-
Notifications
You must be signed in to change notification settings - Fork 59
Description
Hi! I noticed recently when I had a job fail, and I went looking for the reason, that the cli_log
:
miniwdl/WDL/runtime/backend/cli_subprocess.py
Lines 70 to 71 in 6ecfc6f
cli_log_filename = os.path.join(self.host_dir, f"{self.cli_name}.log.txt") | |
cli_log = cleanup.enter_context(open(cli_log_filename, "wb")) |
is overwritten by default. Here is the current state of a workflow:
$ ls analysis/02-lrrnaseq/20250616_095354_LongReadRnaSeq/call-flairCollapse/std*
analysis/02-lrrnaseq/20250616_095354_LongReadRnaSeq/call-flairCollapse/stderr2.txt
analysis/02-lrrnaseq/20250616_095354_LongReadRnaSeq/call-flairCollapse/stdout2.txt
analysis/02-lrrnaseq/20250616_095354_LongReadRnaSeq/call-flairCollapse/stderr.txt
analysis/02-lrrnaseq/20250616_095354_LongReadRnaSeq/call-flairCollapse/stdout.txt
$ ls analysis/02-lrrnaseq/20250616_095354_LongReadRnaSeq/call-flairCollapse/slurm_singularity*
analysis/02-lrrnaseq/20250616_095354_LongReadRnaSeq/call-flairCollapse/slurm_singularity.log.txt
$ cat analysis/02-lrrnaseq/20250616_095354_LongReadRnaSeq/call-flairCollapse/slurm_singularity.log.txt
srun: job 21258585 queued and waiting for resources
$
It would be nice if the cli_log could either be appended to, or use a numeric suffix like
Here's the relevant portion of my miniwdl.cfg
:
[task_runtime]
defaults = {
"maxRetries": 2,
"threads": 4,
"docker": "ubuntu:20.04"
}
It is indeed slurm (via the https://github.com/miniwdl-ext/miniwdl-slurm) and I could find out why the job failed via my DRM, but in general it would be nice to not overwrite the log.
Perhaps this is as simple as change w
for a
in the logger, but perhaps there's a nice way to do the numerical append like with stdout/err so I can keep a good association between them
Thanks for the consideration
Metadata
Metadata
Assignees
Labels
Type
Projects
Status