Tommi,
SLURM does not offer a direct analog to the command that lists the number of
nodes currently reserved for the top priority job. However, there are a couple
SLURM commands that convey why a user's job may not be running when there are
idle nodes.
The first is contribs/sjstat. It starts off with a summary of the idle (free)
nodes for each partition and then invokes squeue using the default sorting
order such that the highest priority pending job will be listed first. From
that, users can see that there are x nodes available for a given partition and
the top priority job (in each partition) is waiting for y nodes to become
available.
The other command is srun using the --test-only flag followed by either a
pending job id or a set of job specifications. However no job is actually
submitted; srun --test-only returns an estimate of when the job or a job with
the specified arguments will/would be scheduled to run.
Don
-----Original Message-----
From: [email protected] [mailto:[email protected]] On
Behalf Of Tommi T
Sent: Friday, April 01, 2011 5:25 AM
To: [email protected]
Subject: [slurm-dev] Reserved node count
Hello
Is is possible to get information about how many nodes are reserved for big
jobs? Users are sometimes confused when there is a couple of free/idle nodes
available but their jobs go to queue instead of that "free node".
LSF has a command bhosts which will show overall status of nodes in one line:
HOST_NAME STATUS JL/U MAX NJOBS RUN SSUSP USUSP RSV
lsfhost ok - 2048 1996 1996 0 0 8
BR,
Tommi
