Semibin error exit status 137

[bioinfo-arctic]

Description of the bug

I am seeing repeated failures in:

NFCORE_MAG:MAG:BINNING:SEMIBIN_SINGLEEASYBIN

with exit status 137. The task stderr shows only "Killed", and dmesg confirms it is being OOM-killed inside the Docker memory cgroup.

Examples from dmesg:

Memory cgroup out of memory: Killed process 1739457 (SemiBin2) total-vm:472319276kB, anon-rss:36381296kB, file-rss:113872kB, shmem-rss:0kB, UID:1007 pgtables:145912kB oom_score_adj:0

Memory cgroup out of memory: Killed process 3556228 (SemiBin2) total-vm:477333324kB, anon-rss:104084912kB, file-rss:112924kB, shmem-rss:0kB, UID:1007 pgtables:304708kB oom_score_adj:0

Memory cgroup out of memory: Killed process 2722988 (SemiBin2) total-vm:473160272kB, anon-rss:364931412kB, file-rss:113376kB, shmem-rss:0kB, UID:1007 pgtables:812484kB oom_score_adj:0

I set:

withName: 'NFCORE_MAG:MAG:BINNING:SEMIBIN_SINGLEEASYBIN' {
cpus = 8
memory = 360.GB
time = 96.h
}

but inspecting .command.run shows the container was actually launched with:

--memory 147456m

which is ~144 GB, not 360 GB. So it looks like the process-specific memory request is being capped somewhere before Docker launch.

Command used and terminal output

Relevant files

No response

System information

No response

[dialvarezs]

Hi @bioinfo-arctic, thanks for reporting this. But to me doesn't seem like a mag bug.
In the capture attached I see --memory 368640m, so in that job the memory setting was correct.

Did you try running the pipeline with Singularity instead? That could confirm that it's a real memory bottleneck (or a Docker issue).