Right now, Snakemake schedules jobs on the cluster by running a LSF scheduling command. When the job is almost finished, it writes a flag-file to disk, which tells Snakemake the jobs has finished.
This approach is brittle because it cannot detect cases in which the cluster kills the job due to exceeded memory of run time limits. Snakemake waits forever for these jobs.
Right now, Snakemake schedules jobs on the cluster by running a LSF scheduling command. When the job is almost finished, it writes a flag-file to disk, which tells Snakemake the jobs has finished.
This approach is brittle because it cannot detect cases in which the cluster kills the job due to exceeded memory of run time limits. Snakemake waits forever for these jobs.