r

slurm
®
sacctmgr - View and modify account information.
--mem=<MB> Memory required per node.
Options:
--mem-per-cpu=<MB> Memory required per
workload manager
--immediate Commit changes immediately.
allocated CPU.
--parseable Output delimited by 'I'
-N<minnodes[-maxnodes]> Node count required for the
Job Submission job.
salloc - Obtain a job allocation. Commands:
sbatch - Submit a batch script for later execution. -n<count> Number oftasks to be add <ENTITY> <SPECS> Add an entity. Identical to
launched. the create command.
srun - Obtain a job allocation (as needed) and execute an create <ENTITY> <SPECS>
application. --nodelist=<names> Specific host names to
delete <ENTITY> where Delete the specified entities.
include in job allocation.
--array=<indexes> Job array specification. <SPECS>
(e.g. "--array= l-10") (sbatch command only) --output=<name> File in which to store job
list <ENTITY> [<SPECS>] Display information about
output.
the specific entity.
--account=<name> Account to be charged for
resources used. --partition=<names> Partition/queue in which to
modify <ENTITY> where Modify an entity.
run the job.
<SPECS>set<SPECS>
--begin=<time> Initiate job after specified
--qos=<name> Quality OfService.
(e.g. "--begin= IS:00:00") time.
Entities:
--signal=[B:]<num>[@time] Signal job when approaching
--clusters=<name> Cluster(s) to run the job. time limit. account Account associated with job.
(sbatch command only)
--time=<time> Wall clock time limit. cluster ClusterName parameter in the
--constraint=<features> Required node features. slurm.conf.
--wrap=<command_string> Wrap specified command in a qos Quality ofService.
--cpus-per-task=<count> Number ofCPUs required simple "sh" shell.
per task. (sbatch command only) user User name in system.
--dependency=<state:jobid> Defer job until specified jobs
reach specified state.
Accounting Job Manai:ement
--error=<filename> File in which to store job sbcast - Transfer file to a job's compute nodes.
error messages. sacct - Display accounting data.

--exclude=<names> Specific host names to sbcast [options] SOURCE DESTINATION

exclude from job allocation. --allusers Displays all users jobs.
--force Replace previously existing file.
--exclusive[=user] Allocated nodes can not be --accounts=<name> Displays jobs with specified
shared with other jobs/users. accounts. --preserve Preserve modification times, access times, and
access permissions.
--export=<name[=value]> Export identified --endtime=<time> End ofreporting period.
environment variables. --format=<spec> Format output. seance) - Signal jobs, job arrays, and/or job steps.
--gres=<name[:count]> Generic resources required --name=<jobname> Display jobs that have any ofthese
per node. name(s). --account=<name> Operate only on jobs charging the
--input=<name> File from which to read job specified account.
--partition=<names> Comma separated list ofpartitions
input data. to select jobs and job steps from. --name=<name> Operate only on jobs with specified
--job-name=<name> Job name. name.
--state=<state_list> Display jobs with specified states.
--label Prepend task ID to output. --partition=<names> Operate only on jobs in the specified
--starttime=<time> Start ofreporting period.
(srun command only) partition/queue.

--licenses=<name[:count]>

\.'-
License resources required
for entire job. SchedMD
Slurm Support and Development
--qos=<name> Operate only on jobs using the
specified quality ofservice.
--reservation=<name> Operate only on jobs using the --format=<spec> Output format to display. SLURM- ARRAY- TASK- ID Set to the task ID if part of
specified reservation. a job array.
--iterate=<seconds> Print the state at specified interval.
--state=<names> Operate only on jobs in the specified
--long Print more detailed information. SLURM- CLUSTER- NAME Name of the cluster
state.
executing the job.
--user=<name> Operate only on jobs from the --Node Print information in a node-oriented
format. SLURM- CPUS- PER- TASK Number of CPUs requested
specified user.
per task.
--nodelist=<names> Operate only on jobs using the --partition=<names> View only specified partitions.
SLURM JOB ACCOUNT Account name.
specified compute nodes. --reservation Display information about advanced
reservations. SLURM JOB ID Job ID.
squeue - View information about jobs. -R Display reasons nodes are in the SLURM JOB NAME Job Name.
down, drained, fail or failing state.
SLURM JOB NODELIST Names of nodes allocated
--account=<name> View only jobs with specified --state=<names> View only nodes specified states. to job.
accounts.
SLURM- JOB- NUM- NODES Number of nodes allocated
--clusters=<name> View jobs on specified clusters. to job.
scontrol - Used view and modify configuration and state.
--format=<spec> Output format to display. Also see the sview graphical user interface version. SLURM JOB PARTITION Partition/queue running the
(e.g. "--format=¾i ¾j") Specify fields, size, order, etc. job.

--jobs<job_id_list> Comma separated list of job IDs --details Make show command print more details. SLURM JOB um User ID of the job's owner.
to display.
--one liner Print information on one line. SLURM JOB USER User name of the job's
--name=<name> View only jobs with specified owner.
names.
Commands: SLURM- RESTART- COUNT Number of times job has
--partition=<names> View only jobs in specified restarted.
create SPECIFICATION Create a new partition or
partitions.
SLURM PROCID Task ID (MPI rank).
--priority Sort jobs by priority.
delete SPECIFICATION Delete the entry with the SLURM STEP ID Job step ID.
--qos=<name> View only jobs with specified specified SPECIFICATION
Qualities Of Service. SLURM- STEP- NUM- TASKS Task count (number of
reconfigure All Slurm daemons will re-read MPiranks).
--start Report the expected start time the configuration file.
and resources to be allocated for
pending jobs in order of requeue JOB_LIST Requeue a running, suspended or Daemons
increasing start time. completed batch job.
slurmctld Executes on cluster's "head" node to
--state=<names> View only jobs with specified show ENTITY ID Display the state of the specified manage workload.
states. entity with the specified
identification slurmd Executes on each compute node to
--users=<names> View only jobs for specified locally manage resources.
users. update SPECIFICATION Update job, step, node, partition,
or reservation configuration per slurmdbd Manages database of resources limits,
the supplied specification. licenses, and archives accounting
sinfo - View information about nodes and partitions. records.

--all Display information about all

partitions.
Environment Variables
SLURM- ARRAY- JOB- ID Set to the job ID if part of a
SchedMD slurm®
Slurm Support and Development workload manager
--dead If set, only report state information job array.
Copyright 2017 SchedMD LLC. All rights reserved.
for non- responding (dead) nodes. http://www.schedmd.com