SLURM jobs, instead always allocating 6.4GB RAM per core for each job. The easiest way to get up and running with this workflow is to set --default-resources mem_mb="XXXX * threads" when running the Snakemake command, replacing XXXX with the RAM available per core on your HPC system in MB (in the case of ours, this was 6400). See the profiles in the GitHub repository as an example.

However, the default resources may not always work, your data may need more memory, or longer runtimes, or maybe you even need shorter if your HPC has shorter runtime limits than some of the defaults we set (up to 7 days). Snakemake makes it easy to alter resources in the command line using the --set-resources and --set-threads options, which will override anything set in the workflow already.

This is even better set up in a profile. PopGLen includes a default profile in the profiles/default folder, which has all the rules already populated with the default resources already set. If you have downloaded the workflow via cloning the repository, all you need to do is change the values to match your needs (which may require adding an additional resource option, like mem_mb). If you have downloaded via Snakedeploy, just download the config file from the main repository, place it in the same directory in your deployed workflow, modify as needed and Snakemake will automatically pick it up when you run it.

Here is an example of what it would look like to change the number of threads for the rule bwa_index from the default of 1 to a new value of 5:

set-threads:
  # Reference Prep
  link_ref: 1
  link_anc_ref: 1
  bwa_index: 5
  samtools_faidx: 1
  ref_chunking: 1
  picard_dict: 1

And if you wanted to update the runtime to give it only a maximum of 1 day and add a limit of 2GB for memory:

set-resources:
#   # Reference Prep
  link_ref:
    runtime:
  link_anc_ref:
    runtime:
  bwa_index:
    runtime: "1d"
    mem_mb: 2048
  samtools_faidx:
    runtime:
  ref_chunking:
    runtime:
  picard_dict:
    runtime:

You can also use the term 'attempt' in these definitions, which allow you to scale the resources with the number of attempts a rule has made, automatically increasing threads, runtime, or memory with each attempt. This is already done for several rules in the config, largely to automatically request more memory due to OOM errors.