Hi,
From the perspective of the job, those directories are mapped to /tmp (and
others, depending on your job_container.conf). There's no need for the user to
be aware of the basepath that is specified in the conf file.
You can easily verify it is working by writing files to /tmp from a new slurm
Hi,
I was wondering if some light could be shed as to why permissions on the
mounted /tmp need to be 700 when using the job_container/tmpfs plugin.
I'm asking this because we are considering allowing for users to run rootless
apptainer as a method for building containers from a slurm job, and I'
The proposed solution will break for values >=1000G.
As sinfo is apparently stuck with megabytes, you will have to do something like:
sinfo -o "%n %m %e %C" | awk '$3 ~ /[0-9]+/ {printf "%s %iG %iG %s\n", $1, $2 /
1024, $3 / 1024, $4}'
numfmt is another option, but it is only with newer version
Hi Bill,
Your best bet is probably /var/log/slurmctld on the server that is acting as
active controller.
Best,
--
Roberto P. Monti
DevOps Engineer I
roberto.mo...@jax.org
The Jackson Laboratory
United States | China | Japan
www.jax.org
From: slurm-users On Behalf Of Bill
Benedetto
Sent: Wedn
